Safe Learning in Robotics

A great deal of research in recent years has focused on robot learning. In many applications, guarantees that specifications are satisfied throughout the learning process are paramount. For the safety specification, we present a controller synthesis technique based on the computation of reachable sets using optimal control. We show recent results in system decomposition to speed up this computation, and how offline computation may be used in online applications. We then present a method combining reachability with machine learning, which uses approximate knowledge of the dynamics to provide a least-restrictive, safety-preserving control law which intervenes only when the computed safety guarantees require it, or when confidence in the computed guarantee decays in light of new observations. We will illustrate these methods on a quadrotor UAV experimental platform which we have at Berkeley.