Cooperative Transportation by Humanoid Robots – Solving Piano Movers' Problem

Abstract

Recent developments in humanoid robots seem very promising for the sake of achieving higher-level tasks, such as those usually done by humans. Cooperative behavior allows multiple robots to tackle a task easily and effectively. In this paper, we try to solve a cooperative transportation task, i.e., a piano movers' problem, using two humanoid robots. We use Q-learning for the purpose of real-world adaptation. More precisely, the two robots gain environmental information through their respective monocular cameras, and learn to cooperatively navigate a narrow L-shaped aisle. We empirically show how the robustness of the acquired cooperative behavior for the test cases differed from the training L-shaped aisle.