Coupling Robots Behavior by Introducing Reactive Motivational Orientations

Recent descriptions of the determinants of human, social behavior from psychology have detailed key impulsive mechanisms that describe unconscious and omniscient influence from others over the flow of action selection. This influence can be represented by introducing the concept of Motivational Orientation (MO), where behaviors and stimuli are associated and classified to have either an approach or avoidance orientation. Approach-oriented elements promote the decreasing of distances, augmenting interactions with the environment. Inversely, avoidance oriented elements promote behaviors that increase distances, decrementing interactions. We hypothesize that, as for humans, the introduction of a reactive mechanism integrating the notion of MO on an artificial agent controller enhances the agent's adaptability to the environment, promoting the emergence of group behavior, beneficial on tasks where collaboration changes performance. Here, we propose a reactive portion of a reactive-deliberative robot controller that incorporates MOs, meant to be coupled to a task-specific value-based deliberative controller. The proposed architecture is first evaluated on a free exploration task, demonstrating that behavior is coherently modulated by the environment increasing behavioral variability and producing group effects when the MO is influenced by other agents. Then, we show that performance can be improved by this group effect confronting the agents to an object-seeking task.