5683
5683
5683
5683
5683
5683
5683
5683
5683 How do folks talk when 5683 they’re underwater? With physique language, 5683 in fact.
5683
5683 Marine environments current a singular 5683 set of challenges that render 5683 a number of applied sciences 5683 that have been developed for 5683 land purposes utterly ineffective. Speaking 5683 utilizing sound, or at the 5683 very least as folks use 5683 sound to speak, is one 5683 among them.
5683
5683 Michael Fulton tackles this problem 5683 along with his presentation at 5683 ICRA 2022 by utilizing physique 5683 language to speak with an 5683 AUV underwater. Tune in for 5683 extra.
5683
5683
5683 His poster may be seen 5683 5683 right here 5683 .
5683
5683 Michael Fulton
5683
5683 Michael Fulton is a Ph.D. 5683 Candidate on the College of 5683 Minnesota Twin Cities. His analysis 5683 focuses totally on underwater robotics 5683 with a deal with purposes 5683 the place robots work with 5683 people. Particularly, human-robot interplay and 5683 robotic notion utilizing laptop imaginative 5683 and prescient and deep studying, 5683 with the intent of making 5683 techniques that may work collaboratively 5683 with people in difficult environments.
5683
5683
5683 transcript
5683
5683
5683 Abate: 5683 [00:00:00] So inform me 5683 just a little bit about 5683 your presentation earlier right now.
5683
5683 Michael Fulton: 5683 Yeah, so I used 5683 to be presenting right now, 5683 my collaborative work with Jungseok 5683 Hong and my advisor Junaed 5683 Sattar on diver strategy. So 5683 mainly the issue of when 5683 you may have an AUV 5683 and a diver working collectively 5683 underwater it’s necessary that they 5683 be shut collectively once they 5683 wanna talk, whether or not 5683 it’s for, you already know, 5683 doing gestures to the AUV, 5683 to inform it, you already 5683 know, Go do that activity, 5683 go take a look at 5683 this space.
5683
5683 Or if it’s the AUV 5683 speaking to the diver, perhaps 5683 they’re telling it, Hey, I 5683 discovered this cool factor over 5683 right here. You must come 5683 test it out in both 5683 of these conditions, you want 5683 to be shut collectively, proper? 5683 Mm-hmm nevertheless, for AUVs to 5683 be helpful underwater, they should 5683 go away the diver. They 5683 should go do looking out 5683 and, you already know, carrying 5683 merchandise or, or instruments and 5683 supplies and stuff like that.
5683
5683 Uh, so that is the 5683 issue that we now have, 5683 proper. We have to be 5683 shut to speak, however we 5683 have to be distant to, 5683 to do stuff. So to 5683 repair this, we’d like a 5683 functionality for diver strategy. We 5683 want to have the ability 5683 to seek for the diver, 5683 discover them and strategy them 5683 to an applicable distance and 5683 orientation for communication.
5683
5683 So our algorithm known as 5683 ADROC autonomous diver relative operator 5683 configuration. And it’s this monocular 5683 imaginative and prescient based mostly 5683 methodology of doing this the 5683 place we we do that 5683 diver strategy based mostly on 5683 solely monocular imaginative and prescient. 5683 Yeah. As a result of 5683 we wished to maintain it 5683 as low cost as doable, 5683 you already know? No, no 5683 sonar, no stereovision and, and 5683 as minimal sensing as we 5683 might, we might handle this 5683 with and mainly the way 5683 in which the algorithm works 5683 is as a substitute of 5683 attempting to do monocular depth 5683 estimation, which is, you already 5683 know, you will get respectable 5683 accuracy on it, however you 5683 generally want excessive computational energy.
5683
5683 Mm-hmm . As a substitute 5683 of doing that, we realized, 5683 okay, what we truly have 5683 to know is, is the 5683 space that the divers is 5683 presently at “Ok”? Is it 5683 shut sufficient for, for us 5683 to work with the communication 5683 a part of issues.
