1
00:00:00,000 --> 00:00:03,493
Audrow Nash: Hey there and
welcome to the Audrow Nash

2
00:00:03,493 --> 00:00:07,250
podcast. Today we're diving into
the fascinating world of

3
00:00:07,250 --> 00:00:11,271
robotics with a guy who's making
robot arms work in the most.

4
00:00:11,271 --> 00:00:15,028
Unexpected places from the vast
expanses of space to this

5
00:00:15,028 --> 00:00:18,851
strawberry field right here on
Earth. I sat down with Dave

6
00:00:18,851 --> 00:00:22,608
Coleman, the Chief Product
Officer and founder of PickNik

7
00:00:22,608 --> 00:00:26,299
Robotics. Now, Dave's not just
any robotics expert, he's

8
00:00:26,299 --> 00:00:29,858
someone who's been deeply
involved in the ROS or robot

9
00:00:29,858 --> 00:00:34,076
operating system community, and
his company has become a pivotal

10
00:00:34,076 --> 00:00:37,833
player in helping over 110
companies harness the power of

11
00:00:37,833 --> 00:00:41,590
robotic arms for diverse
applications. If you are curious

12
00:00:41,590 --> 00:00:45,018
about how robotics can
revolutionize industries like

13
00:00:45,018 --> 00:00:48,972
agriculture, aerospace, or even
your local restaurant, or if

14
00:00:48,972 --> 00:00:53,059
you're just a fan of innovative
tech, this episode is for you.

15
00:00:53,059 --> 00:00:57,343
So buckle up. Grab your favorite
snack and let's explore how Dave

16
00:00:57,343 --> 00:01:01,298
and PickNik are pushing the
boundaries of what robots can do

17
00:01:01,298 --> 00:01:04,000
in our everyday lives. Let's get
started.

18
00:01:04,000 --> 00:01:07,000
Audrow Nash: Hi Dave. Would you
introduce yourself?

19
00:01:07,000 --> 00:01:11,276
Dave Coleman: Hey, this is Dave
Coleman. I am at PickNik

20
00:01:11,276 --> 00:01:16,011
Robotics, and I've been, a long
time, ROS Advocate and someone

21
00:01:16,011 --> 00:01:20,669
in the ROS e or the robotics
ecosystem. I got my PhD in Robot

22
00:01:20,669 --> 00:01:24,945
motion planning, back in 2017
from CU Boulder. and yeah,

23
00:01:24,945 --> 00:01:28,000
that's, I guess that's me in a
nutshell.

24
00:01:28,000 --> 00:01:31,620
Audrow Nash: Now tell me about
PickNik. like some of the

25
00:01:31,620 --> 00:01:35,563
details just for context of the
company. So you said you guys

26
00:01:35,563 --> 00:01:39,312
are in Boulder. How many people
are you, when did you guys

27
00:01:39,312 --> 00:01:43,255
start? and then like high level
of the business model. So how

28
00:01:43,255 --> 00:01:45,000
are you guys getting money,

29
00:01:45,000 --> 00:01:49,235
Dave Coleman: Yeah, I'm, really
glad to be, being interviewed by

30
00:01:49,235 --> 00:01:53,271
you again. This is, I love this
every couple years. so you've

31
00:01:53,271 --> 00:01:57,439
heard this before, but PickNik,
we're, we're a delayed spin out

32
00:01:57,439 --> 00:02:01,343
of Willow Garage in some ways.
There was certainly a couple

33
00:02:01,343 --> 00:02:05,379
years gap, where I was still in
grad school, but, essentially

34
00:02:05,379 --> 00:02:09,614
Willow Garage back in, in Menlo
Park, Silicon Valley started the

35
00:02:09,614 --> 00:02:13,650
robotic operating system. And
one of the projects was MoveIt.

36
00:02:13,650 --> 00:02:17,553
And I was, actually an intern
back then helping develop the

37
00:02:17,553 --> 00:02:21,457
first version of MoveIt. And
after Willow closed, it was in

38
00:02:21,457 --> 00:02:25,427
the hands of SRI for a minute.
but that got, it, got dropped

39
00:02:25,427 --> 00:02:29,265
from there. And then, I picked
it up on behalf of, PickNik

40
00:02:29,265 --> 00:02:33,168
Robotics leading this, open
source project. And, from there

41
00:02:33,168 --> 00:02:37,006
we've just, we've grown a lot.
We've helped over a hundred

42
00:02:37,006 --> 00:02:40,513
companies now, maybe 110,
probably, just do more with

43
00:02:40,513 --> 00:02:44,350
robot arms. It's deceptively
hard using robotic arms for a

44
00:02:44,350 --> 00:02:48,320
lot of classes of applications
beyond traditional industrial

45
00:02:48,320 --> 00:02:52,224
automation. and so we've been
through a bunch of iterations

46
00:02:52,224 --> 00:02:56,128
since, we're at 30 people now
have gone up and down. but we

47
00:02:56,128 --> 00:02:59,767
got just a great team of
engineers, PhDs who understand

48
00:02:59,767 --> 00:03:03,869
how to combine motion control,
kinematics computer vision, and

49
00:03:03,869 --> 00:03:07,574
then most recently machine
learning into a comprehensive

50
00:03:07,574 --> 00:03:11,478
solution that, just a lot of
companies really benefit from.

51
00:03:11,478 --> 00:03:13,000
And, we love partnering

52
00:03:13,000 --> 00:03:14,000
Audrow Nash: Hell

53
00:03:13,000 --> 00:03:15,000
Dave Coleman: with other
robotics companies.

54
00:03:15,000 --> 00:03:16,000
Audrow Nash: Awesome.

55
00:03:16,000 --> 00:03:19,515
Audrow Nash: Now you say, tell
me some of the ways that,

56
00:03:19,515 --> 00:03:23,280
manipulation is hard. what are
some of these challenges that

57
00:03:23,280 --> 00:03:26,795
are not obvious for why
manipulation may be so hard that

58
00:03:26,795 --> 00:03:30,059
you guys are working on very
often and helping these

59
00:03:30,059 --> 00:03:31,000
companies with?

60
00:03:31,000 --> 00:03:34,634
Dave Coleman: Yeah, so part of
my theory on the industry of,

61
00:03:34,634 --> 00:03:38,328
robotic arms is it started with
maybe like the seventies with

62
00:03:38,328 --> 00:03:41,961
automotive, and then it grew
into general manufacturing. And

63
00:03:41,961 --> 00:03:45,655
so a lot of the incumbents have
gotten really good at serving

64
00:03:45,655 --> 00:03:49,289
those needs. And those needs
don't actually align with a lot

65
00:03:49,289 --> 00:03:52,680
of the newer, more modern
applications of robot arms. We

66
00:03:52,680 --> 00:03:56,314
basically wanna get robot arms,
into every part of our life.

67
00:03:56,314 --> 00:03:59,887
Small businesses, restaurants,
your, home. we want it to be

68
00:03:59,887 --> 00:04:03,460
outdoors and farm fields. We
want it to be in hospitals, in

69
00:04:03,460 --> 00:04:07,275
space, building our next, lunar
colony, which is the thing that

70
00:04:07,275 --> 00:04:10,000
we actually are under contract
with NASA for,

71
00:04:10,000 --> 00:04:11,000
Audrow Nash: Super cool.

72
00:04:11,000 --> 00:04:15,182
Dave Coleman: subsea, supporting
like infrastructure for oil and

73
00:04:15,182 --> 00:04:18,711
gas, un under water. Just
there's so many areas that I

74
00:04:18,711 --> 00:04:22,436
think people forget about and
they just think, we have an

75
00:04:22,436 --> 00:04:25,704
assembly line, we have a
conveyor belt, and we're,

76
00:04:25,704 --> 00:04:29,625
palletizing. We are, machine
tending. We are a welding parts

77
00:04:29,625 --> 00:04:33,219
of a car. And, so the, these
incumbents I referred to a

78
00:04:33,219 --> 00:04:37,009
minute ago, they've done a
really good job of serving this

79
00:04:37,009 --> 00:04:40,538
kind of more blue collar
warehouse, like factory space

80
00:04:40,538 --> 00:04:44,459
with, the classic stuff that
people always talk about, teach

81
00:04:44,459 --> 00:04:48,576
pendants and, PLCs and all those
things. And of course, Audrow,

82
00:04:48,576 --> 00:04:52,562
you know a ton about this. You
work for a similar company. So

83
00:04:52,562 --> 00:04:54,000
I'll acknowledge that.

84
00:04:54,000 --> 00:04:57,442
Dave Coleman: But we, actually,
a, term that we borrowed from

85
00:04:57,442 --> 00:05:00,997
academia and that we are trying
to push forward is unstructured

86
00:05:00,997 --> 00:05:04,382
robotics is the niche that we're
looking at. So I think most

87
00:05:04,382 --> 00:05:07,711
robot companies today, they're
still looking at the biggest

88
00:05:07,711 --> 00:05:11,266
market, which is manufacturing
and logistics. And, we certainly

89
00:05:11,266 --> 00:05:14,764
have customers in that space,
that we're working with. you can

90
00:05:14,764 --> 00:05:17,924
see some of the logos on our
website, but our main focus

91
00:05:17,924 --> 00:05:21,478
really is yeah, these, next gen
applications. like we had a big

92
00:05:21,478 --> 00:05:24,807
program picking strawberries
outdoors in California, in the

93
00:05:24,807 --> 00:05:27,967
farm fields. That was a really
rewarding process because

94
00:05:27,967 --> 00:05:31,409
there's a tractor driving and
arms hanging off and things are

95
00:05:31,409 --> 00:05:35,020
very dynamic. You got the, dirt
bumping and even the seasonality

96
00:05:35,020 --> 00:05:38,349
of the, shadows from the sun
changed throughout the season,

97
00:05:38,349 --> 00:05:41,622
as the earth rotates. and then
all those things, they have

98
00:05:41,622 --> 00:05:44,838
these big impacts on your
machine learning models and you

99
00:05:44,838 --> 00:05:48,223
have this like data drift. and
that's just like one, one use

100
00:05:48,223 --> 00:05:51,552
case. It's pretty wild the
number of things that we've been

101
00:05:51,552 --> 00:05:54,712
able to help with. So we're
really industry agnostic. we

102
00:05:54,712 --> 00:05:58,041
just like hard problems that
maybe aren't so served by hard

103
00:05:58,041 --> 00:05:59,000
coding waypoints.

104
00:05:59,000 --> 00:06:02,735
Audrow Nash: And what, so the
traditional factories that we

105
00:06:02,735 --> 00:06:06,218
think of when we think of
robotic arms, where you, they

106
00:06:06,218 --> 00:06:09,953
pick up huge metal and they
repeat the same motion over and

107
00:06:09,953 --> 00:06:13,815
over again. so that's not the
use case you're serving. You're

108
00:06:13,815 --> 00:06:16,727
serving unstructured
environments where things

109
00:06:16,727 --> 00:06:20,653
change, the world changes. the
strawberry is not always in the

110
00:06:20,653 --> 00:06:23,818
same position. These kinds of
things. What, in the

111
00:06:23,818 --> 00:06:27,680
unstructured space of robotics
applications, what are some of

112
00:06:27,680 --> 00:06:31,416
the like classes of problems
that you see? So you mentioned

113
00:06:31,416 --> 00:06:35,341
the environment changes, like
the sun moves over the course of

114
00:06:35,341 --> 00:06:38,000
the year. what are some of the
other ones?

115
00:06:38,000 --> 00:06:42,308
Dave Coleman: I, I wanna clarify
there, there are actually still

116
00:06:42,308 --> 00:06:45,000
plenty of hard problems in
manufacturing

117
00:06:45,000 --> 00:06:46,000
Audrow Nash: Oh, definitely

118
00:06:46,000 --> 00:06:50,069
Dave Coleman: That are barely
solved or just starting to be

119
00:06:50,069 --> 00:06:54,276
solved, like wiring harnesses,
for example, or assembly is a,

120
00:06:54,276 --> 00:06:58,276
big one. or just, what is it?
Low high mix. Low yield, low

121
00:06:58,276 --> 00:07:00,000
volume manufacturing like

122
00:07:00,000 --> 00:07:01,000
Audrow Nash: Yeah. High mix,
high volume kind

123
00:07:01,000 --> 00:07:04,065
Dave Coleman: Yeah, high mix,
high volume. Thank you. Thank

124
00:07:04,065 --> 00:07:05,000
you. and so again,

125
00:07:05,000 --> 00:07:06,000
Audrow Nash: area.

126
00:07:06,000 --> 00:07:09,813
Dave Coleman: we do get pulled
into some of those. a big area

127
00:07:09,813 --> 00:07:13,250
we've been helping out with
lately is in, in big spaces

128
00:07:13,250 --> 00:07:17,063
where you don't want fixed work
cells, but you actually want,

129
00:07:17,063 --> 00:07:20,875
like mobile robots, an example
of a hanger for, aerospace for

130
00:07:20,875 --> 00:07:22,000
like large planes.

131
00:07:22,000 --> 00:07:23,000
Audrow Nash: Hmm.

132
00:07:23,000 --> 00:07:26,224
Dave Coleman: you want actually
robots that can drive around and

133
00:07:26,224 --> 00:07:29,297
when you have a robot with an
arm on it, suddenly you have to

134
00:07:29,297 --> 00:07:32,370
deal with a lot of calibration
issues and understanding where

135
00:07:32,370 --> 00:07:35,393
you are with the respect to the
work piece and the work and,

136
00:07:35,393 --> 00:07:38,416
the, the tool that needs to be.
So there's a lot of scan and

137
00:07:38,416 --> 00:07:41,640
plan. I, do think I missed your
original question, Audrow. That,

138
00:07:41,640 --> 00:07:43,000
that it, what was it again?

139
00:07:43,000 --> 00:07:47,747
Audrow Nash: The classes of
problems that you may encounter

140
00:07:47,747 --> 00:07:50,000
in a structured environment,

141
00:07:50,000 --> 00:07:53,000
Dave Coleman: And you mean
unstructured?

142
00:07:53,000 --> 00:07:54,000
Audrow Nash: unstructured Yes.

143
00:07:53,000 --> 00:07:57,000
Dave Coleman: Yeah.

144
00:07:57,000 --> 00:07:59,889
Audrow Nash: or in a few
categories of problems that

145
00:07:59,889 --> 00:08:01,000
you're often helping

146
00:08:01,000 --> 00:08:04,041
Dave Coleman: there's one
category. It's just pick and

147
00:08:04,041 --> 00:08:07,025
place, which can be a very
simple thing if you have a

148
00:08:07,025 --> 00:08:10,629
conveyor belt and you make sure
that all the parts are coming in

149
00:08:10,629 --> 00:08:14,120
in the same exact position. But
that is such a, huge term that

150
00:08:14,120 --> 00:08:17,667
we have to really, when someone
says we won't pick in place, we

151
00:08:17,667 --> 00:08:21,271
have to ask 'em like, what kind?
'cause okay, let's say we focus

152
00:08:21,271 --> 00:08:24,650
on, bin picking, and you have a,
a bin of parts. And this is

153
00:08:24,650 --> 00:08:27,859
again, more like warehouse
logistics. It, there's just so

154
00:08:27,859 --> 00:08:31,463
much variety to even that one of
like, how chaotic are the parts

155
00:08:31,463 --> 00:08:34,842
in the bin? how exact do you
need to place? Because a lot of

156
00:08:34,842 --> 00:08:38,107
times people just want pick and
drop. That's a lot easier.

157
00:08:38,107 --> 00:08:41,542
'cause you don't have to worry
about like where you've picked

158
00:08:41,542 --> 00:08:44,639
it in your hand. and then
there's the question of okay,

159
00:08:44,639 --> 00:08:48,187
the, if you maybe have various
products that you're fulfilling,

160
00:08:48,187 --> 00:08:51,058
let's say your Amazon
warehouses, how many SKUs are

161
00:08:51,058 --> 00:08:54,381
you supporting? If it's like a
single SKU bin, that's a lot

162
00:08:54,381 --> 00:08:55,000
easier than

163
00:08:55,000 --> 00:08:58,000
Audrow Nash: SKU is like an
inventory code, right?

164
00:08:58,000 --> 00:09:00,426
Dave Coleman: Correct. Yeah.
Every industry has their own

165
00:09:00,426 --> 00:09:02,000
terms. But SKUs a, common term
for a,

166
00:09:02,000 --> 00:09:03,000
Audrow Nash: I have no idea

167
00:09:03,000 --> 00:09:04,000
Dave Coleman: if there's a

168
00:09:03,000 --> 00:09:05,290
Audrow Nash: for, but I know
what it means. Do you know what

169
00:09:05,290 --> 00:09:07,504
it stands for? I've asked in
past interviews and I haven't

170
00:09:07,504 --> 00:09:08,000
looked it up.

171
00:09:08,000 --> 00:09:10,000
Dave Coleman: Nope.

172
00:09:10,000 --> 00:09:13,372
Audrow Nash: Okay, so you have
different skews and the SKU

173
00:09:13,372 --> 00:09:15,000
might be a different object,

174
00:09:15,000 --> 00:09:16,000
Dave Coleman: yeah. So so

175
00:09:16,000 --> 00:09:18,000
Audrow Nash: is a different
object.

176
00:09:18,000 --> 00:09:21,767
Dave Coleman: yeah. So that's an
example of a thing we might do,

177
00:09:21,767 --> 00:09:25,416
but In other environments, you
want to drive up to a table and

178
00:09:25,416 --> 00:09:29,125
move a tray from one work, like,
automation islands. I've heard

179
00:09:29,125 --> 00:09:32,951
of this term like where you have
different machines and you wanna

180
00:09:32,951 --> 00:09:36,482
do this machine tending, and
that's a different kind of like

181
00:09:36,482 --> 00:09:40,190
pick and place. 'cause now it's
a little more structured. Maybe

182
00:09:40,190 --> 00:09:43,899
this tray has little slots where
you put each product in and it

183
00:09:43,899 --> 00:09:47,666
kind of slots in, or the machine
needs a certain format of that.

184
00:09:47,666 --> 00:09:51,138
So there's just such a wide
variety. and then place itself.

185
00:09:51,138 --> 00:09:54,670
if you're just placing onto a,
flat surface, that's easy. If

186
00:09:54,670 --> 00:09:58,025
you're inserting it into a
particular, like a pe and hole

187
00:09:58,025 --> 00:10:01,616
type of problem, like you're
having to pull in all this force

188
00:10:01,616 --> 00:10:05,324
torque control to make sure that
you exactly get it aligned. if

189
00:10:05,324 --> 00:10:08,384
you're assembling, maybe you
need to hold and have a

190
00:10:08,384 --> 00:10:12,034
secondary arm, attach it with
some kind of welding or screwing

191
00:10:12,034 --> 00:10:15,448
operation. So there's just such
a wide variety and, we're,

192
00:10:15,448 --> 00:10:18,685
taking on the really hard
problem and. You probably are

193
00:10:18,685 --> 00:10:22,452
familiar with this problem too,
of a generic toolbox that can do

194
00:10:22,452 --> 00:10:26,160
all of these things, regardless
of industry. So like we say yes

195
00:10:26,160 --> 00:10:29,692
to the hard stuff, bring it on,
we'll, take it. And a lot of

196
00:10:29,692 --> 00:10:33,400
robot companies, they, have a,
arguably this is a strength, but

197
00:10:33,400 --> 00:10:37,049
they're really focused on one
thing and doing that, doing that

198
00:10:37,049 --> 00:10:40,758
well, and we're, going after the
broad stuff. So when there are

199
00:10:40,758 --> 00:10:44,348
things that haven't really been
solved before, a little more,

200
00:10:44,348 --> 00:10:47,703
more novel, that's the stuff
that we get excited about or

201
00:10:47,703 --> 00:10:50,823
more dynamic or changing
environments or has a lot of

202
00:10:50,823 --> 00:10:52,000
sequences and steps.

