Behavioral Optimization in a Robotic Serial Reaching Task Using Predictive Information

Prediction is a powerful approach to minimize errors and control problems in familiar environments tasks. In human motor execution of sequential action, context effects can be observed, such as anticipation or predictive movement toward target objects, where later subactions are affected by the earlier subactions. this article, we present simulation framework for serial reaching task using 4-DoF robotic arm examine learning simulated reinforcement agents. As demonstrate, giving agents access information about future object’s identity results motion optimization, next modulates Specifically, learn anticipate predict location object, move it before appears, thus achieving higher rewards than that were not given information.