Reinforcement learning for discernment behavior acquisition

In this study, we propose an active perception model that autonomously learns discernment behaviors. Discernment behavior, which is a type of exploratory behaviors that support object feature extraction, is a fundamental tool for a robot to orientate itself, operate objects and establish higher classes of knowledge. In this model, a robot learns the discernment behaviors through the interaction with multiple objects. While the interaction, the robot takes reinforcement signal according to the cluster distance of the observed data. We applied the proposed model to a mobile robot simulation to confirm the effectiveness. In this simulation, three different shaped objects were placed beside the robot one by one. After the learning, the robot acquired different behaviors according to each object. Our investigation for behavioral patterns showed the acquisition of intelligent behavioral strategies, which are related to the object shapes. Thus, the proposed model effectively established intelligent strategies according to the relation between object features and robot's configuration.