Achieving Different Levels of Adaptability for Human – Robot Collaboration Utilizing a Neuro-Dynamical System

Collaborative robots are expected to acquire adequate planning ability when achieving complex tasks such as cooking and cleaning with humans in dynamically changing daily-life environment. To fulfill adequate planning, three levels of adaptability—motion modification, action selection, and turn-taking—should be considered. This study demonstrates that a single hierarchically organized neuro-dynamical system called multiple timescale recurrent neural network (MTRNN) can achieve these levels of adaptability by utilizing the socalled multiple timescale property. The system is implemented in a humanoid robot and the robot is required to collaborate with a human partner by sharing a specific task under dynamically changing environment. Experimental results show that in both learned and unlearned situations, the robot can generate adequate behaviors against different situations, and the aforementioned levels of adaptability can be realized by a single system.