Task Feasibility Maximization using Model-Free Policy Search and Model-Based Whole-Body Control

Ryan Lober (1) , Olivier Sigaud (1) , Vincent Padois (1, 2)

1 ISIR - Institut des Systèmes Intelligents et de Robotique
2 AUCTUS - Augmenting human comfort in the factory using cobots

Producing feasible motions for highly redundant robots, such as humanoids, is a complicated and high-dimensional problem. Model-based whole-body control of such robots, can generate complex dynamic behaviors through the simultaneous execution of multiple tasks. Unfortunately, tasks are generally planned without close consideration for the underlying controller being used, or the other tasks being executed, and are often infeasible when executed on the robot. Consequently, there is no guarantee that the motion will be accomplished. In this work, we develop an optimization loop which automatically improves task feasibility using model-free policy search in conjunction with model-based whole-body control. This combination allows problems to be solved, which would be otherwise intractable using simply one or the other. Through experiments on both the simulated and real iCub humanoid robot, we show that by optimizing task feasibility, initially infeasible complex dynamic motions can be realized --- specifically, a sit-to-stand transition.

Keywords: iCub humanoid robot; Humanoids; Policy Search; Whole-body control; Reinforcement learning

Domains: Robotics [cs.RO]

Ryan Lober, Olivier Sigaud, Vincent Padois. Task Feasibility Maximization using Model-Free Policy Search and Model-Based Whole-Body Control. Frontiers in Robotics and AI, 2020, 7, ⟨10.3389/frobt.2020.00061⟩. ⟨hal-01620370v3⟩