5683
5683 Abate: 5683 So that you want 5683 a tough estimate?
5683
5683 Michael Fulton: 5683 Yeah. You want, you 5683 want a really tough common 5683 estimate. I don’t care if 5683 the, if the robotic’s, you 5683 already know, one meter away 5683 or 1.1, you already know, 5683 0.9 0.7.
5683
5683 It doesn’t actually matter to 5683 me so long as it’s 5683 shut sufficient. Yeah. Tough sufficient. 5683 So the way in which 5683 that we did that is 5683 by utilizing shoulder width as 5683 a previous piece of knowledge, 5683 as a result of we 5683 all know from biomedical literature 5683 that there’s a variety that 5683 human shoulder widths are available 5683 in. We all know the 5683 common of that vary.
5683
5683 We all know, you already 5683 know, the place most individuals’s 5683 shoulder widths are fairly near. 5683 From that we will calculate 5683 the anticipated pixel width between 5683 shoulders for a detailed sufficient 5683 tough estimate, distance for communication. 5683 Yeah. After which we simply 5683 evaluate: is the diver shoulder 5683 width smaller than that? Okay. 5683 We have to come nearer.
5683
5683 Is it, is it bigger 5683 than that? Okay. We have 5683 to again up. And the 5683 way in which we do 5683 the the precise calculation of 5683 the shoulder width is a 5683 two-step course of. We both 5683 use a diver detector, which 5683 takes a picture of a, 5683 of the scene and finds. 5683 Diver attracts a bounding field 5683 round them. We are able 5683 to use the width of 5683 that as sort of a 5683 proxy for shoulder width.
5683
5683 Mm-hmm however it’s not tremendous 5683 correct, proper? The diver might 5683 be sort of on their 5683 aspect. Yeah. Uh, there’s a 5683 number of issues that may 5683 change the bounding field width 5683 with out altering shoulder width. 5683 In order that will get 5683 us a really, very tough 5683 estimate. And if we simply 5683 approached based mostly on that, 5683 the, the AUV can be 5683 method off on distance as 5683 a result of the bounding 5683 field adjustments rather a lot.
5683
5683 What doesn’t change rather a 5683 lot is the precise shoulder 5683 width that is still. So 5683 we additionally use the diver 5683 pose estimation algorithm to get 5683 key factors on the shoulders 5683 and calculate the space between 5683 them. Yeah. And so it’s 5683 this cascaded strategy the place 5683 mainly what finally ends up 5683 taking place is from distant, 5683 the detector works.
5683
5683 We’ve truly run this so 5683 far as 15 meters away. 5683 Um, and that allows you 5683 to heart the diver within 5683 the picture and begin getting 5683 nearer to them. After which 5683 as you get nearer throughout 5683 the vary of, I might 5683 say most likely about six 5683 to seven meters is the 5683 efficient vary. Uh, you’ll be 5683 able to truly begin detecting 5683 the important thing factors for 5683 the shoulders and then you 5683 definately get correct distance.
5683
5683 Not distance estimation, however distance 5683 ratio calculation, we name this 5683 the pseudo distance. Yeah. Trigger 5683 it’s not likely distance, however 5683 it features at it. Yeah.
5683
5683 Abate: 5683 So I imply, one 5683 of many good issues that 5683 you just mentioned in your 5683 presentation is that even in 5683 several poses and orientations, the 5683 house between your shoulders stays 5683 comparatively the identical.
5683
5683 However on the flip aspect, 5683 say my shoulders and your 5683 shoulders are completely different lengths.
5683
5683 Michael Fulton: 5683 They’re completely different. However 5683 while you take a look 5683 at the magnitude of the 5683 distinction in comparison with the 5683 magnitude of the scene, it’s 5683 truly very small. Proper. Like, 5683 I might say simply on 5683 a tough guess, I’d say 5683 the distinction between our shoulder 5683 width is a couple of 5683 centimeters mm-hmm proper.
