Researchers at the Massachusetts Institute of Technology (MIT) have developed a control algorithm that allows shape-shifting soft robots to autonomously learn and complete specific tasks. These robots can dynamically change their entire bodies, making traditional learning methods ineffective. By using a coarse-to-fine methodology, the algorithm controls groups of adjacent muscles and optimizes action plans based on a simulated environment. The approach, which outperformed other algorithms, has the potential to enable general-purpose robots capable of adapting to various tasks.