203
00:10:52,000 --> 00:10:55,859
Audrow Nash: That sounds really
awesome. So it sounds fun to

204
00:10:55,859 --> 00:10:59,396
work on all the different
diverse, challenging problems

205
00:10:59,396 --> 00:11:03,320
that you might do. How do you,
balance between that and being

206
00:11:03,320 --> 00:11:06,921
spread too thin? Because you
were saying it's arguably a

207
00:11:06,921 --> 00:11:10,587
strength to just focus on one
thing. but I think there is

208
00:11:10,587 --> 00:11:14,317
strength in being very general,
and maybe there's a larger

209
00:11:14,317 --> 00:11:18,305
market if you're more general.
but tell me about that. Tell me

210
00:11:18,305 --> 00:11:22,228
about being more general, and
how do you counter becoming too

211
00:11:22,228 --> 00:11:23,000
spread thin?

212
00:11:23,000 --> 00:11:26,677
Dave Coleman: Yeah, we've been
doing it for a long time and so

213
00:11:26,677 --> 00:11:30,294
there's just a lot of momentum.
We've and, software and tools

214
00:11:30,294 --> 00:11:33,733
we've developed. We're also
like. we're, pretty pro, using

215
00:11:33,733 --> 00:11:36,936
open source, so we're big
contributors to the ROT ROS,

216
00:11:36,936 --> 00:11:40,316
like I mentioned earlier,
MoveIt, the open source version

217
00:11:40,316 --> 00:11:43,696
is out there. We, we have
developed a, new product called

218
00:11:43,696 --> 00:11:47,372
MoveIt Pro, which I'm sure we'll
be talking about, but because

219
00:11:47,372 --> 00:11:51,168
we've been around for basically
10 years now, we've, done this a

220
00:11:51,168 --> 00:11:54,844
lot. We've seen it with a lot of
customers, and so we've taken

221
00:11:54,844 --> 00:11:58,343
that experience, and our work
with, a lot of big government

222
00:11:58,343 --> 00:12:01,901
agencies like NASA and Space
Force, and we've just built out

223
00:12:01,901 --> 00:12:05,340
this. We, we've already built
most of this, like we have a

224
00:12:05,340 --> 00:12:08,780
huge library. we're counting
like, over 160, and that's, I

225
00:12:08,780 --> 00:12:12,219
think, a conservative count
because some of them maybe are

226
00:12:12,219 --> 00:12:15,837
more like our customers. We've
built something custom for our

227
00:12:15,837 --> 00:12:19,573
customers, but, at least 160 of
these skills and behaviors that

228
00:12:19,573 --> 00:12:23,012
are, available. it's always
growing. And so I just, I look

229
00:12:23,012 --> 00:12:26,748
at it as, I think the robotics
industry, it needs a more common

230
00:12:26,748 --> 00:12:30,484
platform so that we don't keep
reinventing the wheel. I imagine

231
00:12:30,484 --> 00:12:34,220
this is, again, similar to what
you're, doing in Intrinsic, but

232
00:12:34,220 --> 00:12:37,659
I look for inspiration often at
the video gaming industry,

233
00:12:37,659 --> 00:12:41,336
actually. I. I, I've never been
a video game developer myself,

234
00:12:41,336 --> 00:12:44,834
but of course they have video
gaming engines. And when that

235
00:12:44,834 --> 00:12:47,977
was first came out, my
understanding from reading the

236
00:12:47,977 --> 00:12:51,773
history of this, it was, novel.
It was, most shops were building

237
00:12:51,773 --> 00:12:55,271
the whole 3D graphics, ray
tracing, shading, all that stuff

238
00:12:55,271 --> 00:12:58,889
they're building each time on
their own. and some groups like

239
00:12:58,889 --> 00:13:02,328
Unreal Unity came along and
said, why are we, having to do

240
00:13:02,328 --> 00:13:06,005
this? Is this causing way too
much costs? And it's a long time

241
00:13:06,005 --> 00:13:09,622
to get a new video game title to
market? And of course, these

242
00:13:09,622 --> 00:13:13,061
shops had their own code that
they were using for multiple

243
00:13:13,061 --> 00:13:16,738
games. But, the, emergence of
these, big platforms to speed up

244
00:13:16,738 --> 00:13:20,474
and improve the user experience
and the tooling and, make it so

245
00:13:20,474 --> 00:13:23,498
you can focus on your end
application, whether it's

246
00:13:23,498 --> 00:13:27,056
certain storyline for a video
game or it's a certain kind of

247
00:13:27,056 --> 00:13:30,733
con combination of mobility and
robotic arms and manipulation.

248
00:13:30,733 --> 00:13:34,410
And Yeah, that's what we have
available now, and we have a lot

249
00:13:34,410 --> 00:13:38,027
of successful case studies now
of customers using our, I like

250
00:13:38,027 --> 00:13:41,703
to call it like a manipulation
engine. It's a, it's equivalent

251
00:13:41,703 --> 00:13:42,000
to a.

252
00:13:42,000 --> 00:13:45,000
Audrow Nash: Cool.

253
00:13:45,000 --> 00:13:48,250
Audrow Nash: Okay, so
Manipulation engine, how does,

254
00:13:48,250 --> 00:13:52,063
so I'm curious, how does that
fit with MoveIt Pro and can you

255
00:13:52,063 --> 00:13:54,000
tell me a bit about MoveIt Pro?

256
00:13:54,000 --> 00:13:57,761
Dave Coleman: Sure. so MoveIt
Pro is the new commercial grade

257
00:13:57,761 --> 00:14:01,522
version of, MoveIt. I'm sure
some listeners here are aware of

258
00:14:01,522 --> 00:14:05,099
this. It's move. It's very, I
don't know if you introduced

259
00:14:05,099 --> 00:14:08,860
that one. It's, very, popular in
research. My PhD thesis was,

260
00:14:08,860 --> 00:14:12,744
based on it. I contributed a lot
of, my work to it, but there's

261
00:14:12,744 --> 00:14:15,827
been a lot of other, great
researchers, more, more

262
00:14:15,827 --> 00:14:19,404
accomplished than I who have
contributed pretty incredible

263
00:14:19,404 --> 00:14:23,288
algorithms for collision check
and motion control, grasping and

264
00:14:23,288 --> 00:14:27,173
all of that. So that's available
PickNiks still maintaining it,

265
00:14:27,173 --> 00:14:30,996
and we still are contributing ex
fixes to it. Improvements. We

266
00:14:30,996 --> 00:14:35,003
just released the latest version
for the, for ROS to Jazzy, did I

267
00:14:35,003 --> 00:14:38,826
get that right? so that's out.
But MoveIt Pro is okay, this is

268
00:14:38,826 --> 00:14:42,587
move, it's for research and it's
for com. It's a benchmarking

269
00:14:42,587 --> 00:14:46,472
tool comparing algorithms that,
that was the initial design. my

270
00:14:46,472 --> 00:14:49,000
mentor, y Sukin designed it for
that. and

271
00:14:49,000 --> 00:14:51,891
Audrow Nash: mean? So
benchmarking algorithms. So you

272
00:14:51,891 --> 00:14:55,000
give them a similar task and you
see how you, how they do

273
00:14:55,000 --> 00:14:58,297
Dave Coleman: Exactly. Yeah. So
different labs, like Rice

274
00:14:58,297 --> 00:15:01,884
University has developed a lot
of kind of, they call it search

275
00:15:01,884 --> 00:15:05,181
based planning algorithms. And
those have gotten a lot of

276
00:15:05,181 --> 00:15:08,825
mileage. It's our favorite, like
maybe you've heard of RRT. but

277
00:15:08,825 --> 00:15:12,296
then other labs, did more like,
SVP Ls, a library, that came

278
00:15:12,296 --> 00:15:15,593
with a different way. And then,
of course now there's new

279
00:15:15,593 --> 00:15:18,832
machine learning approaches to
this. there are different

280
00:15:18,832 --> 00:15:21,898
optimization algorithms, but
there's a whole bunch of

281
00:15:21,898 --> 00:15:25,542
different ways of solving these
robotic arm problems. And for a

282
00:15:25,542 --> 00:15:28,839
while at least. The researchers
were really interested in

283
00:15:28,839 --> 00:15:32,541
comparing and publishing papers
saying mine's better than yours.

284
00:15:32,541 --> 00:15:36,012
And, so I, I was totally guilty
of that. I've published some

285
00:15:36,012 --> 00:15:39,367
comparison things and, MoveIt
was great for that. And then

286
00:15:39,367 --> 00:15:42,664
when you wanna actually deploy
these algorithms to harden

287
00:15:42,664 --> 00:15:45,961
reliable systems that have to
have the infrastructure and

288
00:15:45,961 --> 00:15:49,201
containerization and, like
introspection, recovery fault

289
00:15:49,201 --> 00:15:52,498
management, being able to do a
lot of complex sequencing,

290
00:15:52,498 --> 00:15:55,853
that's not what movement was
built for. And so we, and we,

291
00:15:55,853 --> 00:15:59,324
kept building all that custom
for our customers and we, so I

292
00:15:59,324 --> 00:16:02,389
said like, why, do we keep doing
this? These are like

293
00:16:02,389 --> 00:16:05,976
multimillion dollar programs.
It's taking customers, year plus

294
00:16:05,976 --> 00:16:09,389
couple years to get to market,
or sometimes they fail. They

295
00:16:09,389 --> 00:16:13,149
would underestimate how hard the
feasibility of all this was. And

296
00:16:13,149 --> 00:16:16,446
we've seen a lot of robotic
startups who try to use robot

297
00:16:16,446 --> 00:16:19,974
arms kinda shutter or have to
really pivot. or maybe they get

298
00:16:19,974 --> 00:16:23,214
acquired, which is sometimes
similar to shuttering. just

299
00:16:23,214 --> 00:16:26,626
merging. and so we, our kind of
thesis is okay, let's, stop

300
00:16:26,626 --> 00:16:29,808
wasting all of the, this
investor money on building the

301
00:16:29,808 --> 00:16:33,510
same thing over and over again.
And, we've built a, first, first

302
00:16:33,510 --> 00:16:36,981
in class application of okay,
now you just drop in all these

303
00:16:36,981 --> 00:16:40,220
different algorithms, depending
on what you need. We can

304
00:16:40,220 --> 00:16:43,864
prototype something very quickly
in like really complex physics

305
00:16:43,864 --> 00:16:47,509
simulators, and then work with
you through different phases. So

306
00:16:47,509 --> 00:16:51,153
like PickNik, we, really love to
get a quick prototype out with

307
00:16:51,153 --> 00:16:54,913
our customers, and that could be
something that's used for market

308
00:16:54,913 --> 00:16:58,499
validation, getting that pilot
out, showing investors, getting

309
00:16:58,499 --> 00:17:01,796
them excited about the vision.
And these aren't just like

310
00:17:01,796 --> 00:17:05,151
concept videos. This is like
we're using the full existing

311
00:17:05,151 --> 00:17:08,911
algorithm suite. We just haven't
done some of the like six sigma,

312
00:17:08,911 --> 00:17:12,498
like last mile reliability and
Deployment system, so we can do

313
00:17:12,498 --> 00:17:16,026
that. but yeah, that's, kinda
where we're at today. So MoveIt

314
00:17:16,026 --> 00:17:19,092
pro I think you were asking
about that. it's, both a,

315
00:17:19,092 --> 00:17:22,447
runtime and it's developer
platform. And so the runtime is

316
00:17:22,447 --> 00:17:25,513
the gaming engine part, and it
provides like the core

317
00:17:25,513 --> 00:17:28,000
algorithms and a lot of the, AI
components.

318
00:17:28,000 --> 00:17:31,789
Audrow Nash: When you, so I,
don't fully understand when you

319
00:17:31,789 --> 00:17:35,389
say it's the gaming engine part,
is, you're saying it's a

320
00:17:35,389 --> 00:17:39,368
metaphor to the gaming engine?
It's, or is it, literally has an

321
00:17:39,368 --> 00:17:43,158
engine for you to it's its own
software suite in, I, guess I

322
00:17:43,158 --> 00:17:46,000
don't understand the parts of
it, completely.

323
00:17:46,000 --> 00:17:49,793
Dave Coleman: yeah, it's the two
parts. It's the, game, the. The

324
00:17:49,793 --> 00:17:53,291
runtime is what goes in your
release robot. So it's got the

325
00:17:53,291 --> 00:17:56,610
actual, some brains. And then,
of course there's a whole

326
00:17:56,610 --> 00:18:00,107
application layer that our
customers build out. and to make

327
00:18:00,107 --> 00:18:03,604
it easy to build out their
applications. We have a bunch of

328
00:18:03,604 --> 00:18:07,457
developer tools, so it's a whole
user interface. And we're not an

329
00:18:07,457 --> 00:18:11,250
IDE, we're not trying to replace
whatever, developer environment

330
00:18:11,250 --> 00:18:14,807
you're used to for coding, so
you can use VS code or Eclipse

331
00:18:14,807 --> 00:18:17,000
or anything. but we certainly
have a,

332
00:18:17,000 --> 00:18:18,000
Audrow Nash: files. Uhhuh.

333
00:18:18,000 --> 00:18:21,252
Dave Coleman: yeah, but we
certainly have a lot of, of 3D

334
00:18:21,252 --> 00:18:24,390
visualizations where you have
multiple panes, different

335
00:18:24,390 --> 00:18:27,870
simulated camera feeds, real
life camera feeds. There's a lot

336
00:18:27,870 --> 00:18:31,064
built around like teleoperation
and recovery, so you can

337
00:18:31,064 --> 00:18:34,430
remotely help your robot out
when you're going through your

338
00:18:34,430 --> 00:18:37,796
development stages. and then
the, core of it is as behavior

339
00:18:37,796 --> 00:18:41,162
trees, which is a very popular
thing in robotics right now.

340
00:18:41,162 --> 00:18:44,471
It's kinda like a state machine,
but we just, we allow our

341
00:18:44,471 --> 00:18:47,951
customers to build out very.
Like complex, like multitask. If

342
00:18:47,951 --> 00:18:51,146
we're gonna do this and then
that we're gonna drive over

343
00:18:51,146 --> 00:18:54,284
here, we're gonna stop, we're
gonna survey, we're gonna

344
00:18:54,284 --> 00:18:57,193
understand what's in the
environment, and then pick

345
00:18:57,193 --> 00:19:00,559
something and we're gonna go
over there, put it down, we're

346
00:19:00,559 --> 00:19:03,868
gonna do a, spraying operation,
maybe a welding operation.

347
00:19:03,868 --> 00:19:07,177
That's all that's enabled by
behavior trees. And then the,

348
00:19:07,177 --> 00:19:10,714
bugging infrastructure where you
can have break points and see

349
00:19:10,714 --> 00:19:14,080
some insights about what the
robot's thinking. and maybe as

350
00:19:14,080 --> 00:19:17,046
you're developing it,
something's not working right?

351
00:19:17,046 --> 00:19:20,013
that's like a lot of the problem
of robotics is like

352
00:19:20,013 --> 00:19:23,607
understanding what the robot's
thinking is. It's a hard problem

353
00:19:23,607 --> 00:19:27,144
that like without good tooling,
it's causing our industry just

354
00:19:27,144 --> 00:19:28,000
to, to stumble.

355
00:19:28,000 --> 00:19:32,542
Audrow Nash: Yeah, for sure.
Yeah, I just interviewed Adrian

356
00:19:32,542 --> 00:19:37,311
from Fox Glove and visualization
is critical for sure. okay, so

357
00:19:37,311 --> 00:19:41,323
you have the runtime. The
runtime is the part that is

358
00:19:41,323 --> 00:19:45,789
actually running on the robot.
So that's, talking to it and

359
00:19:45,789 --> 00:19:50,558
executing your program. Then you
have tooling. That's the other

360
00:19:50,558 --> 00:19:54,948
big section. and so that makes
it easier for them to write

361
00:19:54,948 --> 00:19:59,869
their code introspect. what, and
let's see. The debugging is very

362
00:19:59,869 --> 00:20:04,259
interesting. So actually would
you take me through what it

363
00:20:04,259 --> 00:20:09,104
might look like in, in a cartoon
version of using, MoveIt Pro I.

364
00:20:09,104 --> 00:20:13,645
To do a simple application for
this? what, does it look like

365
00:20:13,645 --> 00:20:18,339
from the developer's perspective
or the company's perspective?

366
00:20:18,339 --> 00:20:22,578
How does debugging work?
introspection in general. gimme

367
00:20:22,578 --> 00:20:25,000
a cartoon of that kind of thing.

368
00:20:25,000 --> 00:20:29,245
Dave Coleman: A cartoon. so we
have, a handful of reference

369
00:20:29,245 --> 00:20:33,489
applications that we've built a
lot of different. To things

370
00:20:33,489 --> 00:20:37,878
already that you can draw from.
And so that's the quick start

371
00:20:37,878 --> 00:20:41,691
is, figure out what most close
closely resembles your

372
00:20:41,691 --> 00:20:46,151
application and then build from
there. And, that's just useful

373
00:20:46,151 --> 00:20:50,468
because this isn't, this is
still pretty low level and yeah,

374
00:20:50,468 --> 00:20:54,640
there's, there is a drag and
drop component of it that you

375
00:20:54,640 --> 00:20:59,029
can use, but we definitely, en
encourage you to make your own

376
00:20:59,029 --> 00:21:03,561
code just so you can do your own
plugins, your own, we call 'em

377
00:21:03,561 --> 00:21:05,000
behaviors or skills.

378
00:21:05,000 --> 00:21:09,515
Audrow Nash: Is it, so when
someone makes a program, or

379
00:21:09,515 --> 00:21:14,358
behavior a skill, is it like one
drop or 1 1 1 node in your

380
00:21:14,358 --> 00:21:16,000
behavior tree. Okay.

381
00:21:16,000 --> 00:21:17,000
Dave Coleman: But you can make
as many nodes as you want

382
00:21:17,000 --> 00:21:19,840
Audrow Nash: Oh yeah. You can
make your whole application from

383
00:21:19,840 --> 00:21:22,725
it, but that's the abstraction
and where you give users access.

384
00:21:22,725 --> 00:21:23,000
I see.

385
00:21:23,000 --> 00:21:26,711
Dave Coleman: Yeah, and we've
always, been very pro ROS. So

386
00:21:26,711 --> 00:21:30,610
we're actually using the, just
the ROS infrastructure. It's, a

387
00:21:30,610 --> 00:21:34,572
ROS package. It's in like the,
the ROS build system. And so you

388
00:21:34,572 --> 00:21:38,220
can easily pull in, in the Ro
ROS ecosystem. There's this,

389
00:21:38,220 --> 00:21:41,679
concept of, actions and the
services and all that's all

390
00:21:41,679 --> 00:21:43,000
built in first class.

391
00:21:43,000 --> 00:21:44,000
Audrow Nash: awesome.

392
00:21:44,000 --> 00:21:46,821
Dave Coleman: we're trying not
to like. the development

393
00:21:46,821 --> 00:21:49,949
workflow that a lot of people
have learned in grad school and

394
00:21:49,949 --> 00:21:52,923
in college. this is a defacto
for development now, but we,

395
00:21:52,923 --> 00:21:55,846
know there's a lot of rough
edges and, I, think with the,

396
00:21:55,846 --> 00:21:58,923
what Adrian's doing at Fox Glove
is it's really exciting and

397
00:21:58,923 --> 00:22:01,846
we're, really compatible with
all those tools. So you can

398
00:22:01,846 --> 00:22:04,974
still launch our visit if you
want. we have our own, we built

399
00:22:04,974 --> 00:22:08,000
a specialized visualizer that's
really focused on robot arm

400
00:22:08,000 --> 00:22:10,000
trajectories and robot arm
manipulation

401
00:22:10,000 --> 00:22:12,000
Audrow Nash: Oh, cool.