5683
5683 And while you have been 5683 utilizing this, I can’t keep 5683 in mind my precise shoulder 5683 width. It was one thing 5683 like 40 one thing centimeters. 5683 I, I don’t keep in 5683 mind once we’re utilizing that 5683 as our, as our, mainly 5683 our sign for the space 5683 a distinction of a few 5683 centimeters does make a distinction, 5683 however it doesn’t wreck issues.
5683
5683 Yeah. We are able to 5683 nonetheless work with it. And, 5683 and like I mentioned, within 5683 the, within the presentation earlier, 5683 we will run it off 5683 of the common diver shoulder 5683 width. However if you’re taking 5683 place with an AUV and 5683 you already know, you’re gonna 5683 work [00:05:00] with it, you 5683 possibly can additionally calibrate it 5683 to your precise shoulder width. 5683 We did this a couple 5683 of occasions and it really 5683 works.
5683
5683 The algorithm works regardless if 5683 you happen to calibrate it 5683 to your precise shoulder width, 5683 you will get very nice 5683 distance like last distance for 5683 strategy. It really works actually 5683 properly if you happen to 5683 calibrate it to the precise 5683 shoulder width, however it works 5683 usually on the common as 5683 effectively. Is there any distinction
5683
5683 Abate: 5683 between say taking these 5683 these measurements and pictures above 5683 floor versus underwater. Does water 5683 distort that measurement?
5683
5683 Michael Fulton: 5683 Yeah, so completely underwater 5683 imaginative and prescient normally. there’s 5683 distortion of coloration. There’s distortion 5683 of turbidity particulate matter and 5683 bubbles, a number of issues. 5683 So, so this aspect of 5683 underwater imaginative and prescient is 5683 sort of it’s the method 5683 it’s.
5683
5683 Mm-hmm all underwater imaginative and 5683 prescient stuff suffers from this. 5683 There’s a, a very vigorous 5683 thread of labor on underwater 5683 picture. Improve. Which principally makes 5683 an attempt to take care 5683 of like gentle or coloration 5683 altering coloration. Yeah. Yeah. Um, 5683 so that truly, it helps 5683 a bit, however doesn’t assist 5683 a ton with this. Um, 5683 the opposite huge factor. So, 5683 in order that’s from the 5683 visible aspect of issues.
5683
5683 After we’re speaking extra concerning 5683 the I don’t know fairly 5683 learn how to say this. 5683 The, the, the educational aspect 5683 of issues, our diver detector 5683 is educated on photographs of 5683 divers, so it is aware 5683 of what they appear like. 5683 It approaches them straightforward. The 5683 physique pose that we use 5683 is TRT pose from nvidia 5683 IOT. it’s educated on terrestrial 5683 imagery. So the factor about 5683 that’s that in these conventional 5683 photographs, persons are standing or 5683 sitting, no person is sideways, 5683 proper? Cuz we, we will’t 5683 go sideways, however within the 5683 water we will, persons are 5683 sideways on a regular basis.
5683
5683 They’re swimming, they’re floating. And 5683 so this truly causes issues 5683 with ADROC. Um, If, if 5683 any individual is in a, 5683 a vastly completely different orientation 5683 it, it, it’s rather a 5683 lot tougher, which is why, 5683 you already know, if you 5683 happen to learn the paper, 5683 you’ll see, we, we made 5683 a few simplifying assumptions. Certainly 5683 one of them was that 5683 there’s just one diver within 5683 the scene as a result 5683 of whereas we’re trying into 5683 discriminating between divers proper now, 5683 the algorithm doesn’t try this.
5683
5683 So, and it’ll strategy whichever 5683 one, it sees first . 5683 Um, the opposite simplifying assumption 5683 that we made was that 5683 the diver is usually upright. 5683 You realize, we didn’t inform 5683 folks, you must keep 100% 5683 straight up and down, however 5683 we mentioned, you already know, 5683 keep principally upright. Yeah. And 5683 once we tried it on 5683 folks, you already know, sideways, 5683 it nonetheless does work, however 5683 not as effectively.
