Learning Rat-Like Behavioral Interaction Using a Small-Scale Robotic Rat

In the existing robot-rat interaction, robots usually exert influence as stimuli to observe the response of target rats. However, the above single-stimulation model from robot to rat lacks a two-way communication. Therefore, we proposed a method to learn the rat-like behavioral interaction. First, we constructed two behavior patterns in the interaction process between the two rats: the individual pattern and the interaction pattern. We divided the roles of the two rats into active stimulator and passive receiver. The stimulator executes the individual pattern, and the receiver switches between the two patterns. Which pattern the receiver executes depends on the probability of interaction. Secondly, we proposed a hypothesis concerning the behavioral interaction between two rats and controlled the interaction process between two robots. Information entropy was used to measure the similarity between two robots and two rats. The results proved that the probability of interaction between rats depends on the distance of centroid and average relative velocity, and we obtained the optimal solution to express the relationship between them. In the future, the improvement of robots' interaction ability with rats will highly benefit from this study.