DARPA Robotics Challenge Finals June 5-6

[GKrotkov]

Originally Posted by asid61

This is super cool. Are they required to use human-like robots, or is that just the easiest way to do things?

It's just the answer that most of them came up with. It may help that the robot provided for teams that do not wish to build their own, ATLAS, is humanoid.

Also, one of the requirements is the "Ability to manipulate and use a diverse assortment of tools designed for humans"


What's interesting is that this challenge might illustrate how close this technology is, even though it looks halting and slow now. The DARPA urban challenge was in 2007 (footage: https://www.youtube.com/watch?v=SQFEmR50HAk), on known ground and with only other robots on the road; but now Uber is buying all the autonomy experts around and Google is thinking about going to market with an autonomous car! Perhaps robotic autonomy and dealing with variable situations isn't quite so far away as we might think.

[RomeroFRC5012]

A few members from 5012 will be their on their own! Sounds Fun!

[AlexanderTheOK]

Originally Posted by GKrotkov

Also, one of the requirements is the "Ability to manipulate and use a diverse assortment of tools designed for humans"

I'm still a bit perplexed as to why so many teams decided to go bipedal. Every task here seems to be navigable with the use of an appropriately modified tracked drivetrain, and I'm sure an appropriate humanoid and dexterous arm can be mounted on said drivetrain.

I'm imagining it has more to do with funding than the solution. I'd think "look our robot is basically a person!" attracts more money than a generic looking rover. I'd really like to know from the teams however.

[mman1506]

Originally Posted by AlexanderTheOK

I'm still a bit perplexed as to why so many teams decided to go bipedal. Every task here seems to be navigable with the use of an appropriately modified tracked drivetrain, and I'm sure an appropriate humanoid and dexterous arm can be mounted on said drivetrain.

I'm imagining it has more to do with funding than the solution. I'd think "look our robot is basically a person!" attracts more money than a generic looking rover. I'd really like to know from the teams however.

I'm not really up to date with things but there usually are a few robots that can switch between tracked and bipedal.

[TheHolyHades1]

The robot who is currently in the lead (per day 1 standings, anyway), CHIMP from Tartan Rescue (CMU), can switch between bipedal and quadruped. Each of the "feet" are tracks, which helps mostly eliminate the need to solve a balancing problem.

[BenGuy]

Originally Posted by AlexanderTheOK

I'm still a bit perplexed as to why so many teams decided to go bipedal. Every task here seems to be navigable with the use of an appropriately modified tracked drivetrain, and I'm sure an appropriate humanoid and dexterous arm can be mounted on said drivetrain.

I'm imagining it has more to do with funding than the solution. I'd think "look our robot is basically a person!" attracts more money than a generic looking rover. I'd really like to know from the teams however.

Tracked and bipedal is the best solution (in my opinion anyways .) There are a few a challenges (like climbing stairs) that require legs, but, tracks are better for generally traveling through other challenges, like traversing the runway.

[AlexanderTheOK]

Originally Posted by BenGuy

Tracked and bipedal is the best solution (in my opinion anyways .) There are a few a challenges (like climbing stairs) that require legs, but, tracks are better for generally traveling through other challenges, like traversing the runway.

if you google search "Stair Climbing robot" and go into "videos" literally NONE of the videos on the first page are bipedal. They all use drive systems that are effective both on flat ground and wonky stairs/debris.

Tracked + biped sorta makes sense. It's a good way to keep your robot bipedal while keeping the benefits of being able to just roll around.

[BenGuy]

Originally Posted by AlexanderTheOK

if you google search "Stair Climbing robot" and go into "videos" literally NONE of the videos on the first page are bipedal. They all use drive systems that are effective both on flat ground and wonky stairs/debris.

Tracked + biped sorta makes sense. It's a good way to keep your robot bipedal while keeping the benefits of being able to just roll around.

If you google that, the results are all of very small robots that can do nothing but climb stairs, and their systems are specifically for climbing stairs.

The team in first right now has a tracked + bipedal system btw...