402
00:22:12,000 --> 00:22:15,260
Dave Coleman: debugging. the,
understanding like how, a

403
00:22:15,260 --> 00:22:18,520
pipeline of grasping and,
planning works. But you could

404
00:22:18,520 --> 00:22:22,254
definitely still connect to Fox
Glove and RVIs, and we actually

405
00:22:22,254 --> 00:22:25,632
provide easy commands for that
outta the box to use those

406
00:22:25,632 --> 00:22:29,307
tools. and so we're, trying to
be very developer friendly. And

407
00:22:29,307 --> 00:22:32,626
so it could be from a high
level, like if you don't know

408
00:22:32,626 --> 00:22:36,182
all that stuff, there's a lot
you can do without it. But you

409
00:22:36,182 --> 00:22:39,502
might want someone on your team
who has some like deeper

410
00:22:39,502 --> 00:22:43,176
computer science to add those
extensions, however. Part of our

411
00:22:43,176 --> 00:22:46,377
business model is that we also
just build that for our

412
00:22:46,377 --> 00:22:50,111
customers often. So you can hire
us to do those integrations as

413
00:22:50,111 --> 00:22:51,000
well. For sure.

414
00:22:51,000 --> 00:22:55,800
Audrow Nash: Yeah. Very cool.
Okay, so you have the runtime.

415
00:22:55,800 --> 00:22:57,000
Yeah, go ahead.

416
00:22:57,000 --> 00:23:00,081
Dave Coleman: not sure I
answered your question about

417
00:23:00,081 --> 00:23:02,000
this character like this
cartoon.

418
00:23:02,000 --> 00:23:06,363
Audrow Nash: I was, I, have an
idea of what you're doing. I was

419
00:23:06,363 --> 00:23:10,448
just meaning you have a specific
task you wanna do then, so

420
00:23:10,448 --> 00:23:14,464
you're grabbing one of these,
reference solutions for this

421
00:23:14,464 --> 00:23:18,688
kind of thing. You can try it.
and then this is how our intro

422
00:23:18,688 --> 00:23:22,358
specter works. And, we have,
when you have some weird

423
00:23:22,358 --> 00:23:26,513
collision thing and your path
planning is failing, then this

424
00:23:26,513 --> 00:23:30,460
is how we debug it. You're just
writing it in your normal

425
00:23:30,460 --> 00:23:34,615
editor, like these kinds of
things. so I can see the process

426
00:23:34,615 --> 00:23:36,000
a little bit better.

427
00:23:36,000 --> 00:23:38,000
Dave Coleman: Yeah.

428
00:23:38,000 --> 00:23:39,835
Audrow Nash: but I mean, I feel
like I have a decent

429
00:23:39,835 --> 00:23:41,000
understanding of it now.
Anyways.

430
00:23:41,000 --> 00:23:44,050
Dave Coleman: Like any, like
offline programming environment,

431
00:23:44,050 --> 00:23:47,150
we, you can prototype the whole
thing in a, physics simulator,

432
00:23:47,150 --> 00:23:50,100
and we've recently upgraded that
simulator. we're not using

433
00:23:50,100 --> 00:23:52,000
Gazebo anymore. And then that's
really

434
00:23:52,000 --> 00:23:53,000
Audrow Nash: are you using?

435
00:23:53,000 --> 00:23:57,093
Dave Coleman: Be because like
gazebos always been really good

436
00:23:57,093 --> 00:24:01,118
about AMRs and I don't know,
like things that aren't as much

437
00:24:01,118 --> 00:24:05,211
contact dynamics, but we just,
we really found limitations in

438
00:24:05,211 --> 00:24:09,304
being able to like, see the,
manipulation of your fingers and

439
00:24:09,304 --> 00:24:13,329
your grippers on that. Mujoku is
actually the new thing that

440
00:24:13,329 --> 00:24:14,000
we've just

441
00:24:14,000 --> 00:24:17,000
Audrow Nash: Mujoku. Huh?

442
00:24:17,000 --> 00:24:19,000
Dave Coleman: that's a pretty
cool library.

443
00:24:19,000 --> 00:24:23,143
Audrow Nash: Gazebo. And when
you said AMRs, that's mobile

444
00:24:23,143 --> 00:24:26,929
robots. Autonomous mobile
robots. but the mu yoku, so

445
00:24:26,929 --> 00:24:30,714
gazebo is like a higher level
abstraction as far as I

446
00:24:30,714 --> 00:24:34,786
understand, than a simulator.
Like you can use mu yoku in

447
00:24:34,786 --> 00:24:38,714
gazebo, but I guess you went
down a lower level and now

448
00:24:38,714 --> 00:24:42,714
you're using mu joco more
directly. and you were able to

449
00:24:42,714 --> 00:24:46,643
maybe visualize your paths and
the gripper f the finger

450
00:24:46,643 --> 00:24:50,429
grippers and all these things,
better with this, like

451
00:24:50,429 --> 00:24:54,571
additional control you gain by
going down to just using mu

452
00:24:54,571 --> 00:24:56,000
joco. Is it Correct?

453
00:24:56,000 --> 00:24:59,863
Dave Coleman: Yeah, we can now
simulate like really complex

454
00:24:59,863 --> 00:25:03,726
admitence control and admitence
control is a, form of force

455
00:25:03,726 --> 00:25:07,917
feedback where we have a closed
loop and we can actually push on

456
00:25:07,917 --> 00:25:11,845
surfaces with a certain amount
of force and certain vectors.

457
00:25:11,845 --> 00:25:15,250
And, it, allows for just
speaking at a little higher

458
00:25:15,250 --> 00:25:19,506
level here, if we wanna do a peg
and hole insertion task, you can

459
00:25:19,506 --> 00:25:23,107
actually see the object
deforming a little bit as we do

460
00:25:23,107 --> 00:25:26,577
like an insertion or an
assembly. we, we can actually

461
00:25:26,577 --> 00:25:30,768
like, push onto things and it's
just, it's, really exciting what

462
00:25:30,768 --> 00:25:34,369
that's enabled so that when we
build, when we deploy to

463
00:25:34,369 --> 00:25:38,298
hardware, which we do a lot,
it's such a smaller gap because

464
00:25:38,298 --> 00:25:40,000
it's so much more similar.

465
00:25:40,000 --> 00:25:43,816
Audrow Nash: That's super cool.
Very interesting. Yeah, I've

466
00:25:43,816 --> 00:25:47,505
used mu joco in the past a
little bit and I actually funny

467
00:25:47,505 --> 00:25:51,194
enough, even coming from open
robotics, I have very little

468
00:25:51,194 --> 00:25:54,502
gazebo experience. I have a
little bit, but, still a

469
00:25:54,502 --> 00:25:58,000
different, I'm mostly on the ROS
2 side of things, but,

470
00:25:58,000 --> 00:26:01,322
Dave Coleman: and we still do
support gazebo. We can actually

471
00:26:01,322 --> 00:26:04,480
plug into any, simulation
environment. We have another one

472
00:26:04,480 --> 00:26:07,911
that's just we call it mock Sim,
but it's no physics. It's just

473
00:26:07,911 --> 00:26:09,000
a, whatever you tell

474
00:26:09,000 --> 00:26:10,000
Audrow Nash: RVIs, probably.

475
00:26:10,000 --> 00:26:13,761
Dave Coleman: It, whatever you
tell it to do, it just does it

476
00:26:13,761 --> 00:26:17,399
magically. and then we also, we
have some integrations into

477
00:26:17,399 --> 00:26:21,160
Nvidia, Isaac Sim. so there's
some advantages there with the,

478
00:26:21,160 --> 00:26:24,736
synthetic data generation for
ml. but we, have some of our

479
00:26:24,736 --> 00:26:28,497
customers now using our platform
for generating training data

480
00:26:28,497 --> 00:26:32,135
because we, for training models,
so that's, I, mentioned ML

481
00:26:32,135 --> 00:26:35,773
earlier, like we're more and
more using that for all of our

482
00:26:35,773 --> 00:26:39,534
computer vision needs, but also
for training models for other

483
00:26:39,534 --> 00:26:42,000
folks. It's been a big area of
interest.

484
00:26:42,000 --> 00:26:45,713
Audrow Nash: Very cool. Yeah,
it's interesting all the AI

485
00:26:45,713 --> 00:26:49,425
stuff and all the generating
data and how that bootstraps

486
00:26:49,425 --> 00:26:53,333
things. Are you guys, so people
are using you, why are using

487
00:26:53,333 --> 00:26:57,437
your new, what's your simulator
called? Are you just calling it

488
00:26:57,437 --> 00:26:59,000
the PickNik simulator or

489
00:26:59,000 --> 00:27:02,457
Dave Coleman: Yeah. it's just
MoveIt Pro and we have the

490
00:27:02,457 --> 00:27:04,000
development platform and,

491
00:27:04,000 --> 00:27:07,000
Audrow Nash: everything is under
the MoveIt Pro umbrella in a

492
00:27:07,000 --> 00:27:08,000
Dave Coleman: Yeah.

493
00:27:08,000 --> 00:27:12,030
Audrow Nash: Okay. And so when
you were saying that it's more

494
00:27:12,030 --> 00:27:15,861
hardened, and you have more of a
deployment, do, you, like

495
00:27:15,861 --> 00:27:19,428
there's a lot of tough
challenges there. There's a lot

496
00:27:19,428 --> 00:27:23,392
of infrastructure that needs to
be handled and you mentioned

497
00:27:23,392 --> 00:27:26,893
several things, but you're
handling like, things like

498
00:27:26,893 --> 00:27:31,055
deploying the code actually to
the runtime so it can run it and

499
00:27:31,055 --> 00:27:34,820
updating and like testing
introspection. It sounds really

500
00:27:34,820 --> 00:27:38,916
cool that you have the tele op
related behaviors. just tell me

501
00:27:38,916 --> 00:27:42,945
about all the infrastructure
that you have. 'cause this isn't

502
00:27:42,945 --> 00:27:46,909
what I think of with MoveIt.
MoveIt is the research and also

503
00:27:46,909 --> 00:27:50,938
very good first, way of doing
manipulation in my head. And so

504
00:27:50,938 --> 00:27:54,638
we haven't talked actually, and
I, since studio has been

505
00:27:54,638 --> 00:27:58,667
released, I think it was very
early days where it was more of

506
00:27:58,667 --> 00:28:02,763
a concept. but tell me about all
the other infrastructure that

507
00:28:02,763 --> 00:28:06,000
you have around it and how it
all works together.

508
00:28:06,000 --> 00:28:09,675
Dave Coleman: yeah. we've. We've
heard a lot of feedback that

509
00:28:09,675 --> 00:28:13,349
there's a lot of applications
where they, don't want it to be

510
00:28:13,349 --> 00:28:16,964
reliant on cloud. That could be
for security reasons. We, do

511
00:28:16,964 --> 00:28:20,518
work with, space Force a fair
amount, and so they have like

512
00:28:20,518 --> 00:28:24,253
air gap requirements. So I think
a differentiation from us and

513
00:28:24,253 --> 00:28:27,627
maybe like Intrinsic, which,
keep bringing up is there's

514
00:28:27,627 --> 00:28:31,060
really actually no cloud
connectivity aspect. I love what

515
00:28:31,060 --> 00:28:34,554
the guys over at Formic are
doing, Jeff, providing all, of

516
00:28:34,554 --> 00:28:37,867
the fleet management. There's a
couple fleet management

517
00:28:37,867 --> 00:28:41,422
companies, of course, and we
have a, we had, an integration

518
00:28:41,422 --> 00:28:45,096
with them. It's still there. We
haven't been using it lately,

519
00:28:45,096 --> 00:28:48,711
but, so if you wanna do cloud
connectivity remote operation,

520
00:28:48,711 --> 00:28:52,506
we recommend you set up a VPN on
your own, like whatever. We're

521
00:28:52,506 --> 00:28:56,301
basically network agnostic. you
do, you'll leave that to you. a

522
00:28:56,301 --> 00:28:59,554
lot of companies have big
opinions on that. we're just

523
00:28:59,554 --> 00:29:01,000
focused more on the, the

524
00:29:01,000 --> 00:29:02,000
Audrow Nash: Getting

525
00:29:02,000 --> 00:29:05,588
Dave Coleman: al algorithmic
deep control engine aspect. Like

526
00:29:05,588 --> 00:29:08,824
we see just so many companies,
they, see, a lot of like

527
00:29:08,824 --> 00:29:12,235
headlines about what the open
source can do. and maybe I'm

528
00:29:12,235 --> 00:29:15,529
guilty of this, MoveIt open
source claims to be, to do a

529
00:29:15,529 --> 00:29:19,059
lot, but when you actually get
into it, like we were talking

530
00:29:19,059 --> 00:29:22,235
about earlier, mobile
manipulation, like there's a lot

531
00:29:22,235 --> 00:29:25,941
of shortcomings and limitations
and, so we've been building out

532
00:29:25,941 --> 00:29:29,706
the things that we're missing to
actually deploy. And so you can

533
00:29:29,706 --> 00:29:33,294
do it yourself. We've helped
companies do it themselves. Just

534
00:29:33,294 --> 00:29:36,471
know that you're in for like
millions of dollars of an

535
00:29:36,471 --> 00:29:40,000
investment and it's gonna to
deploy your time to market. And

536
00:29:40,000 --> 00:29:43,647
then there's also like the long
term support maintenance where

537
00:29:43,647 --> 00:29:46,941
we're, doing that for our
customers. Now we have a, real

538
00:29:46,941 --> 00:29:50,118
time support channel. Like we
usually like share slack

539
00:29:50,118 --> 00:29:53,882
channels with our customers. we
have a QA team who's making sure

540
00:29:53,882 --> 00:29:57,353
that the quality's there and we
would just like a much more

541
00:29:57,353 --> 00:30:01,059
professional operation so that.
for companies that aren't doing

542
00:30:01,059 --> 00:30:04,294
research, this is the, thing we
recommend now to have a

543
00:30:04,294 --> 00:30:08,059
reliable, yeah. Robotic product.
And I think, the industry needs

544
00:30:08,059 --> 00:30:11,529
companies like us, like for
Foxglove to mature the tool set

545
00:30:11,529 --> 00:30:12,000
we have.

546
00:30:12,000 --> 00:30:16,186
Audrow Nash: definitely. Yeah.
And it's, do you know, do you

547
00:30:16,186 --> 00:30:18,000
know Polymath Robotics? I,

548
00:30:18,000 --> 00:30:20,821
Dave Coleman: Yeah. They're
focusing on, the navigation

549
00:30:20,821 --> 00:30:22,000
problem. Is that right?

550
00:30:22,000 --> 00:30:26,184
Audrow Nash: yes. And they, the
way I think of it is it's like

551
00:30:26,184 --> 00:30:30,368
behavior as a service kind of
thing where you communicate via,

552
00:30:30,368 --> 00:30:34,012
h like a normal rest, API
endpoint. So you just send a

553
00:30:34,012 --> 00:30:38,196
simple request and say, go here
or do something like this. and

554
00:30:38,196 --> 00:30:42,245
it will handle that kind of
thing. So they're making it like

555
00:30:42,245 --> 00:30:46,564
behavior as a service is really
how I'm thinking of what they're

556
00:30:46,564 --> 00:30:50,613
doing. And it's, they handle a
lot of the infrastructure for

557
00:30:50,613 --> 00:30:54,865
you and that doesn't really. To
my knowledge that doesn't exist

558
00:30:54,865 --> 00:30:58,712
in the manipulation space for
this kind of thing. But I'm

559
00:30:58,712 --> 00:31:02,761
hearing a lot of like echoes of
what they're doing, which is

560
00:31:02,761 --> 00:31:06,000
interesting in what you're
describing with this,

561
00:31:06,000 --> 00:31:09,220
Dave Coleman: Yeah, I mean we
have a, very high level Python,

562
00:31:09,220 --> 00:31:12,017
API, that if you wanna
integrate, maybe you have some

563
00:31:12,017 --> 00:31:15,025
existing code you've been
building out something, you can

564
00:31:15,025 --> 00:31:17,823
drop us into that. And like
you're, overarching state

565
00:31:17,823 --> 00:31:20,884
machine or you're overarching
like controlling management.

566
00:31:20,884 --> 00:31:24,156
You, can just like trigger our
Python API and we'll, take care

567
00:31:24,156 --> 00:31:27,270
of all the, like the robot
manipulation components for you.

568
00:31:27,270 --> 00:31:30,490
That, that, is one way of using
us. You can also use our user

569
00:31:30,490 --> 00:31:33,815
interface to be the operational
control panel, but we find that

570
00:31:33,815 --> 00:31:37,035
a lot of our customers, they
don't want to use our ui and we,

571
00:31:37,035 --> 00:31:40,307
totally get it. Like you have a
particular kind of operator or

572
00:31:40,307 --> 00:31:43,474
in customer who wants to see
things with certain terminology

573
00:31:43,474 --> 00:31:46,746
and they want things simplified.
And we've really developed a,

574
00:31:46,746 --> 00:31:49,649
developer platform that's built
in the, in mind of like

575
00:31:49,649 --> 00:31:52,816
engineers and roboticists and
what they need to see to build

576
00:31:52,816 --> 00:31:55,561
applications. But once the
application's pretty much

577
00:31:55,561 --> 00:31:58,728
working, you need a different
user interface to like kind of

578
00:31:58,728 --> 00:32:02,000
automation program you have. And
then we totally support that.

579
00:32:02,000 --> 00:32:05,621
Audrow Nash: Yeah, it makes
sense with who, so if you're

580
00:32:05,621 --> 00:32:09,565
targeting developers, you get
the application up and running,

581
00:32:09,565 --> 00:32:13,444
you save a lot of engineering
effort and you handle a lot of

582
00:32:13,444 --> 00:32:17,517
the maintenance burden, because
they're just using your things.

583
00:32:17,517 --> 00:32:20,944
The team doesn't have to
maintain it. The, big reason

584
00:32:20,944 --> 00:32:24,694
with Polymath that I brought it
up is they are doing, it's

585
00:32:24,694 --> 00:32:28,121
basically they're helping
companies with a lot of the

586
00:32:28,121 --> 00:32:31,935
undifferentiated work, and it
sounds like you guys are too,

587
00:32:31,935 --> 00:32:35,362
where it takes millions of
dollars to set up all this

588
00:32:35,362 --> 00:32:39,177
infrastructure and companies
just run out of money. Instead

589
00:32:39,177 --> 00:32:42,603
of, or make it. But it's like
subpar because they had

590
00:32:42,603 --> 00:32:46,547
deadlines and short, they took
shortcuts and things like this

591
00:32:46,547 --> 00:32:50,685
to get it working for one time.
But then in the future it's tech

592
00:32:50,685 --> 00:32:54,629
debt. And so you guys are in a
position where you are able to

593
00:32:54,629 --> 00:32:58,573
spend more time in it. You're
able to get buy-in from several

594
00:32:58,573 --> 00:33:02,517
different companies and then
that allows you to have the time

595
00:33:02,517 --> 00:33:06,267
and resources to really build
something of quality. and so

596
00:33:06,267 --> 00:33:10,211
then it becomes a lot easier to
just go with you guys because

597
00:33:10,211 --> 00:33:13,832
you save them time and money and
work on the maintenance

598
00:33:13,832 --> 00:33:17,000
yourself. 'cause that is the
business in a sense.

