Transfer Learning Method Using Ontology for Heterogeneous Multi-agent Reinforcement Learning

This paper presents a framework, called the knowledge co-creation framework (KCF), for heterogeneous multiagent robot systems that use a transfer learning method. A multiagent robot system (MARS) that utilizes reinforcement learning and a transfer learning method has recently been studied in realworld situations. In MARS, autonomous agents obtain behavior autonomously through multi-agent reinforcement learning and the transfer learning method enables the reuse of the knowledge of other robots’ behavior, such as for cooperative behavior. Those methods, however, have not been fully and systematically discussed. To address this, KCF leverages the transfer learning method and cloud-computing resources. In prior research, we developed ontology-based inter-task mapping as a core technology for hierarchical transfer learning (HTL) method and investigated its effectiveness in a dynamic multi-agent environment. The HTL method hierarchically abstracts obtained knowledge by ontological methods. Here, we evaluate the effectiveness of HTL with a basic experimental setup that considers two types of ontology: action and state. Keywords—Transfer learning; Multi-agent reinforcement learning; Multi-agent robot systems