5683
5683 Abate: 5683 Yeah. So that is 5683 an space that’s like, you’ll 5683 be able to undoubtedly see 5683 a path to enchancment.
5683
5683 Michael Fulton: 5683 Completely
5683
5683 Abate: 5683 not likely a problem. 5683 It’s only a matter of 5683 getting the info and becoming 5683 it to yeah.
5683
5683 Michael Fulton: 5683 With underwater robotics, brown 5683 reality is at all times 5683 an enormous, enormous bother. And 5683 for labeling one thing like 5683 pose. That’s some actually it’s, 5683 it’s not a lot that 5683 it’s like tough work, however 5683 the labeling is gonna take 5683 months for that.
5683
5683 However I truly, I imply, 5683 it, this is the reason 5683 ICRA is nice. Like I 5683 used to be speaking with 5683 any individual on Monday night 5683 time or no sat Sunday 5683 night time. Um, and so 5683 they have been telling me 5683 about some pose community I 5683 ought to attempt. So I’m 5683 gonna go residence and take 5683 a look at, attempt it 5683 for our knowledge and see 5683 if it really works any 5683 higher.
5683
5683 Abate: 5683 Yeah.
5683
5683 Michael Fulton: 5683 Um, I believe the 5683 2 foremost areas of enchancment, 5683 three, three areas of enchancment, 5683 pose estimation, we already talked 5683 about.
5683
5683 Yeah. Second huge one is 5683 search conduct. Our search conduct 5683 for this was actually easy. 5683 For those who don’t see 5683 the diver flip mm-hmm proper, 5683 however there’s, there’s some apparent 5683 enhancements that may be made 5683 there. Issues like if we 5683 lose monitor of the diver, 5683 we must always flip within 5683 the route that we final 5683 noticed them.
5683
5683 Proper. Or if we’re attempting 5683 to cowl a big house, 5683 perhaps turning isn’t gonna be 5683 sufficient. You realize, I, I 5683 mentioned earlier, we, we ran 5683 this from 15 meters away. 5683 I might guess… I don’t 5683 have knowledge. I might guess 5683 that previous 30 meters it’s 5683 not gonna work as a 5683 result of we simply can’t 5683 see something. So for an 5683 area that’s like 30 meters 5683 or bigger, which open water 5683 underwater environments are you’re gonna 5683 want to have the ability 5683 to do extra than simply 5683 turning.
5683
5683 It’s gonna want to love 5683 search the house one way 5683 or the other. Yeah. That 5683 I believe is the entire 5683 huge factor by itself. Um, 5683 after which the opposite huge 5683 factor by itself is what 5683 I mentioned earlier about diver 5683 discrimination. Yeah. Having the ability 5683 to inform the distinction between 5683 diver a and diver B, 5683 you already know, I don’t, 5683 I don’t actually care if 5683 it’s, you already know, this 5683 man versus that man versus 5683 that woman.
5683
5683 It doesn’t matter who particularly, 5683 however I do need the 5683 algorithm to have the ability 5683 to handle a number of 5683 divers within the scene, figuring 5683 out which one it’s … 5683 approached earlier than. And, and 5683 once we truly first got 5683 here up with this concept, 5683 the concept was we’re gonna 5683 activate the robotic and it’s 5683 gonna like go as much 5683 as all people and ask, 5683 Hey, are you my operator?
5683
5683 I actually wish to do 5683 that also. So if we 5683 get the diver discriminator working 5683 effectively sufficient,
5683
5683 Abate: 5683 And that will likely 5683 be via gestures, they’ll say 5683 like, …
5683
5683 Michael Fulton: 5683 yeah. So, so it’ll 5683 come as much as the 5683 diver and it’ll do like 5683 a, so I I’ve performed 5683 this work with movement based 5683 mostly communication, robotic communication by 5683 way of. um, and it, 5683 so the di the robotic’s 5683 gonna come up and it’s 5683 gonna sort of do like 5683 a, you ever seen like 5683 a canine ask to play 5683 fetch with you?
