Researchers at Carnegie Mellon University's Robotics Institute have taught robots to be clever when it comes to dealing with extraneous appendages.
Many robots have more limbs than they actually need—often because there’s an expectation that at some point, one or more of those limbs will stop working and the others will need to compensate. A robot can be programmed in advance to walk on whatever legs it has left.
But as robots become more modular and easier to physically reconfigure, it becomes more useful to have a generalized system to dynamically generate gaits and transitions on the fly. For their recent project, presented at IROS 2017 in Vancouver, Canada, CMU researchers programmed robots to switch limbs from locomotive to non-locomotive tasks. Their new gait-generating method makes it possible, for instance, to 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.
While the work is preliminary, the researchers are planning on extending their method to include dynamic gaits, such as running and jumping; generalizing to other morphologies like bipeds and tripeds; and, ultimately, to respond to high-level task requests. An example would be telling the robot to “go over there and pick up that thing”; the robot would then choose how best to make that happen, regardless of how many legs it has to work with.
For the full story, visit Clever Modular Robots Turn Legs Into Arms on Demand on IEEE Spectrum.
