Human-friendly arm robot using interactive particle swarm optimization for trajectory generation

This work deals with human-friendly trajectory generation of an arm robot. Various methods for the trajectory generation have been proposed so far, but robots must deal with environments including human operators. In this situation, the robot should take a suitable action/motion to the individual operator. This work applies an interactive particle swarm optimization for the trajectory generation using human evaluation. Particle Swarm Optimization offer multiple sets of candidate trajectory and then the best one is shown to the human to be evaluated. Basically human evaluation is very important for generating robotic behavior, but the detail of the human evaluation is not clear. Furthermore, to reduce the number of the human evaluations, a state-value function is used. Therefore, we can search for good trajectory candidates with the estimated human evaluation. The experimental results show that the state-value function can estimate the human evaluation.