[1L4-J-11] Robot and real worlds: learning of robots

[1L4-J-11-03] Integrated Cognitive Model for Robot Learning of Concepts, Actions, and Language

〇Kazuki Miyazawa¹, Tatsuya Aoki¹, Takato Horii¹, Takayuki Nagai^1,2 (1. The University of Electro-Communications, 2. Osaka University)

Keywords:Cognitive Model, Symbol Emergence in Robotics, Concept Formation, Stochastic Model

It is expected to realize an intelligent robot that coexists with us and supports our lives. In order to realize such a robot, it is necessary to act by understanding the surrounding environment and language in the real world. For such a complex understanding, multiple cognitive functions are required. The purpose of this study is to propose an integrated cognitive model for robots that simultaneously learn concepts, actions, and language. This can be achieved by integrating various modules. The proposed model consists of three modules: concepts formation using multimodal information, grammar learning, and reinforcement learning. We verify the potential of the proposed model through experiments using a real robot. We show that the proposed model enables robots to form concepts, make decision, and understand language.