Adaptive Controls of Personified Locomotive Agents

An autonomous control methodology is proposed for locomotive agents to have lifelike personified behaviors. The control introduces a pattern generator that is constructed by connecting oscillators composed of mutually inhibited neurons. The rhythmic signal of the pattern generator drives a proportional derivative controller that determines the legs joint angles of a locomotive agent, and the signal of the controller is fed back to the oscillator by which the physical parameters of the controller are utilized for stably and smoothly controlling the signals. The motions of bodily components can be autonomously synchronized by mutually feeding their oscillator’s state signals. The controller’s signals are intentionally distorted and fluctuated so as to accurately simulate the behaviors of a real human. Keyword: Pattern generator, neural oscillator, locomotive agent, adaptive stochastic motion controls