Creation of Robot's Tool-Body Assimilation Model Using Sparse Recurrent Neural Network

2015 Fiscal Year Final Research Report

• Project/Area Number: 25730159
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Intelligent robotics
• Research Institution: The University of Tokushima (2014-2015); Kyoto University (2013)
• Principal Investigator: Nishide Shun 徳島大学, ソシオテクノサイエンス研究部, 講師 (30613400)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: 再帰結合神経回路モデル / ロボット / スパース結合

Outline of Final Research Achievements

In this research, we investigated the effectiveness of sparse connections using Multiple Timescales Recurrent Neural Network (MTRNN), for applying to robot's motion learning. From experiments with nine types of human motions, sparse networks were confirmed to achieve better training results, compared to full-connected networks. As examples of motion learning, we created models for robot's self body learning and developmental imitation drawing learning model.

Free Research Field

認知発達ロボティクス