Risk-Averse Control via CVaR Barrier Functions: Application to Bipedal Robot Locomotion

Enforcing safety in the presence of stochastic uncertainty is a challenging problem. Traditionally, researchers have proposed statistical mean as measure for systems subject to uncertainty. However, ensuring only reasonable if system's safe behavior large number runs interest, which precludes use practical scenarios. In this letter, we propose risk sensitive notion called conditional-value-at-risk (CVaR) safety. We introduce CVaR barrier functions tool enforce CVaR-safety and conditions their Boolean compositions. Given legacy controller, show that can design minimally interfering CVaR-safe controller via solving difference convex programs (DCPs). elucidate method by applying it bipedal robot locomotion case study.