5683
5683 Yeah. It’s gonna sort of 5683 go like, Hey, Hey, Hey, 5683 Hey, Hey, are you, are 5683 you? Yeah. After which the 5683 diver will say sure or, 5683 or no, I’m not your, 5683 I’m not your operator. After 5683 which it’ll go, okay, I’ll 5683 cross you off the checklist 5683 seek for the following [00:10:00] 5683 individual. Yeah. That’s the place 5683 this work hopefully goes sooner 5683 or later. Um, you already 5683 know, my, my work normally, 5683 my thesis work is about 5683 robotic communication and interplay underwater.
5683
5683 Uh, I believe I discussed 5683 this briefly within the speak, 5683 you already know, underwater human 5683 robotic collaboration is a model 5683 new subject. Yeah. Like this 5683 didn’t exist earlier than the 5683 early two hundreds. Um, partially 5683 as a result of the 5683 AUVs which can be cheap 5683 to, to work with underwater 5683 are like, since 2000’s,
5683
5683 Abate: 5683 they have been, they 5683 have been created within the 5683 2000’s.
5683
5683 Michael Fulton: 5683 Sure.
5683
5683 Abate: 5683 And that was the 5683 impetus for why now working 5683 with a robotic, proper. Underwater 5683 is even an idea that 5683 we’re speaking about.
5683
5683 Michael Fulton: 5683 Sure. Trigger the primary 5683 AUV’s are in just like 5683 the sixties, and these are 5683 these huge ocean going submarine, 5683 issues which can be for 5683 oceanography, nice work, you already 5683 know, actually necessary stuff, however 5683 they’re larger than you and 5683 I are.
5683
5683 Yeah. And you’ll, you’ll be 5683 able to work together with 5683 that, however it’s not likely 5683 what they’re for due to 5683 this fact doing these lengthy 5683 deployments that people can’t do. 5683 We’re now in, in underwater 5683 robotics, seeing the, the appearance, 5683 the approaching of collaborative AUV’s. 5683 It’s, it’s a new factor 5683 that’s arising and you’ll see 5683 it within the work, you 5683 already know, underwater HRI papers 5683 weren’t written 20 years in 5683 the past.
5683
5683 Um, perhaps any individual wrote 5683 one 20 years in the 5683 past that I don’t learn 5683 about and so they’re gonna 5683 get mad at me, however 5683 I’ve solely seen ones relationship 5683 again to early two hundreds. 5683 Um, and now there’s, there’s 5683 a couple of right here 5683 and there. I’ve introduced a 5683 few ICRA now, and whereas 5683 we’re not but on the 5683 level the place the AUVs 5683 and the persons are truly 5683 working collectively you already know, 5683 I, I, I don’t know 5683 of anyone who’s truly doing 5683 collaborative work with AUVs for 5683 like an organization.
5683
5683 Um, however it’s coming. Yeah, 5683 it’s coming quickly. And, and 5683 specifically, for me, I’m actually 5683 interested by like environmental conservation 5683 and organic remediation. So like 5683 trash cleanup, oil spills uh, 5683 observing invasive or so it’s 5683 both eradicating invasive species or 5683 preserving endangered species.
5683
5683 Yeah. This type of factor 5683 the place what’s taking place 5683 proper now’s around the globe. 5683 Some scientist is diving, you 5683 already know, they’re diving with 5683 all these undergrads for hours 5683 lengthy a day. I need 5683 to have the ability to 5683 give them robots which can 5683 be low cost and, and 5683 brazenly out there. And you 5683 already know, my huge a 5683 part of it’s robots that 5683 they will talk with in 5683 a method that’s not onerous 5683 for them to study.
