抄録
We have proposed technique, named Bayesian-discrimination-function-based Reinforcement Learning (BRL). BRL can segmentalize continuous state and action spaces. Because of this, we succeeded acquisition of cooperative behavior in Multi-Robot System under a real environment with noise and uncertainty. In that experiment, robots didn't understand who their partners are. So in this paper, we aim to acquisition of cooperative behavior between robots and a human by BRL. A task is that robots keep a pallet upright and lift up it with a human. We examine the robustness of BRL through the change of an environment by shift to another human.