599
00:33:17,000 --> 00:33:19,409
Dave Coleman: we do find, like
we, we work with a lot of

600
00:33:19,409 --> 00:33:21,000
corporate RD labs, innovation
groups.

601
00:33:21,000 --> 00:33:22,000
Audrow Nash: Mm-Hmm

602
00:33:22,000 --> 00:33:25,504
Dave Coleman: we work with a lot
of robotic startups too. all

603
00:33:25,504 --> 00:33:28,720
sorts of organizations really.
but with the, some of the

604
00:33:28,720 --> 00:33:32,052
robotic startups, we find that
there's, some organizations

605
00:33:32,052 --> 00:33:35,440
where they have a degree in
robotics and, they, they really

606
00:33:35,440 --> 00:33:39,174
wanna just build it all in-house
and more power to you. But what,

607
00:33:39,174 --> 00:33:42,620
we can really help companies
with is like that short time to

608
00:33:42,620 --> 00:33:46,296
market is a, big, one of our big
value propositions. look, we've

609
00:33:46,296 --> 00:33:49,570
done this before. We have the
components. Let us help you

610
00:33:49,570 --> 00:33:52,959
assemble it in a unique way
that's for your application. we

611
00:33:52,959 --> 00:33:56,692
can't like, solve it just to day
one, there's usually some amount

612
00:33:56,692 --> 00:34:00,138
of configuration and let's get
these different skills in the

613
00:34:00,138 --> 00:34:01,000
right order and

614
00:34:01,000 --> 00:34:02,000
Audrow Nash: Yeah. Yeah.

615
00:34:02,000 --> 00:34:05,662
Dave Coleman: we, we find that
also companies, they, always

616
00:34:05,662 --> 00:34:09,448
want different pe, they need
different pieces of hardware. So

617
00:34:09,448 --> 00:34:12,986
they might, different vendors,
or some of them build from

618
00:34:12,986 --> 00:34:16,462
scratch, their own in-house
hardware, and we support all

619
00:34:16,462 --> 00:34:20,186
that. our only requirement is
that you have a ROS driver. if

620
00:34:20,186 --> 00:34:23,910
you speak ROS, we can drop in
the rest for you. So that's, a

621
00:34:23,910 --> 00:34:27,883
big advantage. We see a lot of,
companies that are look like us.

622
00:34:27,883 --> 00:34:31,607
They, don't have that hardware
agnostic component to it. and

623
00:34:31,607 --> 00:34:35,517
then, yeah, another value of us
is like just managing the risks

624
00:34:35,517 --> 00:34:38,000
that, that you were hitting on
just now.

625
00:34:38,000 --> 00:34:41,579
Audrow Nash: Yeah, regarding all
the like, infrastructure that's

626
00:34:41,579 --> 00:34:44,766
required. And it's fundamentally
not differentiating too,

627
00:34:44,766 --> 00:34:48,289
especially if you're doing on a
startup and this kind of thing.

628
00:34:48,289 --> 00:34:51,701
I would imagine that like the
investors would just prefer you

629
00:34:51,701 --> 00:34:55,000
buy something off the shelf so
you can move so much faster.

630
00:34:55,000 --> 00:34:56,000
Dave Coleman: Yeah.

631
00:34:56,000 --> 00:34:59,924
Audrow Nash: tell me about the,
you mentioned having something

632
00:34:59,924 --> 00:35:03,342
like 160 capabilities or
something like this. What are

633
00:35:03,342 --> 00:35:07,076
those and what's, what's an
example of what a capability is

634
00:35:07,076 --> 00:35:11,127
like? depending on how you draw
your abstractions, that could be

635
00:35:11,127 --> 00:35:14,734
a lot or few, depending, very
few. If, depending on what,

636
00:35:14,734 --> 00:35:18,785
where the abstraction is, where
like what, where the line is for

637
00:35:18,785 --> 00:35:21,000
whatever you're calling
capability.

638
00:35:21,000 --> 00:35:24,411
Dave Coleman: Yeah. I'll just
give you some high level

639
00:35:24,411 --> 00:35:28,326
categories of what our, skills
are in. So there's, the classic

640
00:35:28,326 --> 00:35:32,432
stuff you think of when it comes
to, to MoveIt, and that's motion

641
00:35:32,432 --> 00:35:33,000
planning,

642
00:35:33,000 --> 00:35:34,000
Audrow Nash: Yep.

643
00:35:34,000 --> 00:35:36,883
Dave Coleman: Cartesian
planning, global free space

644
00:35:36,883 --> 00:35:40,332
planning, being able to, execute
a trajectory and, all those.

645
00:35:40,332 --> 00:35:43,611
Kind of things. We, have a lot
built around, like not just

646
00:35:43,611 --> 00:35:47,060
doing a position controller, but
being able to switch between

647
00:35:47,060 --> 00:35:50,452
different kinds of controllers.
And I might be losing people

648
00:35:50,452 --> 00:35:53,165
saying that, but there's
velocity, there's force

649
00:35:53,165 --> 00:35:56,501
controllers. We can smoothly
switch between those depending

650
00:35:56,501 --> 00:36:00,063
on what you need. So that's the
classic stuff. And then a layer

651
00:36:00,063 --> 00:36:03,229
above that, you have task
planning and, this is a little

652
00:36:03,229 --> 00:36:06,451
less intuitive for a lot of
folks, but rather than saying

653
00:36:06,451 --> 00:36:09,843
move from A to B, then C, then D
and then you just like hard

654
00:36:09,843 --> 00:36:13,179
code, that's exactly where
you're gonna be. We can actually

655
00:36:13,179 --> 00:36:16,627
tell the robot like, here's a,
pile of boxes and we want them

656
00:36:16,627 --> 00:36:20,019
moved to the other side of the
room and we, yeah, we have to

657
00:36:20,019 --> 00:36:23,468
give them some structure to that
they need to be stacked like

658
00:36:23,468 --> 00:36:26,860
this and that, but then you can
give it to a task planner to

659
00:36:26,860 --> 00:36:29,970
break that into all the
substeps. And we have different

660
00:36:29,970 --> 00:36:33,475
flavors of that. a lower level
version of that is like a, task

661
00:36:33,475 --> 00:36:36,923
constructor for motion planning
where it can add in different

662
00:36:36,923 --> 00:36:40,316
intermediate steps. If you have
multiple arms, it can choose

663
00:36:40,316 --> 00:36:43,821
which arm to use, things like
that. And then a higher level is

664
00:36:43,821 --> 00:36:47,213
more semantic reasoning about,
to get to object A, you gotta

665
00:36:47,213 --> 00:36:50,605
move object B off of it, and
then going through a puzzle. So

666
00:36:50,605 --> 00:36:53,658
we've built out a lot of
functionality there. We don't

667
00:36:53,658 --> 00:36:56,880
honestly use task planning as
much. a lot of applications

668
00:36:56,880 --> 00:37:00,216
don't need that yet. But it's a
thing that some of our more

669
00:37:00,216 --> 00:37:03,834
research, programs, like when we
work with nasa, they, want that

670
00:37:03,834 --> 00:37:07,000
for like fully autonomous space
operations, for example.

671
00:37:07,000 --> 00:37:08,000
Audrow Nash: Cool.

672
00:37:07,000 --> 00:37:10,280
Dave Coleman: And then that
beyond that, there's machine

673
00:37:10,280 --> 00:37:13,969
learning and perception, which
is, recognize all the objects in

674
00:37:13,969 --> 00:37:17,424
the scene, be able to reason
about, we gotta open this door

675
00:37:17,424 --> 00:37:20,997
and here's the handle and here's
the hinge. And once you know

676
00:37:20,997 --> 00:37:24,627
where the door, the handle hinge
is, you know how big the door

677
00:37:24,627 --> 00:37:28,200
is, and you can figure out the
arc that you need to swing the

678
00:37:28,200 --> 00:37:31,596
door to open it. and then,
things like understanding where

679
00:37:31,596 --> 00:37:35,052
to properly grasp an object.
Like where's the wide point of

680
00:37:35,052 --> 00:37:38,565
it? Where's the narrow point?
How can I configure my arm to,

681
00:37:38,565 --> 00:37:42,138
to use that? So that, that kinda
leads us in the grasping. We

682
00:37:42,138 --> 00:37:45,827
have a lot of different grasping
algorithms. We, do some vacuum

683
00:37:45,827 --> 00:37:48,873
suction, which is the most
reliable form of robotic,

684
00:37:48,873 --> 00:37:52,211
grasping. But we also support
two finger gripping, and in

685
00:37:52,211 --> 00:37:55,314
theory we can support more
complex than that. there's

686
00:37:55,314 --> 00:37:58,887
mobile navigation, so that's
whole body planning where we can

687
00:37:58,887 --> 00:38:02,576
actually to reach something. We
not only use our arm, but we're

688
00:38:02,576 --> 00:38:06,266
gonna actually enroll our whole
body. So the wheels, if we have

689
00:38:06,266 --> 00:38:09,780
a torso joint, if we have a lift
joint. Kind of planning all

690
00:38:09,780 --> 00:38:13,469
those together. And then lastly,
something that a lot of groups

691
00:38:13,469 --> 00:38:17,217
maybe don't think about, is user
input. We do, we have built out

692
00:38:17,217 --> 00:38:20,789
a pretty complex set of, we call
it supervised autonomy, but,

693
00:38:20,789 --> 00:38:24,537
it's, not just teop. It's let's
say that you're, you've deployed

694
00:38:24,537 --> 00:38:28,285
robots with arms into production
and it's running, 99% reliable.

695
00:38:28,285 --> 00:38:31,624
And that's actually pretty
impressive in this day and age

696
00:38:31,624 --> 00:38:35,079
of what robots can do. But,
there's, always some, use cases

697
00:38:35,079 --> 00:38:38,534
where something's out of the,
way and, not where you expect

698
00:38:38,534 --> 00:38:42,282
it. We do have a lot of built in
capabilities to fall back to a,

699
00:38:42,282 --> 00:38:45,679
remote operator so that, we
don't dictate how that network

700
00:38:45,679 --> 00:38:49,134
is built for the remoteness
part, but we, can easily have a

701
00:38:49,134 --> 00:38:52,121
human say, okay, you're
confused. Let me give you a

702
00:38:52,121 --> 00:38:55,459
little nudge about how to
complete this task. And, that's

703
00:38:55,459 --> 00:38:59,148
just kinda where we are today in
deploying automation. And, the

704
00:38:59,148 --> 00:39:02,486
goal is to get that fan out,
which is a term for how many

705
00:39:02,486 --> 00:39:06,000
robots to humans? You need to
get that really positive. one,

706
00:39:06,000 --> 00:39:11,000
Audrow Nash: fan? What is it? is
that an acronym or what's the

707
00:39:11,000 --> 00:39:12,746
Dave Coleman: fan out. I'm not,
I don't think it's an acronym.

708
00:39:12,746 --> 00:39:13,000
It's just

709
00:39:13,000 --> 00:39:14,000
Audrow Nash: fan out.

710
00:39:14,000 --> 00:39:17,447
Dave Coleman: for one operator,
one like remote robot, human

711
00:39:17,447 --> 00:39:20,779
operator. how many robots can
they manage? so this is more

712
00:39:20,779 --> 00:39:24,283
advanced than Teop 'cause Teop
is usually one-to-one. in some

713
00:39:24,283 --> 00:39:27,787
of our customers we've seen that
you have actually two people

714
00:39:27,787 --> 00:39:31,521
overviewing a console screen for
one robot. But we wanna get down

715
00:39:31,521 --> 00:39:35,198
to where these robots are mostly
fully autonomous. And then once

716
00:39:35,198 --> 00:39:38,817
an hour, once a day, a human has
to drop in and give it a help.

717
00:39:38,817 --> 00:39:41,000
But we, built out that aspect as
well.

718
00:39:41,000 --> 00:39:44,860
Audrow Nash: That's really cool.
Yeah, I've seen a lot of

719
00:39:44,860 --> 00:39:48,923
companies reinvent that and then
some companies make it like

720
00:39:48,923 --> 00:39:52,782
their core competence in a
sense, like plus one robotics,

721
00:39:52,782 --> 00:39:57,048
is a good example in my mind of
they, I. Do logistics, but when

722
00:39:57,048 --> 00:40:01,314
the robot gets a crazy package,
like something I sent back from

723
00:40:01,314 --> 00:40:05,310
my Amazon returns, that's just
like incredibly poorly taped

724
00:40:05,310 --> 00:40:09,440
together. then they handle it by
having someone just like it,

725
00:40:09,440 --> 00:40:13,638
like, they have, some operator
sitting there and then they get

726
00:40:13,638 --> 00:40:17,769
the request, Hey, I don't know
what to do with this. And then

727
00:40:17,769 --> 00:40:21,764
they can give it direction to
handle. So you have something

728
00:40:21,764 --> 00:40:26,030
like that. I think that's going
to be an increasingly big thing

729
00:40:26,030 --> 00:40:30,296
for robotics as a, it just helps
with so many things to be able

730
00:40:30,296 --> 00:40:34,088
to quickly get a human in the
loop. How are you, how are

731
00:40:34,088 --> 00:40:37,000
customers receiving that? How's
that going?

732
00:40:37,000 --> 00:40:41,196
Dave Coleman: Yeah, the irony of
that is. No one wants it because

733
00:40:41,196 --> 00:40:45,134
everyone wants a robot that's
fully automated. They just want

734
00:40:45,134 --> 00:40:49,072
it to be so reliable, that you
don't need a human to drop in.

735
00:40:49,072 --> 00:40:53,204
So it's like not something that
you really lead, into of Hey, we

736
00:40:53,204 --> 00:40:57,077
can make your robot so good that
a human has to control. You

737
00:40:57,077 --> 00:41:01,144
don't wanna say that. it's like
a, the dirty secret though of a

738
00:41:01,144 --> 00:41:04,437
lot of autonomy that's in the
market already. So my

739
00:41:04,437 --> 00:41:08,246
understanding of self-driving
cars today, Waymo and Cruise,

740
00:41:08,246 --> 00:41:12,054
they all have big call centers
in the background where when

741
00:41:12,054 --> 00:41:15,089
your car runs into a
construction site, there's

742
00:41:15,089 --> 00:41:18,704
humans guiding it along and
helping it. And plus one's a

743
00:41:18,704 --> 00:41:22,835
great example. my understanding,
those, founders are, great, but

744
00:41:22,835 --> 00:41:26,967
they've been using that less and
less. That's always the hope is

745
00:41:26,967 --> 00:41:28,000
that you, go to,

746
00:41:28,000 --> 00:41:30,000
Audrow Nash: or whatever it is,
you fan out.

747
00:41:30,000 --> 00:41:33,019
Dave Coleman: Yeah, you, wanna
go to market as soon as

748
00:41:33,019 --> 00:41:36,373
possible, so you start getting
customers feedback, you prove

749
00:41:36,373 --> 00:41:39,503
traction, you get your early
revenue. It's important for

750
00:41:39,503 --> 00:41:42,913
getting the next investments
and, then, and playing the whole

751
00:41:42,913 --> 00:41:46,323
VC game. but the, value of this
human in the loop is that you

752
00:41:46,323 --> 00:41:49,677
can do that really early and
know that there's a, man behind

753
00:41:49,677 --> 00:41:52,919
the curtains to use a Wizard of
Oz analogy to help just to

754
00:41:52,919 --> 00:41:56,273
massage it along and, you gather
data. And a lot of that can

755
00:41:56,273 --> 00:41:59,627
become training data to improve
your models, to improve your

756
00:41:59,627 --> 00:42:03,149
software. but it's an important
thing that gets overlooked, and

757
00:42:03,149 --> 00:42:06,000
that's something that we've,
built in from day one.

758
00:42:06,000 --> 00:42:09,570
Audrow Nash: Ah, hell yeah.
Yeah. I think I'm seeing more

759
00:42:09,570 --> 00:42:13,579
and more companies do it, and as
you're saying with the data and

760
00:42:13,579 --> 00:42:17,149
things like this, I think it's
just a clever approach. I.

761
00:42:17,149 --> 00:42:21,032
Because as you reduce the number
of humans you need per robot,

762
00:42:21,032 --> 00:42:25,104
that gets better and better. You
get better margins for that kind

763
00:42:25,104 --> 00:42:28,924
of thing. You also learn more
about your unknown unknowns for

764
00:42:28,924 --> 00:42:32,683
this kind of thing, so you could
fit your problem better. So

765
00:42:32,683 --> 00:42:36,378
that's, really cool that you
guys have that. And so I guess

766
00:42:36,378 --> 00:42:39,134
you're not providing
infrastructure for that

767
00:42:39,134 --> 00:42:42,516
necessarily, like on a network
level, but you have the

768
00:42:42,516 --> 00:42:46,086
capability for someone to
control a robot when they need.

769
00:42:46,086 --> 00:42:49,594
So I imagine it's like code
running says, Hey, flag this

770
00:42:49,594 --> 00:42:53,102
condition. Then it goes out to
the network and says, Hey

771
00:42:53,102 --> 00:42:56,985
person, come help me with this.
And then you have someone take

772
00:42:56,985 --> 00:43:00,868
over, do something, it sends the
response back to them and the

773
00:43:00,868 --> 00:43:04,000
robot executes it or controls it
for a little bit.

774
00:43:04,000 --> 00:43:07,233
Dave Coleman: yeah. I think you
could easily put in whatever

775
00:43:07,233 --> 00:43:10,681
network you want. We'd recommend
just an SVN to start with, but,

776
00:43:10,681 --> 00:43:14,021
yeah, we, we have the. We have
the U user interface built out.

777
00:43:14,021 --> 00:43:17,416
There's all different levels of
recovery. Some of them are like

778
00:43:17,416 --> 00:43:20,756
pretty high level and easy. So
for example, imagine looking at

779
00:43:20,756 --> 00:43:24,204
a camera feed of a robot's view
of the world, and it's stuck and

780
00:43:24,204 --> 00:43:27,437
you're, it's not sure what to do
next, and you just use your

781
00:43:27,437 --> 00:43:30,131
mouse, you click on something
and our software can

782
00:43:30,131 --> 00:43:33,202
automatically then pick that
object. Maybe there was like

783
00:43:33,202 --> 00:43:36,650
three options and I didn't know
which one to go with. and that's

784
00:43:36,650 --> 00:43:39,936
a lot easier than using a game
controller to tele op or using

785
00:43:39,936 --> 00:43:42,953
little joint controls. and we
offer that too, but I just

786
00:43:42,953 --> 00:43:45,809
wouldn't recommend it as a fast
way of doing anything

787
00:43:45,809 --> 00:43:48,880
particularly. And there's other
companies that are really

788
00:43:48,880 --> 00:43:52,220
looking at like the, VR approach
where you, maybe you put on a

789
00:43:52,220 --> 00:43:55,507
headset, you have wireless hand
controllers, and, you, it's a

790
00:43:55,507 --> 00:43:58,578
one-to-one tele op. and we've
explored that. We built out

791
00:43:58,578 --> 00:44:01,864
versions of that as well. But
we're, one, people get dizzy in

792
00:44:01,864 --> 00:44:05,151
those headsets, but we wanna
provide something that's smarter

793
00:44:05,151 --> 00:44:08,168
than that. So that's why we
don't call it pure teop, but

794
00:44:08,168 --> 00:44:10,000
it's, just give it a little
nudge.

795
00:44:10,000 --> 00:44:11,000
Audrow Nash: Totally.

796
00:44:11,000 --> 00:44:14,920
Audrow Nash: Yeah, I think like
from plus one, like they just,

797
00:44:14,920 --> 00:44:18,777
like you click and the robot
figures out how to execute that.