5683
5683 Yeah. I don’t need these 5683 scientists to must study Python 5683 or must study C++ or 5683 ROS and learn to program 5683 these robots. I need them 5683 to have the ability to 5683 use my communication frameworks, and 5683 my activity administration frameworks in 5683 order that they will activity 5683 these AUVs with completely different 5683 items. Work go discover me 5683 this, the sort of Marine 5683 life.
5683
5683 Go discover me this trash. 5683 Inform me the place to 5683 go decide up this trash. 5683 Uh, deliver me instruments, carry 5683 samples for me. Yeah. This 5683 type of stuff I believe 5683 may be very a lot 5683 throughout the realm of chance 5683 and the work that I, 5683 and the opposite nice Ph.D. 5683 college students and grasp college 5683 students and undergrad college students 5683 and our advisor of the 5683 interactive robotics and imaginative and 5683 prescient lab do is actively 5683 transferring us in direction of 5683 that.
5683
5683 Yeah. We’re getting, you already 5683 know, notion, capabilities, and navigation 5683 mapping. Capabilities you noticed within 5683 the Marine, robotics talks, all 5683 these various things. You realize, 5683 the acoustic localization, the GoPro-based 5683 imaginative and prescient for mapping 5683 all these things. It’s all 5683 items of the puzzle. And 5683 the piece that I’m most 5683 interested by is the human-robot 5683 interplay half as a result 5683 of it’s, it’s such an 5683 attention-grabbing, difficult atmosphere.
5683
5683 There’s so many assumptions that 5683 you just make terrestrially that 5683 simply aren’t there. Like the 5683 massive, the. Know, if you 5683 happen to’re speaking with a 5683 robotic, you sort of anticipate 5683 to speak to it and 5683 have it speak again. You 5683 possibly can’t try this underwater. 5683 You gotta,
5683
5683 Abate: 5683 yeah. There’s no voice.
5683
5683 Michael Fulton: 5683 There’s no voice. There’s 5683 a respiration equipment in your 5683 mouth.
5683
5683 Yeah. And you’ll hear, however 5683 not likely effectively. Yeah. So 5683 I’ve developed, you already know, 5683 movement, light-based communication. I’m attempting 5683 sound, however nonverbal sounds so 5683 like tones as a substitute 5683 of phrases.
5683
5683 Abate: 5683 Yeah. And what’s attention-grabbing 5683 too, is like as in 5683 there are a number of 5683 business examples like offshore wind 5683 and like offshore buildings which 5683 can be being constructed the 5683 place The divers aren’t gonna 5683 get changed.
5683
5683 Michael Fulton: 5683 No, no. Very quickly 5683 quickly.
5683
5683 Abate: 5683 Yeah. They’ve such an 5683 extremely tough job to automate. 5683 Sure. That, and due to 5683 that, they’re additionally there, a 5683 few of laborious to search 5683 out yep. Should be costly. 5683 Yep. Um,
5683
5683 Michael Fulton: 5683 it’s harmful too
5683
5683 Abate: 5683 and harmful.
5683
5683 Michael Fulton: 5683 Yeah. Folks die yearly.
5683
5683 Abate: 5683 So that you don’t, 5683 you, we wish to do 5683 all the things you’ll be 5683 able to to make that 5683 dive probably the most environment 5683 friendly model of themselves doable.
5683
5683 Michael Fulton: 5683 And protected and, and 5683 simpler. Yeah. You realize, it’s, 5683 it’s, it’s laborious, work. It, 5683 such as you mentioned, it’s 5683 laborious to search out individuals 5683 who do that as a 5683 result of there’s a number 5683 of scuba dive licensed folks, 5683 proper?
