Curiosity-Driven Intrinsic Motivation Cognitive Model in Perception-Action Loop

Abstract

To study the cognitive problems for autonomous robotic pigeon in an unknown environment, a curiosity-driven intrinsic motivation cognitive model in the perception-action loop is proposed. The model simulates the cognitive mechanism of intrinsic motivation in psychology, through the process of critic, probabilistic behavior selection, and orientation updating, which can implement the behavior cognition of an intelligent agent. By the introduction of curiosity, which is the main element of intrinsic motivation, the exploration degree at different times with the same state can be updated, and the certainty is increased during the cognitive process. Measure by information entropy illustrates that the application of the proposed cognitive model can achieve better cognition for behavior selection. By simulation experiment, we verify the adaptability of the proposed model in a changing environment, which also reflected the cognitive processes similar to biological cognition, and the model is proved to be effective.