[AlexanderTheOK]

Originally Posted by BenGuy

If you google that, the results are all of very small robots that can do nothing but climb stairs, and their systems are specifically for climbing stairs.

Correction. They are all proofs of concepts of drivebases that can climb stairs. All of these can be enlarged and can have 6DOF arms mounted to them.

Originally Posted by BenGuy

The team in first right now has a tracked + bipedal system btw...

Yes. I am well aware of that, and in fact, acknowledged in my post that it was a very good design decision considering that there are likely other constraints that forced teams to make legged robots.


Edit: This doesn't look very small by the way.

[BenGuy]

Originally Posted by AlexanderTheOK

Correction. They are all proofs of concepts of drivebases that can climb stairs. All of these can be enlarged and can have 6DOF arms mounted to them.



Yes. I am well aware of that, and in fact, acknowledged in my post that it was a very good design decision considering that there are likely other constraints that forced teams to make legged robots.


Edit: This doesn't look very small by the way.

Your EXAMPLE is a robot that can't do anything but climb stairs. Try getting a robot with tracks to drive the atv, you need two legs to do that, as well as traverse the ground and getting in and out of it.

[Electronica1]

Talking to the members of my schools team. The reason that humanoid is because they are built to operate in areas built for humans. Driving a car, walking up stairs, opening a door are just some of the things that are built for humans. Just because the DRC checked for certain tasks, doesn't mean they are the only tasks they will go against in a real world application.

[SenorZ]

The bipedal, humanoid form has its benefits. Chief among them is the ability to put on an enemy's uniform after you knock him out during a covert op.

[AlexanderTheOK]

Originally Posted by BenGuy

Your EXAMPLE is a robot that can't do anything but climb stairs. Try getting a robot with tracks to drive the atv, you need two legs to do that, as well as traverse the ground and getting in and out of it.

Well, a leg is A method of pushing a pedal, another method would be a linear actuator mounted in a specific location. Also, wheeled robots can get out of the car, as one evidently has.

My point is that while the humanoid robots CAN do each of these tasks, (well, some of them at least) there is no logical design decision that leads to a bipedal conclusion when focusing on the tasks in this competition.

This doesn't mean there isn't a good reason and everyone is just crazy and wasting money. in fact, Electronica1 has already explained that the robots were less designed for the competition and more for the real world. It just means that I am confident that a robot designed to complete this range of tasks (along with those in the DRC trials of course) would not be bipedal.

I would, however, have to disagree that a bipedal design would be beneficial in any human environment. All of the quadrupeds in my home seem to be much more adept at climbing the stairs than any of the bipeds. I'm glad to have seen a few of those in the finals.

[JesseK]

I thought there was supposed to be a ladder at the end, not stairs. When I first read the challenge years ago, I thought 'yea! that'll push the frontier in mobility!'. Did they change that at some point, or am I misremembering?

Disasters have a wonderful history of taking out stairs. Robots can't exactly take stairs with them to climb something. If a robot finds stairs in-situ it can't move them to where it wishes they were. If stairs are detached they usually break up into pieces and therefore can't be re-used. If a single stair is missing most robots simply fall through the stairs. If stairs are slippery they're basically impassible for a robot, especially if the railing is gone. If stairs are partially submerged, then it's basically game over for most robots, even if they're waterproof and float. In the world of moving up and down, stairs are to FRC arms as ladders are to FRC elevators: both work, but the simpler version is remarkably more effective in the vast majority of scenarios.

Maybe next year's FRC game should include a ladder climbing function, but without the arbitrary silliness of the 2013 rules. That'll show DARPA there's no need to water down the challenge. Maybe the ladder is contained behind a barrier, and the only way to get into the zone is to duck through a hole, or something.

Alex, I'm with you on the quadrupeds. They can tackle stairs and ladders equally with ease.

[Nate Laverdure]

Originally Posted by JesseK

I thought there was supposed to be a ladder at the end, not stairs.

"Ladder-climbing" was part of the DRC Trials in 2013, but (1) they looked more like steep industrial stairs like the kind you'd see between two different roof levels and (2) for some reason they morphed into standard stairs for the Finals.