5683
5683 It’s a, it’s a standard 5683 pastime, however technical diving and 5683 diving for, for business functions. 5683 There’s not too lots of 5683 them on the market. There’s. 5683 I imply, [00:15:00] in, in, 5683 within the grand scheme of 5683 issues, you already know, it’s, 5683 it’s, it’s a rarer subject 5683 and a lot necessary work 5683 is, is in there. Uh, 5683 there’s this quote, I actually. 5683 um, it’s a, I, I, 5683 I don’t know if it’s 5683 truly, it’s attributed to Leonardo 5683 DaVinci water is the driving 5683 drive of all life on 5683 our planet.
5683
5683 Mm-hmm I actually imagine that. 5683 Like, clearly there’s the, the 5683 scientific causes, you already know, 5683 photosynthesis, local weather local weather 5683 stuff, but additionally identical to 5683 a lot commerce depends upon 5683 ocean environments, the web. I 5683 imply, we now have cables 5683 underneath sea, all of these 5683 things. You want AUVs. There 5683 are some locations the place 5683 we wanna exchange divers with 5683 AUV’s.
5683
5683 However we actually wanna increase 5683 the divers who’re presently doing 5683 work underwater with AUVs, with 5683 these collaborative AUVs, partially since 5683 you’re proper. It’s gonna be 5683 a very long time earlier 5683 than they’re changed if ever 5683 it’s such a difficult subject, 5683 but additionally personally, I’m, I, 5683 I actually like the concept 5683 of robots making folks’s lives 5683 higher.
5683
5683 Mm-hmm and generally changing them 5683 in jobs is the way 5683 in which in direction of 5683 that. There are some jobs. 5683 So harmful, so boring, so, 5683 so soiled that you just 5683 don’t need anyone to do 5683 them, however there’s a number 5683 of jobs the place like, 5683 folks rely on this for 5683 his or her livelihood. I 5683 don’t wanna exchange these folks. 5683 I wanna make their lives 5683 simpler.
5683
5683 I wanna make their lives 5683 simpler and I wanna make 5683 it doable for them to 5683 do extra attention-grabbing work. You 5683 realize, there’s we take into 5683 consideration, we consider ourselves as 5683 such a sophisticated society, proper? 5683 Like we go to house, 5683 we go to Mars, a 5683 ridiculous quantity of our ocean 5683 is unexplored. We don’t understand 5683 how a lot of the 5683 life that exists in our 5683 ocean is. We don’t, we 5683 there’s a lot primary science 5683 there that’s undone as a 5683 result of the atmosphere is 5683 so inhospitable.
5683
5683 You want air tanks, there’s 5683 strain concerns. There’s a most 5683 restrict you’ll be able to 5683 dive to. So something that 5683 you just’re doing underwater is 5683 mechanically 100 occasions tougher, 100 5683 occasions extra expensive, extra effortful.
5683
5683 And that is the place 5683 AUVs, my advisor mentioned this 5683 actually, rather well within the 5683 session. So we wish to 5683 improve underwater divers by having 5683 underwater divers do the issues, 5683 AUVs can’t and having AUVs 5683 do the issues underwater divers 5683 can’t. Yeah, I believe that’s 5683 an ideal summation of the 5683 place this subject is headed.
5683
5683 Superior. Thanks. Yeah, no drawback. 5683 Thanks for asking me.
5683
5683
5683
5683 transcript
5683 tags: 5683 bio-inspired 5683 , 5683 c-Analysis-Innovation 5683 , 5683 cx-Analysis-Innovation 5683 , 5683 human-robot interplay 5683 , 5683 podcast 5683 , 5683 Service Skilled Underwater 5683 , 5683 software program 5683 , 5683 video
5683
5683
5683
56835683
5683 Abate De Mey 5683
Founding father of 5683 Fluid Dev, Hiring Platform for 5683 Robotics
5683
5683
5683 5683
5683
5683
5683
5683 5683
Abate De Mey
Founding father 5683 of Fluid Dev, Hiring Platform 5683 for Robotics
5683
5683
5683 5683
5683
5683
5683