Automated learning of muscle-actuated locomotion through control abstraction

Abstract

We present a learning technique that automatically syn- thesizes realistic locomotion for the animation of physics-based models of animals. The method is especially suitable for animals with highly flexible, many-degree-of-freedom bodies and a consid- erable number of internal muscle actuators, such as snakes and fish. The multilevel learning process first performs repeated loco- motion trials in search of actuator control functions that produce efficient locomotion, presuming virtually nothing about the form of these functions. Applying a short-time Fourier analysis, the learn- ing process then abstracts control functions that produce effective locomotion into a compact representation which makes explicit the natural quasi-periodicities and coordination of the muscle actions. The artificial animals can finally put into practice the compact, efficient controllers that they have learned. Their locomotion learn- ing abilities enable them to accomplish higher-level tasks specified by the animator while guided by sensory perception of their vir- tual world; e.g., locomotion to a visible target. We demonstrate physics-based animation of learned locomotion in dynamic models of land snakes, fishes, and even marine mammals that have trained themselves to perform SeaWorld stunts.