CHANNELS

Aerospace

Automotive

Automated & Autonomous Vehicles

Battery & Electrification Technology

Defense

Drone TV

Electronics

Manufacturing & Prototyping

Materials

Medical

Motion Control

RF & Microwave Technology

Robotics & Automation

Sensors & IoT

Test & Measurement

Aerospace
Automotive
Automated & Autonomous Vehicles
Battery & Electrification Technology
Defense
Drone TV
Electronics
Manufacturing & Prototyping
Materials
Medical
Motion Control
RF & Microwave Technology
Robotics & Automation
Sensors & IoT
Test & Measurement
Aerospace & Defense
Search Results

Robust-Legged Robot Can Do (Almost) Anything

Researchers have designed a robotic system that enables a low-cost and relatively small-legged robot to climb and descend stairs nearly its height, traverse varied terrain, cross gaps, and even operate in the dark.

"Empowering small robots to climb stairs and handle a variety of environments is crucial to developing robots that will be useful in people's homes as well as search-and-rescue operations," said Deepak Pathak  , an assistant professor in the Robotics Institute. "This system creates a robust and adaptable robot that could perform many everyday tasks."

Topics:
Automation Robotics

More From SAE Media Group

    Transcript

    00:00:00 so the goal here in this particular application is to make legged robots work any kind of terrain previously we have had shown results on legged robots which were blind and they could work any kind of terrain from just their own body sensing but if you truly want legged robots to walk in any kind of scenario any kind of setting walk upstairs walk downstairs walk over rocks

    00:00:22 a blind policy cannot do that and the robot has to make use of vision and what is in front of it and this is what this project is about so this uses Vision directly as input to Output command to the motors of the robot and what is nice about this is that vision and control both are integrated together as a single system there is no

    00:00:44 separate Vision part making plans and map and then robot part making work over the plan it's a single system which takes direct camera images input captures that and outputs Motors command there is no mapping or planning involved in that process this allows the system to be very robust in the real world like if it slips on stairs it can recover and walk up it can

    00:01:07 go to unseen scenarios and adapt to those settings now why does this Vision as a first class citizen is a nice thing because if you were to make follow the classical pipeline make maps and plans using Vision then you will always have some kind of error in making map like your camera May vibrate as robot is walking things may fall around and your map will be noisy but here we directly

    00:01:30 use the camera image so robot has the freedom to really go out of the box and and adapt it motors from a raw sensory input so it's and this motivation here is uh it's very biologically inspired in the sense that vision and control even in humans as well as other animals is very tightly covered like it's very hard to balance yourself on a single leg if you close your eyes

    00:01:54 otherwise it's trivial but if you close your eyes it becomes very hard to balance yourself on a single leg which seems uh very confusing because standing does not even require any vision it's because Vision helps you in so many ways we do not even realize that in our body and this is what this robot is doing here vision and control both are tightly coupled into a single system

    00:02:15 what makes this application very nice is this is a very small robot it's very safe it's meant to be put in homes and help people so it's a very safe and small robot but then the problem with small robots is that any normal size stair or any obstacle that is normal for humans that becomes like a very high obstacle for this small robot because the height of the robot at a normal

    00:02:38 terrain is like 28 centimeter the stairs in your home are like 20 centimeter maybe slightly higher so like 24 so you are already reaching 80 percent of the height of the robot what you can call State climbing for humans is actually obstacle climbing for this robot like when human climb hurdles since the robot is learning or seeing only from egocentric vision

    00:02:59 if you notice it climbing over obstacles its rear legs do not actually see the present picture they have to remember from their memory as to how to move the rear leg if you go to YouTube and find cat videos over walking obstacles you will realize that cats look in the front and their real legs move as if the rear legs have a camera in them it is because you can

    00:03:21 remember from memory how to move your legs despite the fact you are few seconds ahead of them this is what the same robot does there is no map no planning what it does it remembers the terrain from the memory as to how to move its rear leg so in that sense it's much more closer to this biological way of thinking about control