Clever Modular Robots Turn Legs Into Arms on Demand

Lots of legs is good, but slightly fewer legs plus a couple of arms can be more useful at times

3 min read
This CMU modular robot turns legs into arms on demand
CMU’s new gait-generating method means you can teach a dodecapod robot to transition into a nonapod robot that can carry stuff with two arms while using a third to point a camera.
Photo: Carnegie Mellon University

Robots that can be physically reconfigured to do lots of different things are, in theory, a great way to maximize versatility while saving time and effort. In theory. The idea is that you could have a robot with a bunch of different limbs that would be arms or legs depending on what worked best at the time, but in practice, this means coming up with entirely new gaits along with a way to transition from the old gait to the new one. You could, if you had a lot of time to kill and nothing better to do, pre-compute every possible combination of gaits and transitions in advance, but who would want to do that when you could instead “create new gaits online to enable rapid deployment minutes after reconfiguration.” Okay, yeah, that may not sound super exciting, but it means you can teach a dodecapod robot to transition into a septapod robot that can carry stuff with two arms while using a third to point a camera.

CMU researchers have developed a method to allow robots with multiple limbs to switch them from locomotive to non-locomotive tasks. That is, turn them from legs into arms that can manipulate things

There are plenty of robots out there that have more legs than they need, often because there’s some expectation that at some point one (or more) of those legs will cease to function and the other legs will have to compensate. If this happens, the robot is programmed to change its gait to walk on whichever legs it has left. Programmed in advance, that is, which is fine, except that as robots get more modular and easier to physically reconfigure, it becomes more and more useful to have a generalized system that can dynamically generate gaits (and transitions between gaits) on the fly no matter what the leg configuration of your robot happens to be.

While dealing with dead legs is certainly useful, a recent project by CMU researchers Julian Whitman, Shuang Su, Stelian Coros, Alex Ansari, and Howie Choset, focuses on something different: allowing robots with multiple limbs to switch them from locomotive to non-locomotive tasks. That is, turn them from legs into arms that can manipulate things.

Here’s how they explain it:

We demonstrated our approach to gait and transition generation with a hexapod and a dodecapod robot. We emphasize the utility of our method for redundant locomotors, like our robots, where limbs can be reassigned simultaneously to peripheral tasks while still leaving enough limbs for quasistatic locomotion. In each demonstration, the robot begins with all limbs in locomotive roles. We choose a subset of limbs to be reassigned, and create a new gait. We then create a transition that maintains robot heading, orientation, and speed. These examples show different sets of limbs used to pick up and carry objects or to position a camera while the robots locomote.

As useful as this seems, it’s still very preliminary, and there’s a lot more to be done. The researchers are planning on extending their method to include dynamic gaits, which means things like (we hope) running and jumping, and they’re also going to generalize to other morphologies like bipeds and tripeds. Ultimately, the idea is to develop a system that you can feed high level tasks to—something like “go over there and pick up that thing”—and the robot will choose how best to make that happen, no matter how many legs it starts (or finishes) with.

“Generating Gaits for Simultaneous Locomotion and Manipulation,” by Julian Whitman, Shuang Su, Stelian Coros, Alex Ansari, and Howie Choset from Carnegie Mellon University, was presented this week at IROS 2017 in Vancouver, Canada.

The Conversation (0)

How the U.S. Army Is Turning Robots Into Team Players

Engineers battle the limits of deep learning for battlefield bots

11 min read
Robot with threads near a fallen branch

RoMan, the Army Research Laboratory's robotic manipulator, considers the best way to grasp and move a tree branch at the Adelphi Laboratory Center, in Maryland.

Evan Ackerman

This article is part of our special report on AI, “The Great AI Reckoning.

"I should probably not be standing this close," I think to myself, as the robot slowly approaches a large tree branch on the floor in front of me. It's not the size of the branch that makes me nervous—it's that the robot is operating autonomously, and that while I know what it's supposed to do, I'm not entirely sure what it will do. If everything works the way the roboticists at the U.S. Army Research Laboratory (ARL) in Adelphi, Md., expect, the robot will identify the branch, grasp it, and drag it out of the way. These folks know what they're doing, but I've spent enough time around robots that I take a small step backwards anyway.

The robot, named RoMan, for Robotic Manipulator, is about the size of a large lawn mower, with a tracked base that helps it handle most kinds of terrain. At the front, it has a squat torso equipped with cameras and depth sensors, as well as a pair of arms that were harvested from a prototype disaster-response robot originally developed at NASA's Jet Propulsion Laboratory for a DARPA robotics competition. RoMan's job today is roadway clearing, a multistep task that ARL wants the robot to complete as autonomously as possible. Instead of instructing the robot to grasp specific objects in specific ways and move them to specific places, the operators tell RoMan to "go clear a path." It's then up to the robot to make all the decisions necessary to achieve that objective.

Keep Reading ↓ Show less