A study of reinforcement learning with knowledge sharing
Kazuyuki Ito, Y. Imoto, Hideaki Taguchi, Akio Gofuku · Institutional Repositories DataBase (IRDB) · 2004
In this paper, we consider multi-agent system in which every agents have own tasks that differs each other. We propose a method that decreases learning time of reinforcement learning by using the model of environment. In the proposed algorithm, the model is created by sharing the experiences of agents each other. To demonstrate the effectiveness of the proposed method, simulations of a puddle world and experiments of a maze world have been carried out. As a result effective behaviors have been obtained quickly.