New algorithm improves robot movement adaptation speed

Nature : Instead of attempting to preprogram robots with solutions to all possible problems, such as failed joints or broken extremities, researchers are attempting to have the robots find a solution themselves. In 2006 a hexapod robot was developed that could identify problems with its legs and create a new way of moving to compensate for the “injury,” but it took quite a while to do so. Now Jean-Baptiste Mouret of the French Institute for Research in Computer Science and Automation and his colleagues have simplified the process for the robot so it can find a new movement pattern in under one minute. The robot now has an on-board camera to identify where a problem is, but it does not attempt to diagnose what is wrong. Mouret and his team also include a library of preselected gaits, which they say is similar to an animal’s innate knowledge. The library helps the robot rule out nonsensical movements, which significantly reduces the amount of time the robot needs to fix the problem. The improvements allow the robot to compensate not just for the loss of a limb but also for different terrains such as slippery or uneven floors.