798
00:44:18,777 --> 00:44:22,824
So they'll click on a package on
a specific location and it will

799
00:44:22,824 --> 00:44:26,617
go and pick it up there. which
uns gets the robot unstuck in

800
00:44:26,617 --> 00:44:30,664
the case. It has a broken box to
pick up or something like that.

801
00:44:30,664 --> 00:44:34,647
So that high level just this is
some guidance and I assume like

802
00:44:34,647 --> 00:44:38,441
Waymo or some companies that are
doing teop at some point or

803
00:44:38,441 --> 00:44:42,298
human assisted driving, I assume
that they probably have them

804
00:44:42,298 --> 00:44:46,218
give more high level commands
'cause driving it at a low level

805
00:44:46,218 --> 00:44:49,000
would be pretty error prone, I
would imagine

806
00:44:49,000 --> 00:44:52,289
Dave Coleman: and a lot of times
all you need is to like, have

807
00:44:52,289 --> 00:44:55,473
the robot say, okay, I'm not
fully sure what to do here, and

808
00:44:55,473 --> 00:44:58,709
this is like a, the probability
of correctness is low. So I'm

809
00:44:58,709 --> 00:45:01,945
gonna present my proposed plan
to the human operator and say,

810
00:45:01,945 --> 00:45:05,341
this is the trajectory I want to
take. This is what I'm thinking

811
00:45:05,341 --> 00:45:08,736
is right. And oftentimes you can
say, yep, that should work. And

812
00:45:08,736 --> 00:45:12,132
you just accept it. You approve
it. But if it doesn't work, then

813
00:45:12,132 --> 00:45:15,209
you need to get, get a little
more involved and modify the

814
00:45:15,209 --> 00:45:18,339
trajectory or maybe go more into
a manual control. But just

815
00:45:18,339 --> 00:45:21,629
having that approval step can be
very valuable when the robots

816
00:45:21,629 --> 00:45:22,000
lesser.

817
00:45:22,000 --> 00:45:24,871
Dave Coleman: And that's
something that we built out a

818
00:45:24,871 --> 00:45:28,115
lot for our work with nasa. I
brought them up a couple times,

819
00:45:28,115 --> 00:45:31,145
but, they've been a, great
partner in all this. and we're

820
00:45:31,145 --> 00:45:33,910
bringing that technology from,
very autonomous space

821
00:45:33,910 --> 00:45:36,888
operations, bringing it to
terrestrial applications. but

822
00:45:36,888 --> 00:45:39,706
because NASA's a more
conservative organization, they

823
00:45:39,706 --> 00:45:42,683
feel very strongly that there
should be a, human kind of

824
00:45:42,683 --> 00:45:45,980
overseeing it and they want it
to be autonomous, but also have

825
00:45:45,980 --> 00:45:48,000
the, a chance to intercept,
interrupt,

826
00:45:48,000 --> 00:45:49,000
Audrow Nash: Mm-Hmm

827
00:45:49,000 --> 00:45:51,000
Dave Coleman: go no go kind of
thing.

828
00:45:51,000 --> 00:45:54,148
Audrow Nash: What? Yeah. Since
you brought it up a few times,

829
00:45:54,148 --> 00:45:57,245
what is the NASA work, what are
you guys doing? What kind of

830
00:45:57,245 --> 00:45:59,000
problem? How have you helped
them?

831
00:45:59,000 --> 00:46:03,314
Dave Coleman: so MoveIt first
ran on the International Space

832
00:46:03,314 --> 00:46:07,268
Station a long time ago, 10
years ago, over that on the

833
00:46:07,268 --> 00:46:11,582
robonaut 2 program. and MoveIt
was used for controlling this

834
00:46:11,582 --> 00:46:15,752
humanoid, and humanoid. It
wasn't balancing on gravity, so

835
00:46:15,752 --> 00:46:20,137
it was more like a four legged
crawler through a, low gravity

836
00:46:20,137 --> 00:46:21,000
environment.

837
00:46:21,000 --> 00:46:24,431
Audrow Nash: Yeah, I think of it
like a spider kind of thing.

838
00:46:24,431 --> 00:46:27,919
'cause it's like the knees were
backwards or something. 'cause

839
00:46:27,919 --> 00:46:30,000
that was better for navigating
Uhhuh.

840
00:46:30,000 --> 00:46:33,034
Dave Coleman: It's, really cool.
Yeah. Unfortunately the program

841
00:46:33,034 --> 00:46:35,974
was pretty, pretty short lived.
I think there was something to

842
00:46:35,974 --> 00:46:38,819
do with like space exchange,
their payload, the, size of the

843
00:46:38,819 --> 00:46:41,000
door and we couldn't get the
robot in anymore.

844
00:46:41,000 --> 00:46:42,000
Audrow Nash: No way.

845
00:46:42,000 --> 00:46:45,821
Dave Coleman: And yeah, so other
space applications is, We're

846
00:46:45,821 --> 00:46:49,517
working with NASA on this, like
creating a, permanent lunar

847
00:46:49,517 --> 00:46:53,338
presence on the moon. And this
is one of the, big objectives,

848
00:46:53,338 --> 00:46:57,222
and it's good staging ground for
that. the big picture Martian

849
00:46:57,222 --> 00:47:01,106
colony. so it's, called like in
the NASA parlance like Moon to

850
00:47:01,106 --> 00:47:04,802
Mars initiative. And we've got
like Elon saying straight to

851
00:47:04,802 --> 00:47:08,435
the, to Mars. But there's a lot
of, problems with that. So

852
00:47:08,435 --> 00:47:12,382
NASA's more let's prove it all
out on the moon again before we,

853
00:47:12,382 --> 00:47:15,890
we go on. And so we're, helping
them with, robots inside

854
00:47:15,890 --> 00:47:19,272
pressurized vehicles on, on the
moon or on the future.

855
00:47:19,272 --> 00:47:22,718
Commercial Leo destinations,
space stations, as well as

856
00:47:22,718 --> 00:47:26,288
robots that would be driving
around on the surface of the

857
00:47:26,288 --> 00:47:29,733
moon, with wood arms picking
things up, assembling just

858
00:47:29,733 --> 00:47:33,054
general purpose, helping
astronauts do their job when

859
00:47:33,054 --> 00:47:36,499
the, especially 'cause the,
first, 10 years of the moon

860
00:47:36,499 --> 00:47:40,383
program, I. Is the astronauts
will be there for a short period

861
00:47:40,383 --> 00:47:44,392
every year to help out, but then
the rest of the year we'll have

862
00:47:44,392 --> 00:47:48,401
it be more autonomous. And this
is all published, NASA strategy,

863
00:47:48,401 --> 00:47:52,034
but we're, very involved with
helping that program. And, I

864
00:47:52,034 --> 00:47:55,793
also will, we'll, bookmark a
topic of, we've been working on

865
00:47:55,793 --> 00:47:59,301
Space ROS to enable our
technologies, our MoveIt pro and

866
00:47:59,301 --> 00:48:03,122
all of that to work on the moon
because for things that go in

867
00:48:03,122 --> 00:48:06,880
space, there's a, pretty high
level of, quality and software

868
00:48:06,880 --> 00:48:10,576
quality you want because you
can't just walk over there and

869
00:48:10,576 --> 00:48:14,147
unplug it. You can't have a,
human just plug in a, like a

870
00:48:14,147 --> 00:48:17,780
keyboard into your robot and
start debugging it if it goes

871
00:48:17,780 --> 00:48:21,476
down, it's a lot more high
stakes. And we've been hardening

872
00:48:21,476 --> 00:48:24,921
MoveIt pro not only for
commercial applications but for

873
00:48:24,921 --> 00:48:28,680
space applications. That's what
I mean by MoveIt is good for

874
00:48:28,680 --> 00:48:32,000
research and then MoveIt pro is
the hardened version.

875
00:48:32,000 --> 00:48:33,000
Audrow Nash: Mm-Hmm,

876
00:48:33,000 --> 00:48:35,259
Dave Coleman: and I don't dunno
if you wanna talk about Space

877
00:48:35,259 --> 00:48:37,000
ROS at all, but I'd be love to
talk about that.

878
00:48:37,000 --> 00:48:39,890
Audrow Nash: sure. Yeah. I
worked on Space ROS, actually

879
00:48:39,890 --> 00:48:42,781
when I was at Open Robotics
before we were acquired, but

880
00:48:42,781 --> 00:48:45,000
yeah. I'd love to. Tell me about
Space ROS.

881
00:48:45,000 --> 00:48:48,879
Dave Coleman: we're, hoping to
continue your work. after Open

882
00:48:48,879 --> 00:48:52,822
Robotics has gone through these
transitions that you know more

883
00:48:52,822 --> 00:48:56,764
about than I, we talked to Brian
Gerkey and Hey, PickNiks very

884
00:48:56,764 --> 00:49:00,707
involved with Space Robotics now
and, with these different na,

885
00:49:00,707 --> 00:49:04,586
NASA and Space Force programs.
So we'd love to just help lead

886
00:49:04,586 --> 00:49:08,592
this open source initiative. So
working with OSRF and, Geoffrey

887
00:49:08,592 --> 00:49:12,153
Biggs. we're currently the
chairman of the committee and

888
00:49:12,153 --> 00:49:15,397
Space ROS. We've had a couple
releases now. So it's

889
00:49:15,397 --> 00:49:19,530
essentially, think of it as like
a special distribution of, ROS 2

890
00:49:19,530 --> 00:49:23,345
humble, or ROS 2 jazzy. I, I
actually forget which one we're

891
00:49:23,345 --> 00:49:27,097
on for Space ROS, but it's
always gonna be a little behind.

892
00:49:27,097 --> 00:49:30,913
It's not gonna be the cutting
edge. It's definitely not. ROS

893
00:49:30,913 --> 00:49:35,046
rolling. For those who know what
that means, it's like the latest

894
00:49:35,046 --> 00:49:36,000
and brightness.

895
00:49:36,000 --> 00:49:37,000
Audrow Nash: bleeding edge
version?

896
00:49:37,000 --> 00:49:40,573
Dave Coleman: Yeah, nightly.
and, so we're gonna go through

897
00:49:40,573 --> 00:49:44,388
more quality checks. We're gonna
be a little more careful about

898
00:49:44,388 --> 00:49:47,961
what's included, and it's not
done yet, but the overarching

899
00:49:47,961 --> 00:49:51,413
goal is to get it up to, a
higher safety standards, which

900
00:49:51,413 --> 00:49:54,986
it's not only useful for space,
but it's gonna be great for

901
00:49:54,986 --> 00:49:58,620
medical, healthcare, surgical
robots, maybe even automotive,

902
00:49:58,620 --> 00:50:02,314
even though there's like ware
doing great work there. so it's

903
00:50:02,314 --> 00:50:05,402
good for the whole ROS
community. 'cause we've been

904
00:50:05,402 --> 00:50:09,278
really, and this is really your
work and, open robotics work. We

905
00:50:09,278 --> 00:50:12,609
don't want to fork it. We want
to just, push as much as

906
00:50:12,609 --> 00:50:16,000
possible upstream. Do you wanna
say anything about that?

907
00:50:16,000 --> 00:50:20,320
Audrow Nash: no, I mean that,
that was a great overview of it.

908
00:50:20,320 --> 00:50:24,640
let's see. Yeah, I was involved
on it for a few months. It was

909
00:50:24,640 --> 00:50:28,820
really fun, really cool work.
And I was working with Michael

910
00:50:28,820 --> 00:50:33,140
Jeronimo who was leading it, and
it was just so fun. but no, a

911
00:50:33,140 --> 00:50:37,600
great summary and I really agree
with you that the applications,

912
00:50:37,600 --> 00:50:41,711
like it's not just limited to
space. And I think one of the

913
00:50:41,711 --> 00:50:45,543
big things was all the
certifications that we're trying

914
00:50:45,543 --> 00:50:49,584
to get it ready for. So that
might mean testing things and

915
00:50:49,584 --> 00:50:53,695
specific ways of making sure
that you're not leaking memory

916
00:50:53,695 --> 00:50:57,806
or I don't know, just you're,
writing safe code in a sense.

917
00:50:57,806 --> 00:51:02,195
those we're all working towards
space certification, but as you

918
00:51:02,195 --> 00:51:06,655
said, it, makes it. A lot better
for medical or possibly driving

919
00:51:06,655 --> 00:51:10,069
cars because those
certifications are much easier

920
00:51:10,069 --> 00:51:14,040
to get because there's a good
bit of overlap between say,

921
00:51:14,040 --> 00:51:18,012
medical devices and space ROS,
as I understand, or, space

922
00:51:18,012 --> 00:51:22,262
certifications or at least it's
easier to prove to committees

923
00:51:22,262 --> 00:51:25,955
that might authorize it or
whatever the certification

924
00:51:25,955 --> 00:51:27,000
process may be.

925
00:51:27,000 --> 00:51:30,900
Dave Coleman: Yeah. Yep. So we
did, just, we just won a small

926
00:51:30,900 --> 00:51:34,928
NASA grant to fund some of this
work. 'cause a lot of it's been

927
00:51:34,928 --> 00:51:38,956
unfunded to date, just us doing
it as a side volunteer project.

928
00:51:38,956 --> 00:51:43,048
But it's certainly important to
our, future strategy. but we, do

929
00:51:43,048 --> 00:51:47,075
have a small amount of funding
now. We, encourage anyone to get

930
00:51:47,075 --> 00:51:50,656
in the GitHub repositories. We
have, I think fortnightly

931
00:51:50,656 --> 00:51:54,492
meetings, technical meetings,
and then every, once a quarter

932
00:51:54,492 --> 00:51:58,136
we have more of like a big
picture meeting. And so we got

933
00:51:58,136 --> 00:52:01,972
all the big players, all the big
government primes, like the

934
00:52:01,972 --> 00:52:05,425
Lockheed Martins of the world.
they're coming to these

935
00:52:05,425 --> 00:52:06,000
meetings.

936
00:52:06,000 --> 00:52:09,655
Dave Coleman: and yeah, there's
exciting momentum, but there's,

937
00:52:09,655 --> 00:52:13,309
it's also difficult, this week
we had a, meeting for Space ROS,

938
00:52:13,309 --> 00:52:16,500
and it's kind a dark week
because there's been a lot of

939
00:52:16,500 --> 00:52:19,749
kind of failed space robotics
programs. and I know we're

940
00:52:19,749 --> 00:52:23,403
really changing topics here from
terrestrial applications, but,

941
00:52:23,403 --> 00:52:27,000
you know a lot about Viper, I'm
sure. And it just got canceled

942
00:52:27,000 --> 00:52:30,066
Audrow Nash: For sure. I worked
on Viper for probably two years

943
00:52:30,066 --> 00:52:32,841
or something like this, so yeah,
it's a bummer. the Viper

944
00:52:32,841 --> 00:52:35,810
project, they were gonna put a
rover on the moon to go to the

945
00:52:35,810 --> 00:52:38,000
dark side and look for ice and
water and this

946
00:52:38,000 --> 00:52:39,000
Dave Coleman: Such a cool
project.

947
00:52:39,000 --> 00:52:42,237
Audrow Nash: Such a cool
project. But yeah, it got

948
00:52:42,237 --> 00:52:46,122
scrapped. I think you're right,
like in the last week or so,

949
00:52:46,122 --> 00:52:48,000
which, yeah. Bummer for sure.

950
00:52:48,000 --> 00:52:51,227
Dave Coleman: And then we've
seen that with, there are these

951
00:52:51,227 --> 00:52:54,724
missions from, nasa, Goddard os,
Sam one, os Sam two. we've hired

952
00:52:54,724 --> 00:52:57,897
some engineers from there and
those also have basically got

953
00:52:57,897 --> 00:53:00,910
scraped. It might be in limbo
land right now officially.

954
00:53:00,910 --> 00:53:04,083
'cause Congress is slow on
stuff. But, it's looking bad for

955
00:53:04,083 --> 00:53:07,526
space robotics right now. and a
lot of the reports coming out of

956
00:53:07,526 --> 00:53:10,914
like NASA committees on this is
okay, the software budgets were

957
00:53:10,914 --> 00:53:14,303
overrun and it was too complex.
And there just seems to be this

958
00:53:14,303 --> 00:53:17,154
growing evidence. To me at
least, that software reuse

959
00:53:17,154 --> 00:53:20,166
modularity. a developer
platform, if you will, is really

960
00:53:20,166 --> 00:53:23,447
necessary for these programs to
be successful. So that's what

961
00:53:23,447 --> 00:53:26,729
we've been really working on
with NASA and what we're excited

962
00:53:26,729 --> 00:53:29,902
about. And I think these lessons
are equally applicable to,

963
00:53:29,902 --> 00:53:32,000
anyone who's into terrestrial
robotics.

964
00:53:32,000 --> 00:53:35,426
Dave Coleman: Like we just see
so many companies who get

965
00:53:35,426 --> 00:53:38,912
excited about applying cutting
edge robotics to, bathroom

966
00:53:38,912 --> 00:53:42,582
cleaning, hotels, restaurants,
you name it. And then getting

967
00:53:42,582 --> 00:53:46,130
smacked in the face with the
reality of, how primitive our

968
00:53:46,130 --> 00:53:49,862
tooling is and how, building
everything from scratch is maybe

969
00:53:49,862 --> 00:53:53,410
not the best strategy anymore.
we're excited to be part of

970
00:53:53,410 --> 00:53:55,000
this, changing initiative.

971
00:53:55,000 --> 00:53:58,775
Audrow Nash: Hell yeah. Yeah. I
do think the building things

972
00:53:58,775 --> 00:54:02,550
from scratch, it's like even
like when I started programming

973
00:54:02,550 --> 00:54:06,388
or when I was still in school,
it's okay, ROS Complex to use.

974
00:54:06,388 --> 00:54:09,974
I'm gonna build my own. And
then. I did and we got pretty

975
00:54:09,974 --> 00:54:13,875
far. But then what it was is I
wanted something like Arvis and

976
00:54:13,875 --> 00:54:17,713
then it's okay, I'm starting to
build my own arvis thing. And

977
00:54:17,713 --> 00:54:21,677
then it's okay, this is far too
much work. Or then ROS bag two,

978
00:54:21,677 --> 00:54:25,641
it's I want something recording
all of the message on, messages

979
00:54:25,641 --> 00:54:29,604
I'm sending acROS. And then at
some point the pragmatic part of

980
00:54:29,604 --> 00:54:33,442
me wins. And I then learned ROS
two and then thought it was a

981
00:54:33,442 --> 00:54:37,280
very good initiative for the
community. but I feel like a lot

982
00:54:37,280 --> 00:54:41,055
of companies also do this where
they start being purists and

983
00:54:41,055 --> 00:54:45,145
being, ah, we don't need all the
overhead. And then they see that

984
00:54:45,145 --> 00:54:48,983
there are reasons why people are
using these, complex kind of

985
00:54:48,983 --> 00:54:52,884
prickly software. for some of
the ROS two stuff where it's not

986
00:54:52,884 --> 00:54:56,785
that easy to set up and you need
to be on a specific operating

987
00:54:56,785 --> 00:55:00,622
system and things like this, but
there's still a lot of value

988
00:55:00,622 --> 00:55:01,000
there.

989
00:55:01,000 --> 00:55:05,610
Dave Coleman: Yeah. I've always
told people that you should use

990
00:55:05,610 --> 00:55:09,927
as much existing software out
there to get your, your basic

991
00:55:09,927 --> 00:55:14,390
prototype together. I. There's a
lot of options paid and open

992
00:55:14,390 --> 00:55:18,927
source. And then after you have
that initial thing going, look

993
00:55:18,927 --> 00:55:23,098
at what part is really gonna
differentiate you and, maybe

994
00:55:23,098 --> 00:55:27,195
break off and, like really
optimize one piece of it, but

995
00:55:27,195 --> 00:55:31,732
don't try to bite off all of it
because, only the largest tech

996
00:55:31,732 --> 00:55:36,195
companies can afford to do that.
And even those tend to fail.

