An hybrid approach for robots learning folding tasks

We tackle the problem of using cooperative manipulators to perform towel folding tasks. Differently from other recent approaches, our method executes what we call a momentum fold a swinging motion that exploits the dynamics of the manipulated object. We propose a new learning algorithm that combines imitation and reinforcement learning. Human demonstrations are used to reduce the search space of the reinforcement learning algorithm, which then converges quickly. The strengths of the algorithm come from its efficient processing, fast learning capabilities, absence of an object model, and applicability to other problems exhibiting temporally incoherent parameter spaces. Experiments were performed on a robotic platform, demonstrating the algorithm’s capability.