Emotionally influenced coordination of behaviors for autonomous mobile robots

Abstract

A model of amygdala is used for an emotional intervention on a behavior coordination mechanism for autonomous mobile robot navigation. Behaviors such as goal following and obstacle avoidance ones are a priori derived by Q-learning. A collision free goal following behavior is carried out by combining of previously learned behaviors. The basic mechanism of this behavior coordination is based on simple summation of corresponding action value functions. The artificial emotion mechanism is proposed to weigh the different action value functions in the summation process. The proposed behavior coordination method improves the performance of the navigation system with respect to its obstacle avoidance capability. The results are confirmed by simulations.