997
00:55:36,195 --> 00:55:40,000
I've seen a lot of big, Toyota
and Amazon and Google

998
00:55:40,000 --> 00:55:44,756
initiatives come and go and they
had unlimited budget, but bit so

999
00:55:44,756 --> 00:55:49,293
much that, still it didn't quite
work. And I'd say that it's a

1000
00:55:49,293 --> 00:55:53,537
funny thing that robotics does
that I don't see as much in

1001
00:55:53,537 --> 00:55:58,000
other industries, but maybe I'm
just not in those industries.

1002
00:55:58,000 --> 00:56:02,404
Audrow Nash: I think to me,
robotics, so I, do a lot of side

1003
00:56:02,404 --> 00:56:06,734
projects. and lately for the
last like maybe three years or

1004
00:56:06,734 --> 00:56:10,991
something, a lot of them have
been web related 'cause it's

1005
00:56:10,991 --> 00:56:14,000
just so damn easy to make web
stuff work.

1006
00:56:14,000 --> 00:56:15,000
Dave Coleman: Yeah.

1007
00:56:15,000 --> 00:56:19,173
Audrow Nash: but my thought is
that robotics could become

1008
00:56:19,173 --> 00:56:23,272
something like web where we have
very good tooling, very

1009
00:56:23,272 --> 00:56:27,884
established infrastructure, like
good, players in very specific

1010
00:56:27,884 --> 00:56:32,130
hard problems. Like in web
there's like auth and databases

1011
00:56:32,130 --> 00:56:36,522
and all sorts of things where
it's I don't want to learn how

1012
00:56:36,522 --> 00:56:41,061
to do all that myself. Or like
the hosting of it and whatever.

1013
00:56:41,061 --> 00:56:45,453
But robotics, I think we are
like where web was 20 years ago

1014
00:56:45,453 --> 00:56:49,699
or. 10, 15 years ago, something
like that where it's still

1015
00:56:49,699 --> 00:56:54,238
painful to do things and it's
starting to get good enough that

1016
00:56:54,238 --> 00:56:58,264
we're seeing companies like
Locus Robotics that have, I

1017
00:56:58,264 --> 00:57:02,510
don't think Locus has technical
founders, or at least they

1018
00:57:02,510 --> 00:57:06,902
started with people that are in
the industry. I may be wrong

1019
00:57:06,902 --> 00:57:08,000
about this, but

1020
00:57:08,000 --> 00:57:11,405
Dave Coleman: there's CTO, he's,
a pretty big ROS expert. Sean.

1021
00:57:11,405 --> 00:57:14,757
Yeah, I'd say he's a technical
founder, but they, did, they're

1022
00:57:14,757 --> 00:57:18,270
very smart, bringing in the CEO,
who's like an industry knowledge

1023
00:57:18,270 --> 00:57:21,405
expert, and I think that's the
winning formula that you're

1024
00:57:21,405 --> 00:57:22,000
getting at.

1025
00:57:22,000 --> 00:57:25,569
Audrow Nash: Yeah, exactly.
Where you have people that are

1026
00:57:25,569 --> 00:57:29,446
coming from the industries that
they're trying to work in doing

1027
00:57:29,446 --> 00:57:33,015
these companies, and that's
making it much more pragmatic.

1028
00:57:33,015 --> 00:57:36,585
But I do think we need, there's
a, lot of like rough edges

1029
00:57:36,585 --> 00:57:39,846
relative to something as
established as web like, and

1030
00:57:39,846 --> 00:57:43,600
there's probably A hundred times
or more web programmers than

1031
00:57:43,600 --> 00:57:47,169
roboticists too. And there's
probably a hundred times more

1032
00:57:47,169 --> 00:57:50,738
money also because of all the
big companies. so that will,

1033
00:57:50,738 --> 00:57:54,000
move all those edits down, but I
think it will change

1034
00:57:54,000 --> 00:57:56,995
Dave Coleman: Yeah, because like
in the, web world, no, no one

1035
00:57:56,995 --> 00:57:59,749
would try to just pull in all
these com, sorry, build all

1036
00:57:59,749 --> 00:58:02,889
these components from scratch to
the level that I still see robot

1037
00:58:02,889 --> 00:58:04,000
companies trying to do.

1038
00:58:04,000 --> 00:58:05,000
Audrow Nash: Totally.

1039
00:58:05,000 --> 00:58:08,476
Dave Coleman: and then there
seems to be some kind of like

1040
00:58:08,476 --> 00:58:12,313
belief that there's really deep
value in the IP and the robotics

1041
00:58:12,313 --> 00:58:15,969
ip, the robotics algorithms. and
I don't wanna, I'll just say

1042
00:58:15,969 --> 00:58:19,685
this generically, it's not to
insult any of our, past customer

1043
00:58:19,685 --> 00:58:23,341
base, but, some of the customers
we worked with, we, sent all

1044
00:58:23,341 --> 00:58:27,057
these NDAs, we have all these
legal agreements and we, look at

1045
00:58:27,057 --> 00:58:30,294
their code and it's. This is,
they've advertised it as

1046
00:58:30,294 --> 00:58:33,950
something really valuable and
secretive and we look at it and

1047
00:58:33,950 --> 00:58:37,487
we're like, oh, that's from
textbooks. That's not something

1048
00:58:37,487 --> 00:58:41,143
like that. Here. Here's another
version of it, like over here

1049
00:58:41,143 --> 00:58:44,440
that's very similar. So it's
there's I think a, misheld

1050
00:58:44,440 --> 00:58:48,096
belief that there's a lot of a
value. and really what I think

1051
00:58:48,096 --> 00:58:51,752
to Locus example, Rick, the CEO,
like it's, more about market

1052
00:58:51,752 --> 00:58:55,049
knowledge execution of
penetrating that market, knowing

1053
00:58:55,049 --> 00:58:58,765
what your competitive advantage
is. And maybe that competitive

1054
00:58:58,765 --> 00:59:02,301
advantage isn't the robotics
side so much, but it's how you

1055
00:59:02,301 --> 00:59:05,957
package the business model to
work with that industry. 'cause

1056
00:59:05,957 --> 00:59:09,194
there's still a lot of like
resistance to robotics and

1057
00:59:09,194 --> 00:59:12,910
automation from people outside
of the average people listening

1058
00:59:12,910 --> 00:59:16,267
to this podcast and robotics,
I'd assume. and I think we

1059
00:59:16,267 --> 00:59:19,743
forget that as folks who are
really in this industry, that

1060
00:59:19,743 --> 00:59:22,860
there's such expertise to
penetrate. 'cause PickNiks

1061
00:59:22,860 --> 00:59:26,516
looked at specific industries
that we wanted to maybe tackle.

1062
00:59:26,516 --> 00:59:30,173
you still hear me? You hear me
now? Yeah, so I, I got cut off

1063
00:59:30,173 --> 00:59:33,829
for a second, but I, think the
winning model is yeah, knowing

1064
00:59:33,829 --> 00:59:37,665
how to package the robotics in a
way that works for the business

1065
00:59:37,665 --> 00:59:41,022
model and the needs of that
specific industry. And, even

1066
00:59:41,022 --> 00:59:44,618
something as simple as what to
call the buttons in the ui. a

1067
00:59:44,618 --> 00:59:48,334
lot of people don't understand
what the word Cartesian control

1068
00:59:48,334 --> 00:59:52,110
means outside of our niche. So
maybe you need, like I hear some

1069
00:59:52,110 --> 00:59:55,766
industries call it like endpoint
control. It's a really silly

1070
00:59:55,766 --> 00:59:59,363
thing, but modifying those,
nuances can really make or break

1071
00:59:59,363 --> 01:00:02,000
how successful your robotics
application is.

1072
01:00:02,000 --> 01:00:05,996
Audrow Nash: Yeah. Especially
depending on who's using it. So

1073
01:00:05,996 --> 01:00:09,533
if the people are not
roboticists, that makes a lot of

1074
01:00:09,533 --> 01:00:13,267
sense. which probably most
people are not roboticists. So

1075
01:00:13,267 --> 01:00:16,870
that probably make a lot of
sense. Yeah. I'm seeing one

1076
01:00:16,870 --> 01:00:20,931
thing that's very cool to me is
that. So I've been podcasting,

1077
01:00:20,931 --> 01:00:25,124
we talked about before or before
the recording, I think maybe 11

1078
01:00:25,124 --> 01:00:29,251
or 12 years now. Robohub, Sense
Think Act. Now this, and I feel

1079
01:00:29,251 --> 01:00:33,116
like the companies that I'm
interviewing in the last couple

1080
01:00:33,116 --> 01:00:36,981
years are doing this way better
than any time before. in my

1081
01:00:36,981 --> 01:00:40,845
experience and I, feel like I
was talking to good companies

1082
01:00:40,845 --> 01:00:44,514
before, but they are finding a
market and finding a good

1083
01:00:44,514 --> 01:00:48,444
business model. And it's more
about how do you deploy in the

1084
01:00:48,444 --> 01:00:52,244
market and stay and create value
right away, but then stay

1085
01:00:52,244 --> 01:00:56,174
valuable and maybe you grow to
something even more valuable.

1086
01:00:56,174 --> 01:01:00,170
And it's so exciting, like these
pragmatic founders that are,

1087
01:01:00,170 --> 01:01:04,297
starting companies and the great
direction and traction that it

1088
01:01:04,297 --> 01:01:06,000
seems like they're getting

1089
01:01:06,000 --> 01:01:09,732
Dave Coleman: Yeah. And we say
all this in, I think PickNik and

1090
01:01:09,732 --> 01:01:13,345
myself we're equally guilty of
we focus just on the robotics.

1091
01:01:13,345 --> 01:01:16,721
from day one, we've always done
that. We don't have an in

1092
01:01:16,721 --> 01:01:20,453
application that we're not just
targeting like robots that make

1093
01:01:20,453 --> 01:01:23,000
burritos really well and auto
autonomously,

1094
01:01:23,000 --> 01:01:24,000
Audrow Nash: Yeah.

1095
01:01:24,000 --> 01:01:27,525
Dave Coleman: And, so I'm not
like, maybe not taking my own

1096
01:01:27,525 --> 01:01:31,110
medicine here, but I think we
are in a unique position where

1097
01:01:31,110 --> 01:01:34,814
we're a really recognized brand
of expertise for robotic arms.

1098
01:01:34,814 --> 01:01:38,637
We've got really well positioned
in the, ROS ecosystem. and I do

1099
01:01:38,637 --> 01:01:42,102
think the industry could use
players who are providing the

1100
01:01:42,102 --> 01:01:45,807
tools and the engines and the
runtimes. And I think we have a,

1101
01:01:45,807 --> 01:01:49,331
good head start here and we want
to keep doing that, but we

1102
01:01:49,331 --> 01:01:52,558
really are dependent on our
customers and our partners

1103
01:01:52,558 --> 01:01:56,202
tackling those applications.
And, we want to, we wanna enable

1104
01:01:56,202 --> 01:01:59,787
them. So I actually think it's a
good segue to talk about, a

1105
01:01:59,787 --> 01:02:03,610
customer that I mentioned to you
earlier before the call that we

1106
01:02:03,610 --> 01:02:06,000
just published a, a case study
with 'em.

1107
01:02:06,000 --> 01:02:09,149
Dave Coleman: It's on our
website. It's top of the list

1108
01:02:09,149 --> 01:02:12,412
for case studies, so if you
wanna see the video. But Hive

1109
01:02:12,412 --> 01:02:15,904
Botics, is an example. So their
CEO, Rashab, he's like really

1110
01:02:15,904 --> 01:02:19,396
focused. He's more of a business
person focused on how do we,

1111
01:02:19,396 --> 01:02:22,430
bring. Bring like bathroom
cleaning robots quickly to

1112
01:02:22,430 --> 01:02:25,751
market. And so he's, instead of
spending years in software

1113
01:02:25,751 --> 01:02:29,128
development and working on the
research of it and technical

1114
01:02:29,128 --> 01:02:32,792
risks, they, chose to work with
us. And, so we're really honored

1115
01:02:32,792 --> 01:02:36,227
for that. And we've been able
to, in a matter of months, get

1116
01:02:36,227 --> 01:02:39,777
them a working prototype. We got
a video on our website of it,

1117
01:02:39,777 --> 01:02:43,097
driving through, a bathroom and
cleaning various sinks and

1118
01:02:43,097 --> 01:02:46,417
toilets and stalls. And so I
wanna, I'm gonna give a shout

1119
01:02:46,417 --> 01:02:49,681
out to their success. And I, I
hope that, we can continue

1120
01:02:49,681 --> 01:02:52,886
serving 'em really well. But
that's kinda the model that

1121
01:02:52,886 --> 01:02:56,035
we're looking at now, is
founders who do have this very

1122
01:02:56,035 --> 01:02:59,470
clear application and business
model who can get those early

1123
01:02:59,470 --> 01:03:02,619
contracts and pilots with
customers. knowing that there

1124
01:03:02,619 --> 01:03:05,653
are companies like us who can
just de-risk that whole

1125
01:03:05,653 --> 01:03:08,000
technical side and, do it really
cheaply.

1126
01:03:08,000 --> 01:03:09,000
Audrow Nash: Mm-Hmm.

1127
01:03:09,000 --> 01:03:11,548
Dave Coleman: It's kinda, a good
example of what we're talking

1128
01:03:11,548 --> 01:03:12,000
about here.

1129
01:03:12,000 --> 01:03:15,985
Audrow Nash: Yeah. You have the
expertise. It's similar to the

1130
01:03:15,985 --> 01:03:19,520
problems you're already
approaching. That's super cool.

1131
01:03:19,520 --> 01:03:23,312
I wonder, so it sounds like
they, one thing that'd be super

1132
01:03:23,312 --> 01:03:27,361
cool, and this would be funny as
like a, silly. But potentially

1133
01:03:27,361 --> 01:03:31,410
good business model, is you do
that kind of thing, but you take

1134
01:03:31,410 --> 01:03:35,524
equity in the companies that you
help for this kind of thing. So

1135
01:03:35,524 --> 01:03:39,573
it'd be like, you're like almost
like a vc, but you take equity

1136
01:03:39,573 --> 01:03:43,558
by helping them solve some of
their technical risk. that would

1137
01:03:43,558 --> 01:03:46,000
be pretty cool, I think. I don't
know.

1138
01:03:46,000 --> 01:03:49,612
Dave Coleman: Yeah. Never really
considered that. But that's,

1139
01:03:49,612 --> 01:03:53,342
it's a good point because we are
really like deeply embedded as

1140
01:03:53,342 --> 01:03:55,000
partners with our customers.

1141
01:03:55,000 --> 01:03:59,071
Audrow Nash: you help them move
a lot faster and, you could be

1142
01:03:59,071 --> 01:04:03,208
more long-term partners for this
kind of thing. Like they could

1143
01:04:03,208 --> 01:04:07,280
pay you but also give you equity
in a sense. I don't know. I'm

1144
01:04:07,280 --> 01:04:11,154
just, I don't know why I've been
thinking about like the VC

1145
01:04:11,154 --> 01:04:15,356
angle. I've been thinking about
like in, in some past interviews

1146
01:04:15,356 --> 01:04:19,625
because if you have say way more
information or you can help them

1147
01:04:19,625 --> 01:04:23,827
technically de-risk or something
like that, there's not really a

1148
01:04:23,827 --> 01:04:27,505
reason you shouldn't take
advantage of that and get some

1149
01:04:27,505 --> 01:04:30,000
of the upside if you help the
company.

1150
01:04:30,000 --> 01:04:31,000
Dave Coleman: Yeah.

1151
01:04:31,000 --> 01:04:34,891
Audrow Nash: And one company,
I'm editing the interview now is

1152
01:04:34,891 --> 01:04:38,343
Koop, and they do insurance. And
during the, during the

1153
01:04:38,343 --> 01:04:41,983
interview, it's like they have
all of these metrics on the

1154
01:04:41,983 --> 01:04:45,936
success or not of the company.
It's like they could use that to

1155
01:04:45,936 --> 01:04:49,576
be a very powerful VC arm
because typical VCs are not even

1156
01:04:49,576 --> 01:04:53,404
some, are people in technology.
but it's interesting if. Like

1157
01:04:53,404 --> 01:04:57,233
you and your company, you are
literally de-risking a problem,

1158
01:04:57,233 --> 01:05:00,998
the technology for them. And
that's one of the most valuable

1159
01:05:00,998 --> 01:05:04,512
things, and that's one of the
biggest risks that VCs are

1160
01:05:04,512 --> 01:05:08,090
trying to get around. Like they,
they need to look at the

1161
01:05:08,090 --> 01:05:11,980
founders and the credentials and
things like this, whereas you

1162
01:05:11,980 --> 01:05:15,934
guys can just be like, oh yeah,
I'll do it. I'll de-risk it for

1163
01:05:15,934 --> 01:05:19,951
you. Now you guys figure out the
market strategy and we can help

1164
01:05:19,951 --> 01:05:23,653
you going forward. We can tell
you that RRT is gonna be way

1165
01:05:23,653 --> 01:05:27,419
better for planning than some
other algorithm or whatever it

1166
01:05:27,419 --> 01:05:30,933
might be. and show them the
implementation. Just a silly

1167
01:05:30,933 --> 01:05:32,000
thought. I dunno.

1168
01:05:32,000 --> 01:05:35,946
Dave Coleman: yeah, Our current
strategy is to come in really

1169
01:05:35,946 --> 01:05:39,893
affordable at the start. and so
we, we think we've got a, for

1170
01:05:39,893 --> 01:05:43,839
what we're providing really low
price point, like I'm talking

1171
01:05:43,839 --> 01:05:47,720
like the cost of, two months of
an engineer. it's what we're

1172
01:05:47,720 --> 01:05:51,408
charging essentially. and, we
can, but to do what we, can

1173
01:05:51,408 --> 01:05:55,354
provide you would take probably
a year or two. So it's a huge

1174
01:05:55,354 --> 01:05:59,171
initial investment, return on
investment. And then, as. If,

1175
01:05:59,171 --> 01:06:03,053
and as were, us, let's say as,
our, customers are successful

1176
01:06:03,053 --> 01:06:06,611
and they, scale up, then we're
looking for a very small

1177
01:06:06,611 --> 01:06:10,428
percentage of return on that.
it's not equity, but there is

1178
01:06:10,428 --> 01:06:14,568
some similarities I'd say. But,
we, want to meet, projects where

1179
01:06:14,568 --> 01:06:18,579
they're at. So oftentimes in the
first phase of the prototype,

1180
01:06:18,579 --> 01:06:22,331
the pilot stage budgets are
small and you're really trying

1181
01:06:22,331 --> 01:06:26,342
to prove to your higher ups or
to your investors that, this is

1182
01:06:26,342 --> 01:06:30,030
a good thing to invest in. So
we're, trying to match that

1183
01:06:30,030 --> 01:06:31,000
budget profile.

