A Survey on Policy Search Algorithms for Learning Robot Controllers in a Handful of Trials

• Published in: IEEE transactions on robotics 2020-04, Vol.36 (2), p.328-347
• Main Authors: Chatzilygeroudis, Konstantinos, Vassiliades, Vassilis, Stulp, Freek, Calinon, Sylvain, Mouret, Jean-Baptiste
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial Intelligence; Automatic; Autonomous agents; Biological system modeling; Computer Science; Computer simulation; Computing time; Data models; Dynamic models; Engineering Sciences; Europe; Heuristic algorithms; learning and adaptive systems; Machine learning; micro-data policy search (MDPS); Optimization; Robot kinematics; robot learning; Robotics; Robots; Search algorithms; Simulators; Trajectory
• ISSN: 1552-3098; 1941-0468
• DOI: 10.1109/TRO.2019.2958211

Most policy search (PS) algorithms require thousands of training episodes to find an effective policy, which is often infeasible with a physical robot. This survey article focuses on the extreme other end of the spectrum: how can a robot adapt with only a handful of trials (a dozen) and a few minutes? By analogy with the word "big-data," we refer to this challenge as "micro-data reinforcement learning." In this article, we show that a first strategy is to leverage prior knowledge on the policy structure (e.g., dynamic movement primitives), on the policy parameters (e.g., demonstrations), or on the dynamics (e.g., simulators). A second strategy is to create data-driven surrogate models of the expected reward (e.g., Bayesian optimization) or the dynamical model (e.g., model-based PS), so that the policy optimizer queries the model instead of the real system. Overall, all successful micro-data algorithms combine these two strategies by varying the kind of model and prior knowledge. The current scientific challenges essentially revolve around scaling up to complex robots, designing generic priors, and optimizing the computing time.