Combining Static and Runtime Methods to Achieve Safe Standing-Up for Humanoid Robots

Francesco Leofante, Simone Vuotto, Erika Ábrahám, Armando Tacchella, Nils Jansen

Abstract: Due to its complexity, the standing-up task for robots is highly challenging, and often implemented by scripting the strategy that the robot should execute per hand. In this paper we aim at improving the approach of a scripted stand-up strategy by making it more stable and safe. To achieve this aim, we apply both static and runtime methods by integrating reinforcement learning, static analysis and runtime monitoring techniques.