1184
01:06:31,000 --> 01:06:34,516
Audrow Nash: So it's a
beachhead, basically into the

1185
01:06:34,516 --> 01:06:38,708
business. And from there you can
do more work for this kind of

1186
01:06:38,708 --> 01:06:42,833
thing. I wonder just, put it on
your radar as like a consider

1187
01:06:42,833 --> 01:06:47,025
taking equity, especially in the
ones that you think are super

1188
01:06:47,025 --> 01:06:50,000
promising. let's see. Okay.
Super cool. And,

1189
01:06:50,000 --> 01:06:53,452
Dave Coleman: had a thought
about, about ROS. if we wanted

1190
01:06:53,452 --> 01:06:56,904
to circle back a little bit
before moving on. So this is a

1191
01:06:56,904 --> 01:07:00,594
little about ROS. Yeah. just a
thought from earlier when I was

1192
01:07:00,594 --> 01:07:04,106
bookmarked in my head is I'm
seeing a trend in our industry

1193
01:07:04,106 --> 01:07:07,915
as well where, ROS continues, I
think to grow maybe, it's slowed

1194
01:07:07,915 --> 01:07:11,605
down its growth, but like it's
still very prominent in all the

1195
01:07:11,605 --> 01:07:15,235
robot companies that I talk to
at least. but what I'm excited

1196
01:07:15,235 --> 01:07:18,806
about and this is weird to hear
it say this, but I think the

1197
01:07:18,806 --> 01:07:22,080
scope of ROS is somewhat
shrinking and I think that's a

1198
01:07:22,080 --> 01:07:25,770
good thing. the ROS that first
came to market in like 2010, it

1199
01:07:25,770 --> 01:07:29,460
was attempting to do a lot of
things. So people would think of

1200
01:07:29,460 --> 01:07:32,912
it as the navigation and the
motion control and arvis, the

1201
01:07:32,912 --> 01:07:36,304
visualizer and their simulator.
And there's just it's got

1202
01:07:36,304 --> 01:07:39,994
everything. And I think that the
more proper scope of ROS, and

1203
01:07:39,994 --> 01:07:43,446
this is, I. This is a little
bit, I dunno a hot take here,

1204
01:07:43,446 --> 01:07:47,077
but it's, the middleware, it's
the message passing, it's the,

1205
01:07:47,077 --> 01:07:50,767
standardized, interface so that
all these different actuators,

1206
01:07:50,767 --> 01:07:54,100
motors, sensors, we want them
all to be talking the same

1207
01:07:54,100 --> 01:07:57,373
language so that when you buy
different components from

1208
01:07:57,373 --> 01:08:01,123
different vendors, you can plug
'em all together. and then what

1209
01:08:01,123 --> 01:08:04,634
I'm seeing in the industry is
you're seeing these companies

1210
01:08:04,634 --> 01:08:08,086
like PickNik and like others
providing software components

1211
01:08:08,086 --> 01:08:11,896
that provide those higher level
capabilities or, you can develop

1212
01:08:11,896 --> 01:08:15,169
yourself. That's always a,
perfectly good answer, but I

1213
01:08:15,169 --> 01:08:18,859
think that's a good trend for,
our industry so that people are

1214
01:08:18,859 --> 01:08:22,430
specializing in and, obviously
open robotics is they need to

1215
01:08:22,430 --> 01:08:25,465
focus because they're, not a
huge organization post

1216
01:08:25,465 --> 01:08:28,917
acquisition. And, so that's,
kinda my take on that. And I,

1217
01:08:28,917 --> 01:08:32,310
still am a big fan of the me
the, middleware. I'm excited

1218
01:08:32,310 --> 01:08:35,762
about. they're changing the
middleware maybe away from DDS

1219
01:08:35,762 --> 01:08:36,000
and,

1220
01:08:36,000 --> 01:08:37,000
Audrow Nash: I can't wait.

1221
01:08:37,000 --> 01:08:39,360
Dave Coleman: because, that's,
there's a lot of good things

1222
01:08:39,360 --> 01:08:40,000
happening there.

1223
01:08:40,000 --> 01:08:44,530
Audrow Nash: for sure. Yeah. I
like your hot take. I agree with

1224
01:08:44,530 --> 01:08:48,916
you. I think that it getting
smaller is good and the, like I,

1225
01:08:48,916 --> 01:08:53,230
I agree we're, giving more to
the community and we're seeing

1226
01:08:53,230 --> 01:08:57,400
high quality maintainers like
you guys that are working on

1227
01:08:57,400 --> 01:09:01,714
things or Fox Glove and their
mcap, which helps with ROS Bag

1228
01:09:01,714 --> 01:09:06,100
and things like this. so I think
it's really nice that we can

1229
01:09:06,100 --> 01:09:10,701
give back some of the parts that
we. Or not maintaining as well,

1230
01:09:10,701 --> 01:09:15,159
or they were like edgy projects
and something in the community

1231
01:09:15,159 --> 01:09:19,473
is coming back to replace them
better. that to me feels very

1232
01:09:19,473 --> 01:09:23,572
healthy. And I also, I would add
that in addition to just

1233
01:09:23,572 --> 01:09:28,029
middleware, it's infrastructure
because we have a lot of stuff

1234
01:09:28,029 --> 01:09:32,128
that all makes the community
connect. so there's like the

1235
01:09:32,128 --> 01:09:36,298
build farm and the way of
getting all the packages that we

1236
01:09:36,298 --> 01:09:40,468
have. There's the, all the
testing related things we have,

1237
01:09:40,468 --> 01:09:44,494
and these as well as like the
documentation and then the

1238
01:09:44,494 --> 01:09:48,809
discourse and things like this
that I think are a big reason

1239
01:09:48,809 --> 01:09:53,266
that the community is as big and
flourishing as much as it is,

1240
01:09:53,266 --> 01:09:57,580
because we've set all that up
and it's almost like we get to

1241
01:09:57,580 --> 01:10:01,894
play mediator now in some way
and just have people use a lot

1242
01:10:01,894 --> 01:10:05,705
of what we've built, and
maintain it and that feature

1243
01:10:05,705 --> 01:10:09,300
slowly. But I think
consolidating what we actually

1244
01:10:09,300 --> 01:10:13,758
have as our core value add is
probably very smart and probably

1245
01:10:13,758 --> 01:10:17,281
a very good sign for the
maturity of the robotics

1246
01:10:17,281 --> 01:10:18,000
community.

1247
01:10:18,000 --> 01:10:22,000
Dave Coleman: I agree. Yep.

1248
01:10:22,000 --> 01:10:23,000
Audrow Nash: Oh yeah.

1249
01:10:22,000 --> 01:10:24,609
Dave Coleman: Yeah, that, that
was, only thought I wanted to

1250
01:10:24,609 --> 01:10:27,000
bookmark, and I think you were
going on to a new topic.

1251
01:10:27,000 --> 01:10:30,910
Audrow Nash: I don't know, I was
just talking, but I did wanna

1252
01:10:30,910 --> 01:10:34,568
make sure we talked about,
mobile manipulation and because

1253
01:10:34,568 --> 01:10:38,477
you've mentioned it and it's not
what I think of for MoveIt, I

1254
01:10:38,477 --> 01:10:42,198
think of robotic arms doing
stuff, and mobile manipulation.

1255
01:10:42,198 --> 01:10:45,730
I know it's been technically
possible to do, but I think

1256
01:10:45,730 --> 01:10:49,703
you're supporting it more now as
like a first part, first party

1257
01:10:49,703 --> 01:10:53,676
service or first party part of
your application. So I'd love to

1258
01:10:53,676 --> 01:10:55,000
hear more about that.

1259
01:10:55,000 --> 01:10:57,000
Dave Coleman: Yeah.

1260
01:10:57,000 --> 01:11:00,791
Dave Coleman: if, for those who
might recall the PR two robot,

1261
01:11:00,791 --> 01:11:04,277
that was the initial flagship
robot at Willow Garage, the

1262
01:11:04,277 --> 01:11:07,763
company that started ROS
effectively, and so move, it was

1263
01:11:07,763 --> 01:11:11,493
actually developed for the PR
two. It had two arms and it had

1264
01:11:11,493 --> 01:11:15,345
wheels. And I think people often
forget that. But from the very

1265
01:11:15,345 --> 01:11:18,709
beginning, move's always had
some level of concept of a

1266
01:11:18,709 --> 01:11:22,255
holonomic base, which is a robot
base that can move in all

1267
01:11:22,255 --> 01:11:25,863
directions. And I think that's
been kinda lost. It's, never

1268
01:11:25,863 --> 01:11:29,471
fully developed, but it's,
always been in MoveIt, but we're

1269
01:11:29,471 --> 01:11:33,385
really making that a first class
focus. So we've, done some open

1270
01:11:33,385 --> 01:11:36,932
source, contributions back to
MoveIt, to support that. And

1271
01:11:36,932 --> 01:11:40,662
then we're also building out a
lot more MoveIt Pro, the, that

1272
01:11:40,662 --> 01:11:44,209
is going to, it's already out
there really allow for doing

1273
01:11:44,209 --> 01:11:47,939
advanced, planning with wheels
and I, and one, one reason why

1274
01:11:47,939 --> 01:11:51,669
we think this is a hot topic
right now is humanoids are huge.

1275
01:11:51,669 --> 01:11:55,399
They just got funded so much and
there's a lot of other folks

1276
01:11:55,399 --> 01:11:59,129
who've been in the industry like
myself who are a little more

1277
01:11:59,129 --> 01:12:02,737
skeptical of the, of, the full,
craze of it. I think it's a

1278
01:12:02,737 --> 01:12:05,000
bubble personally. What do you
think?

1279
01:12:05,000 --> 01:12:09,179
Audrow Nash: I think it's like
autonomous cars. I'm a lot of

1280
01:12:09,179 --> 01:12:13,566
the same. Types of things seem
to be occurring with, autonomous

1281
01:12:13,566 --> 01:12:17,815
vehicles, or with humanoids as a
human, autonomous cars. It's

1282
01:12:17,815 --> 01:12:21,784
like I was, feeling like I'm
reliving the same phenomenon

1283
01:12:21,784 --> 01:12:26,311
when there was all the craze and
then it's like, why is this, oh,

1284
01:12:26,311 --> 01:12:30,212
like the dynamics, the hype is
pretty similar. There are

1285
01:12:30,212 --> 01:12:33,763
several things that are
different, but I think that

1286
01:12:33,763 --> 01:12:37,176
there, so for example, an
autonomous car might be

1287
01:12:37,176 --> 01:12:41,355
traveling at a speed where it's
unsafe to have a glitch, and

1288
01:12:41,355 --> 01:12:45,185
then you really need super safe
systems and that's very

1289
01:12:45,185 --> 01:12:49,434
expensive. And a at a humanoid
can just like pause and freeze

1290
01:12:49,434 --> 01:12:53,473
most likely in most tasks. so
there are slightly different

1291
01:12:53,473 --> 01:12:58,000
dynamics, but I think it'll be a
little bit like autonomous cars.

1292
01:12:58,000 --> 01:13:01,701
Dave Coleman: and by that, I,
what I'm hearing you say is like

1293
01:13:01,701 --> 01:13:04,866
a, long tail investment and
we're gonna be, it can go

1294
01:13:04,866 --> 01:13:08,507
through a couple waves of it's
almost here and then it's not,

1295
01:13:08,507 --> 01:13:10,000
we're gonna disappointed.

1296
01:13:10,000 --> 01:13:14,448
Audrow Nash: Yeah. and also I
talked, with, let's see, I

1297
01:13:14,448 --> 01:13:18,975
talked with an investor at F
Prime. And they, what he was

1298
01:13:18,975 --> 01:13:23,264
saying is that from his
perspective, investing in them

1299
01:13:23,264 --> 01:13:28,188
doesn't really make any sense
because there is no clear payoff

1300
01:13:28,188 --> 01:13:32,953
date when the technology's ready
for this kind of thing, and

1301
01:13:32,953 --> 01:13:37,877
there's no clear market. So it's
a lot like I, autonomous cars

1302
01:13:37,877 --> 01:13:42,563
had like the okay, robo taxi or
whatever it might be, but I

1303
01:13:42,563 --> 01:13:47,487
think. Humanoids. There's some
use cases, but they're probably

1304
01:13:47,487 --> 01:13:52,570
small markets like Melanie Wise,
from Agility Robotics convinced

1305
01:13:52,570 --> 01:13:57,653
me that they're good for her use
case and they were doing Amazon

1306
01:13:57,653 --> 01:14:02,181
tote delivery and things like
this. and so there are some

1307
01:14:02,181 --> 01:14:06,866
advantages to lags in that case,
but I think that the, it's

1308
01:14:06,866 --> 01:14:11,155
narrow markets that'll be
useful. It's not the general

1309
01:14:11,155 --> 01:14:15,841
thing, at least for the next
several years, I think five or

1310
01:14:15,841 --> 01:14:16,000
10

1311
01:14:16,000 --> 01:14:18,664
Dave Coleman: I have a
contrarian view of this. one

1312
01:14:18,664 --> 01:14:21,642
example I like to give is, most
so humanoid would be more

1313
01:14:21,642 --> 01:14:24,776
expensive, and of course we're
seeing prices drop from China

1314
01:14:24,776 --> 01:14:27,806
and so forth, but I think
they're inherently gonna be more

1315
01:14:27,806 --> 01:14:30,000
expensive hardware than like
wheeled robot

1316
01:14:30,000 --> 01:14:32,000
Audrow Nash: Yeah,

1317
01:14:32,000 --> 01:14:33,000
Dave Coleman: and

1318
01:14:33,000 --> 01:14:35,000
Audrow Nash: are already super
expensive.

1319
01:14:35,000 --> 01:14:39,016
Dave Coleman: Correct. But I
just, an observation I have is

1320
01:14:39,016 --> 01:14:43,236
that most applications that can
afford a, a humanoid robot are

1321
01:14:43,236 --> 01:14:46,911
gonna be already handicap
accessible. I don't think an

1322
01:14:46,911 --> 01:14:50,791
average consumer can afford a
humanoid in their house and

1323
01:14:50,791 --> 01:14:54,942
houses where you tend to have
like multi-levels and steps and

1324
01:14:54,942 --> 01:14:59,162
staircases. But with handicap
laws, we have elevators, we have

1325
01:14:59,162 --> 01:15:01,000
pretty flat surfaces, so I,

1326
01:15:01,000 --> 01:15:02,000
Audrow Nash: the

1327
01:15:01,000 --> 01:15:02,000
Dave Coleman: have a hard time.

1328
01:15:02,000 --> 01:15:04,316
Audrow Nash: that's not
necessarily true elsewhere, but

1329
01:15:04,316 --> 01:15:06,000
the US is the biggest market and
I think

1330
01:15:06,000 --> 01:15:07,000
Dave Coleman: It's true.

1331
01:15:07,000 --> 01:15:10,313
Audrow Nash: or a huge market
and very consumption like

1332
01:15:10,313 --> 01:15:13,807
driven, whereas I don't know,
China for example, is not as

1333
01:15:13,807 --> 01:15:17,000
much of a consumption economy,
these kinds of things.

1334
01:15:17,000 --> 01:15:19,677
Dave Coleman: That's true.
That's a good point. And that's

1335
01:15:19,677 --> 01:15:22,354
a very US viewpoint of mine,
like when I go to Europe, but

1336
01:15:22,354 --> 01:15:23,000
things are way

1337
01:15:23,000 --> 01:15:25,000
Audrow Nash: different. Not
handicap accessible generally.

1338
01:15:25,000 --> 01:15:28,832
Dave Coleman: And that's just
for historical reasons, right?

1339
01:15:28,832 --> 01:15:32,536
The buildings are so old. okay,
that, that's one view. But

1340
01:15:32,536 --> 01:15:36,432
anyway, I, just think that the
mobile manipulation's a little

1341
01:15:36,432 --> 01:15:40,392
bit, it's, certainly not in the
hype right now, and that might

1342
01:15:40,392 --> 01:15:44,415
be a signal that it's a mature
technology. not that like we, we

1343
01:15:44,415 --> 01:15:47,928
don't see a, lot of. Like
applications out deployed yet

1344
01:15:47,928 --> 01:15:51,951
for it. But I just think that's
gonna be ready much sooner than

1345
01:15:51,951 --> 01:15:56,039
humanoids. So that's what we're
focusing on. And the people have

1346
01:15:56,039 --> 01:15:59,934
been working on this for a long
time. the, toilet cleaning Hi

1347
01:15:59,934 --> 01:16:03,000
Botics demonstration I just
talked about. That's

1348
01:16:03,000 --> 01:16:04,000
Audrow Nash: out the video.

1349
01:16:04,000 --> 01:16:07,751
Dave Coleman: that's all mobile
manipulation. We've got a couple

1350
01:16:07,751 --> 01:16:11,150
others. NASA's really into it
right now with us. and a few

1351
01:16:11,150 --> 01:16:14,725
more that we'll be announcing
soon. I just wanted to get that

1352
01:16:14,725 --> 01:16:18,183
message out is that I think
there's still a lot of room for

1353
01:16:18,183 --> 01:16:20,000
wheels that we're undervaluing.

1354
01:16:20,000 --> 01:16:21,000
Audrow Nash: Oh, totally.

1355
01:16:21,000 --> 01:16:22,000
Dave Coleman: kind my take.

1356
01:16:21,000 --> 01:16:25,381
Audrow Nash: I completely agree.
I think a lot of the best

1357
01:16:25,381 --> 01:16:30,140
companies that I've been talking
to are companies that are just

1358
01:16:30,140 --> 01:16:34,597
doing robots with wheels and
having them like do some task,

1359
01:16:34,597 --> 01:16:38,525
whether it be mowing or just
driving around, sensing

1360
01:16:38,525 --> 01:16:43,209
something, whatever it is. or
some, simple, I don't know, some

1361
01:16:43,209 --> 01:16:47,892
simple tasks that kind of comes
with the vehicle. I think that

1362
01:16:47,892 --> 01:16:51,820
my, my thought on it is that
manipulation and mobile

1363
01:16:51,820 --> 01:16:56,277
manipulation, particularly
exciting, I suppose is, probably

1364
01:16:56,277 --> 01:17:00,960
where just navigation was maybe
like five or eight or 10 years

1365
01:17:00,960 --> 01:17:05,644
ago. And so I think probably the
next wave of robotic startups

1366
01:17:05,644 --> 01:17:10,403
that we're gonna start seeing,
Chef Robotics is a great example

1367
01:17:10,403 --> 01:17:14,482
of a company that's doing
manipulation and they're, it

1368
01:17:14,482 --> 01:17:19,241
seems like they're killing it.
my last interview was with them.

1369
01:17:19,241 --> 01:17:24,000
The last one that was published
before this recording. and they

1370
01:17:24,000 --> 01:17:26,000
Dave Coleman: them well. I'm an
advisor to Rajat.

1371
01:17:26,000 --> 01:17:30,385
Audrow Nash: yeah, Rajat is
awesome. but they're killing it.

