Biopsychically inspired cognitive control for autonomous agents based on motivated learning

A biopsychically inspired integrated cognitive control system architecture for autonomous agents is proposed in this paper, whose hardware is modeled after the biological Central Nervous System (CNS) of human's, and algorithms are modeled after human psychological processes. The pain-driven system architecture and motivated learning algorithms allow the agent to set goals autonomously so as not only to survive but also thrive in uncertain or unknown environment and they endow the agent with the ability of improvisation, innovation and imagination. Familiar concepts in control engineering will be reexamined in the context of human cognition and psychological behaviors. The control law for decision making derived from psychological process is exemplified with case studies of three psychological behaviors by the method of Linear Quadratic Regulator (LQR) with a state-dependent quadratic loss function. This paper is mainly an inquiry outside of the box of conventional control paradigms; therefore it provides more questions than answers.