1372
01:17:30,385 --> 01:17:34,917
Like they're a company that I
really would bet on and they are

1373
01:17:34,917 --> 01:17:39,229
probably the start of a bunch of
promising companies around

1374
01:17:39,229 --> 01:17:43,980
manipulation, in my opinion. And
we're gonna see more and more of

1375
01:17:43,980 --> 01:17:48,000
them, most likely in the next 10
years say, would be my

1376
01:17:48,000 --> 01:17:51,429
Dave Coleman: Yeah, I agree.
I'm, really excited about all

1377
01:17:51,429 --> 01:17:55,153
the, applications that I, think
are being under, considered for

1378
01:17:55,153 --> 01:17:58,405
robot arms and I see it a lot
because we, talk to these

1379
01:17:58,405 --> 01:18:02,070
companies at least in like kind
of preliminary discovery calls

1380
01:18:02,070 --> 01:18:05,321
on should we collaborate.
There's just, there's so many

1381
01:18:05,321 --> 01:18:08,809
out there that are. Are emerging
that maybe haven't come to

1382
01:18:08,809 --> 01:18:12,356
market. But yeah, chef had a,
that big release, I think last

1383
01:18:12,356 --> 01:18:16,081
week, before big announcement.
So I've been following 'em for a

1384
01:18:16,081 --> 01:18:19,569
couple years. I'm, glad they're
finally outta stealth. it's

1385
01:18:19,569 --> 01:18:23,293
amazing what they've done with,
manipulation and, it looks like

1386
01:18:23,293 --> 01:18:26,367
a standard hardcoded
manufacturing problem, but it's

1387
01:18:26,367 --> 01:18:29,914
not, like the, different styles
of scooping and the variable

1388
01:18:29,914 --> 01:18:33,225
ness level of the different bins
and the different, like

1389
01:18:33,225 --> 01:18:36,772
viscosities and like material
handling of different kinds of

1390
01:18:36,772 --> 01:18:40,378
food. There's a lot of sensing
happening there. There's a lot

1391
01:18:40,378 --> 01:18:43,925
of dynamicism. And so I would
call that a, unstructured or a

1392
01:18:43,925 --> 01:18:47,531
semi-structured problem. And
there is some structure in that,

1393
01:18:47,531 --> 01:18:51,078
where the bins are and like
roughly how deep the bin is. And

1394
01:18:51,078 --> 01:18:54,744
there, there's a lot of things
you do know, but it's, a really

1395
01:18:54,744 --> 01:18:58,231
impressive example of what can
be done in this next wave of

1396
01:18:58,231 --> 01:18:59,000
manipulation.

1397
01:18:59,000 --> 01:19:03,461
Audrow Nash: totally. So I think
we're gonna see more and more of

1398
01:19:03,461 --> 01:19:07,578
that. And I still do not think
we're out of applications for

1399
01:19:07,578 --> 01:19:11,627
mobile robots in general. and
then some mobile manipulators

1400
01:19:11,627 --> 01:19:15,471
that's gonna open up a whole
nother, whole nother set of

1401
01:19:15,471 --> 01:19:19,794
robots, robot applications. So
exciting times I think, ahead of

1402
01:19:19,794 --> 01:19:20,000
us.

1403
01:19:20,000 --> 01:19:24,593
Audrow Nash: Let's see, what do
you, what are, so in my thought,

1404
01:19:24,593 --> 01:19:28,970
man, mobile robots, that's I
don't know, say that in the last

1405
01:19:28,970 --> 01:19:33,060
10 years that's really been
becoming a thing, especially,

1406
01:19:33,060 --> 01:19:37,366
and you're seeing like outdoor
mobile robots and things like

1407
01:19:37,366 --> 01:19:41,815
this, like tons of applications
there. Now we're thinking that

1408
01:19:41,815 --> 01:19:46,264
mobile manipulation is gonna be
such a great opportunity. What

1409
01:19:46,264 --> 01:19:50,282
are some of the challenges of
why it hasn't been in more

1410
01:19:50,282 --> 01:19:51,000
companies?

1411
01:19:51,000 --> 01:19:53,000
Dave Coleman: great question.

1412
01:19:53,000 --> 01:19:57,813
Audrow Nash: what, yeah. and,
then also what has been recently

1413
01:19:57,813 --> 01:20:02,470
solved and then what ones are
looking like they'll be solved

1414
01:20:02,470 --> 01:20:07,283
soon. And like how do you look
at the space, 'cause you have a

1415
01:20:07,283 --> 01:20:10,000
much better idea of this than I
do.

1416
01:20:10,000 --> 01:20:13,466
Dave Coleman: I, kind of wanna
just give more background for a

1417
01:20:13,466 --> 01:20:16,821
second and say there's like kind
of two categories of mobile

1418
01:20:16,821 --> 01:20:20,231
manipulation that we consider,
and we put this on our website

1419
01:20:20,231 --> 01:20:23,530
recently. There's a page on
this, but one category is whole

1420
01:20:23,530 --> 01:20:26,884
body planning, and that's when
you're moving your base, your

1421
01:20:26,884 --> 01:20:30,350
wheels and your arm at the same
time. And that can result in a

1422
01:20:30,350 --> 01:20:33,537
faster cycle time. So the
robot's actually, it's, this is

1423
01:20:33,537 --> 01:20:37,059
what we do as humans. We might
walk by, we're walking outta the

1424
01:20:37,059 --> 01:20:40,469
door to get to our car of our
house, and you grab the keys on

1425
01:20:40,469 --> 01:20:43,712
the coffee table and you don't
even stop walking. You just

1426
01:20:43,712 --> 01:20:47,066
swipe it. And that's where we
wanna get to with, robots. And

1427
01:20:47,066 --> 01:20:50,533
we're working on that. We, have
some examples of that working.

1428
01:20:50,533 --> 01:20:53,775
And then there's the, more
traditional way of you consider

1429
01:20:53,775 --> 01:20:57,242
the navigation of your base as a
separate step before you move

1430
01:20:57,242 --> 01:21:00,540
your arms. So classically you
would fold up your arm into a

1431
01:21:00,540 --> 01:21:03,950
stove position, let's say, and
then you drive acROS the room,

1432
01:21:03,950 --> 01:21:07,305
you stop driving, and then you
unfurl your arms and you grab

1433
01:21:07,305 --> 01:21:10,436
something that's, kinda the
classic way. So we call that

1434
01:21:10,436 --> 01:21:13,007
like navigation plus
manipulation, kind of two

1435
01:21:13,007 --> 01:21:16,082
separate steps. And the reason
why I brought that up is

1436
01:21:16,082 --> 01:21:19,045
because, when you stop and then
re like you scan your

1437
01:21:19,045 --> 01:21:22,120
environment, you use your
cameras to understand things.

1438
01:21:22,120 --> 01:21:25,531
That is easier because you have
to recalibrate basically. You

1439
01:21:25,531 --> 01:21:28,885
don't know exactly where you
stop. The wheels have a lot of,

1440
01:21:28,885 --> 01:21:32,351
skid and, slip, and there's just
you're not exactly sure where

1441
01:21:32,351 --> 01:21:35,203
you ended up. So you have to
have a higher level of

1442
01:21:35,203 --> 01:21:37,998
proprioception or like
self-awareness of where the

1443
01:21:37,998 --> 01:21:41,464
robot is in a room with respect
to a table or to an object you

1444
01:21:41,464 --> 01:21:44,819
wanna manipulate with. And so
all that combines into a, more

1445
01:21:44,819 --> 01:21:48,285
complicated AI and algorithmic
problem, of understanding where

1446
01:21:48,285 --> 01:21:51,304
you are. So that's part of it.
It's just it's, harder,

1447
01:21:51,304 --> 01:21:54,938
especially if you're gonna do it
dynamically where you're driving

1448
01:21:54,938 --> 01:21:56,000
and grabbing at the

1449
01:21:56,000 --> 01:21:58,000
Audrow Nash: that's hard For
sure.

1450
01:21:58,000 --> 01:22:01,297
Dave Coleman: Yeah. And then the
other side of it, I would say

1451
01:22:01,297 --> 01:22:04,381
from a business perspective,
it's just it's a higher, it's

1452
01:22:04,381 --> 01:22:07,518
more expensive robot. And the
more expensive your robot is,

1453
01:22:07,518 --> 01:22:10,602
the harder it's to make the
business case work. So we talk

1454
01:22:10,602 --> 01:22:13,579
to companies all the time who
they're like an innovation

1455
01:22:13,579 --> 01:22:16,610
group. They're doing some
analysis and we're working with

1456
01:22:16,610 --> 01:22:19,800
them and they come back from
their executives and after they

1457
01:22:19,800 --> 01:22:22,778
do the. The cost benefit
analysis, the ROI, it's doesn't

1458
01:22:22,778 --> 01:22:25,968
quite work. the robot's too
expensive and the, and so forth.

1459
01:22:25,968 --> 01:22:29,159
And so that's still what we're
battling. So you said earlier

1460
01:22:29,159 --> 01:22:32,402
that the bases, the AMRs, the
AGVs, that they're pricey. That

1461
01:22:32,402 --> 01:22:35,646
does continue to be a, thing
that suppresses the application,

1462
01:22:35,646 --> 01:22:38,783
whereas like a $30,000 robot
arm, maybe it's 20, maybe it's

1463
01:22:38,783 --> 01:22:41,760
less, that's a lot easier to
drop in as it compared to I

1464
01:22:41,760 --> 01:22:44,632
don't know, an $80,000 mobile
manipulator or a hundred

1465
01:22:44,632 --> 01:22:47,715
thousand dollars system. So
that's, another version of it.

1466
01:22:47,715 --> 01:22:51,172
So there, the sum it up, there's
like the, cost of the robot goes

1467
01:22:51,172 --> 01:22:54,468
up and then the complexity of
the software and the reliability

1468
01:22:54,468 --> 01:22:55,000
goes down.

1469
01:22:55,000 --> 01:22:58,764
Audrow Nash: Definitely. Yeah.
And you might need more

1470
01:22:58,764 --> 01:23:03,154
expensive hardware to do that.
More challenging action, so then

1471
01:23:03,154 --> 01:23:07,197
the cost goes up even further.
Yeah. I think mobile basis,

1472
01:23:07,197 --> 01:23:11,100
that, that case you were
mentioning where you just drive

1473
01:23:11,100 --> 01:23:14,793
up to something and then
manipulate it from like a BA

1474
01:23:14,793 --> 01:23:18,418
mobile Mobility plus,
manipulation. There's a lot of

1475
01:23:18,418 --> 01:23:22,251
applications there still too,
and that's probably a bit

1476
01:23:22,251 --> 01:23:26,502
easier, but, and also you can
just have an arm, but I suppose

1477
01:23:26,502 --> 01:23:30,684
it's probably a bit easier if it
is mobile for it to move to

1478
01:23:30,684 --> 01:23:35,075
whatever the problem is. but one
thing that I was talking about

1479
01:23:35,075 --> 01:23:38,838
with Machina Labs. what they
were doing is, or we were

1480
01:23:38,838 --> 01:23:43,090
talking about robotic arms and
one of the things that came up

1481
01:23:43,090 --> 01:23:47,480
was a robotic arm is probably,
you were mentioning that they're

1482
01:23:47,480 --> 01:23:51,314
cheaper, but it's like an
interesting way to do it that

1483
01:23:51,314 --> 01:23:55,635
makes me think they'll be more
valuable than humanoids in many

1484
01:23:55,635 --> 01:23:59,538
applications is it's the
cheapest, most efficient way to

1485
01:23:59,538 --> 01:24:03,859
move something in space. whereas
if you have the whole rest of

1486
01:24:03,859 --> 01:24:08,250
the robot, you have a bunch of
joints that aren't doing things,

1487
01:24:08,250 --> 01:24:12,222
or maybe you're actively
balancing and it's, or you could

1488
01:24:12,222 --> 01:24:16,752
just have a six foot robotic arm
that is like motor link, for the

1489
01:24:16,752 --> 01:24:20,864
full size of the arm and then it
can just move in the whole

1490
01:24:20,864 --> 01:24:25,325
workspace. And it's, a much more
optimal form factor for doing a

1491
01:24:25,325 --> 01:24:29,437
lot of tasks. And I can imagine
that this is probably true,

1492
01:24:29,437 --> 01:24:33,688
especially in the US as you're
mentioning, where there's very

1493
01:24:33,688 --> 01:24:38,079
good handicap, accessibility in
most public buildings, which is

1494
01:24:38,079 --> 01:24:42,052
probably gonna be the place
you'll see more robots sooner

1495
01:24:42,052 --> 01:24:46,164
because they have more money
versus an individual in a home

1496
01:24:46,164 --> 01:24:47,000
with stairs.

1497
01:24:47,000 --> 01:24:50,540
Dave Coleman: Yeah. Yeah. Yeah.
And there's, also, we, have a

1498
01:24:50,540 --> 01:24:54,138
number of customers using just
gantries, a linear rail. That's

1499
01:24:54,138 --> 01:24:57,621
another way of increasing your
workspace a lot. You can hang

1500
01:24:57,621 --> 01:25:01,277
'em from a ceiling and make it
still very accessible to humans.

1501
01:25:01,277 --> 01:25:04,991
so that, that's another way of,
we add another degree of freedom

1502
01:25:04,991 --> 01:25:08,531
to a robot arm. but I personally
prefer the, wheeled version.

1503
01:25:08,531 --> 01:25:11,955
It's a lot less impact on your
physical building, having to

1504
01:25:11,955 --> 01:25:13,000
bolt a rail to it.

1505
01:25:13,000 --> 01:25:17,624
Audrow Nash: Yeah. Ceilings are
very exciting though for robotic

1506
01:25:17,624 --> 01:25:22,247
applications. Like I'm seeing a
lot, there's a few very good use

1507
01:25:22,247 --> 01:25:26,437
cases and they seem to fit
really good, like security. but

1508
01:25:26,437 --> 01:25:30,844
let's see. I wanted to like, so
starting to wrap up, where do

1509
01:25:30,844 --> 01:25:32,000
you see PickNik?

1510
01:25:32,000 --> 01:25:34,625
Audrow Nash: what do you think
for the next, five years? Where,

1511
01:25:34,625 --> 01:25:36,000
do you think you guys are
headed?

1512
01:25:36,000 --> 01:25:39,501
Dave Coleman: Yeah. love that
question. we're all in on this

1513
01:25:39,501 --> 01:25:43,118
MoveIt Pro platform. I, started
this podcast with, and. pretty

1514
01:25:43,118 --> 01:25:46,269
much going forward, all of our
projects and all of our

1515
01:25:46,269 --> 01:25:49,887
engagements with the customer,
they all are building on top of

1516
01:25:49,887 --> 01:25:52,979
MoveIt Pro. And right now I
mentioned that we do some

1517
01:25:52,979 --> 01:25:56,188
services. So some of our
customers we're really working

1518
01:25:56,188 --> 01:25:59,806
heavily with them to take the
MoveIt Pro platform and apply it

1519
01:25:59,806 --> 01:26:03,423
to their application. we call
those like turnkey solutions. we

1520
01:26:03,423 --> 01:26:06,983
do a lot of that right now. but
where we want to go is making

1521
01:26:06,983 --> 01:26:10,017
this as self-service as
possible. and, there's still

1522
01:26:10,017 --> 01:26:13,809
more work to do there, honestly.
we are a, a growing company. but

1523
01:26:13,809 --> 01:26:17,485
we, wanna get to this point to
where it's like, a gaming engine

1524
01:26:17,485 --> 01:26:21,103
where you just download it off
the internet, maybe watch a few

1525
01:26:21,103 --> 01:26:24,603
video tutorials, and just really
equipping, the next wave of

1526
01:26:24,603 --> 01:26:28,221
roboticists to, be able to do a
lot more than they can do now.

1527
01:26:28,221 --> 01:26:31,897
in a very reliable way that has
a clear path to like deployment

1528
01:26:31,897 --> 01:26:35,164
and a production system, even a
flight system for, space

1529
01:26:35,164 --> 01:26:38,665
applications. So that's, one
dimension of where we're going.

1530
01:26:38,665 --> 01:26:41,816
We, wanna just find more
industries that aren't really

1531
01:26:41,816 --> 01:26:45,317
being served by robotics right
now outside of manufacturing,

1532
01:26:45,317 --> 01:26:49,109
and really change their mindset.
So we, we do a lot of these like

1533
01:26:49,109 --> 01:26:52,610
solution studies as, a service
that we have where we go into

1534
01:26:52,610 --> 01:26:55,761
like large companies and
enterprises and we really sit

1535
01:26:55,761 --> 01:26:58,970
down with them over a series of
meetings. We'll fly out

1536
01:26:58,970 --> 01:27:02,588
sometimes and talk to them about
work where they wanna go with

1537
01:27:02,588 --> 01:27:06,030
automation. And they have these
preconceived notions of the

1538
01:27:06,030 --> 01:27:09,298
limitations of PLCs and
industrial robot arms. And we're

1539
01:27:09,298 --> 01:27:12,740
like, no, we can do way more
than that. And they get really

1540
01:27:12,740 --> 01:27:16,416
excited. And we really emphasize
that we're not here to just do

1541
01:27:16,416 --> 01:27:19,450
it for you, but this is a
collaborative thing. So we

1542
01:27:19,450 --> 01:27:23,126
augment their teams. They, have
their internal automation team.

1543
01:27:23,126 --> 01:27:26,860
We, give them the tools to like
own it themselves. We just kinda

1544
01:27:26,860 --> 01:27:30,128
give 'em a, early push saying,
all right, here's like a,

1545
01:27:30,128 --> 01:27:33,570
prototype and simulation and
then let them run with it. And

1546
01:27:33,570 --> 01:27:36,954
so that's, kinda where we're
going of just like rethinking

1547
01:27:36,954 --> 01:27:40,222
what traditionally we were
limited to, like an area that

1548
01:27:40,222 --> 01:27:43,373
we're excited about its
assembly. and that's, again, a

1549
01:27:43,373 --> 01:27:46,582
little more manufacturing
focused, but there's a lot of

1550
01:27:46,582 --> 01:27:49,908
hard assembly problems that we
think we have the building

1551
01:27:49,908 --> 01:27:53,000
blocks to solve now that weren't
previously possible.

1552
01:27:53,000 --> 01:27:57,236
Audrow Nash: Oh, that's awesome.
Yeah. Assembly is an awesome

1553
01:27:57,236 --> 01:27:58,000
area. Okay.

1554
01:27:58,000 --> 01:28:01,139
Dave Coleman: And then I guess
the last thing I wanna say is we

1555
01:28:01,139 --> 01:28:04,078
wanna have robots building moon
bases. And that robot's not

1556
01:28:04,078 --> 01:28:07,167
launched yet, but we're working
hard on it. I can't wait to be

1557
01:28:07,167 --> 01:28:09,808
on a podcast in the future
talking about how our, our

1558
01:28:09,808 --> 01:28:12,000
robots have built a, human
habitat up there.

1559
01:28:12,000 --> 01:28:15,561
Audrow Nash: That's awesome.
Hell yeah. Yeah. I can't wait

1560
01:28:15,561 --> 01:28:19,000
for that too. Alright, thank you
Dave. This was a blast.

1561
01:28:19,000 --> 01:28:22,268
Dave Coleman: Yeah. Thank you
Audrow. This is, an honor to be

1562
01:28:22,268 --> 01:28:25,000
here again and, thanks for all
the great questions.

1563
01:28:25,000 --> 01:28:28,000
Audrow Nash: All right. Bye
everyone.

1564
01:28:28,000 --> 01:28:29,000
Dave Coleman: See you.

1565
01:28:28,000 --> 01:28:31,781
Audrow Nash: So we've journeyed
through the world of robotics

1566
01:28:31,781 --> 01:28:35,561
with Dave Coleman from PickNik
Robotics, and I hope you're as

1567
01:28:35,561 --> 01:28:38,660
amazed as I am by the
versatility and potential of

1568
01:28:38,660 --> 01:28:42,069
robotic arms. We've touched on
how they're not just for

1569
01:28:42,069 --> 01:28:45,601
assembly lines or manufacturing
anymore. They're becoming

1570
01:28:45,601 --> 01:28:48,948
integral in dynamic and
unstructured environments like

1571
01:28:48,948 --> 01:28:52,419
farms, hospitals, and even
space. Now I leave you with a

1572
01:28:52,419 --> 01:28:56,199
question. Where do you think
robotics could make its next big

1573
01:28:56,199 --> 01:28:59,670
leap? Could it be in home
automation, disaster response,

1574
01:28:59,670 --> 01:29:02,769
or maybe just in art and
entertainment? Thanks for

1575
01:29:02,769 --> 01:29:05,000
listening, and until next time,
bye.
