Abstract

Optimal control theory has become a dominant mathematical framework for studying human movement, but major gaps remain in the understanding of how such control schemes are implemented in neural hardware. In particular, most of modern control theory assumes that control laws are implemented using a perfect, centralized, and infinitely fast computer. In contrast, a human must implement control using a distributed network of slow, noisy neurons. Motivated by the presence of delays in the human motor system, I will discuss the architecture of optimal decentralized controllers when communication between subsystems is limited by delays. I will present state feedback and output feedback control schemes that are, to my knowledge, the first known explicit solutions to some classical problems in decentralized control. In the case of decentralized state feedback, the structure that emerges as the result of optimization resembles a management hierarchy. When output feedback is considered, however, the hierarchical structure is less apparent. Throughout the talk, I will attempt to relate the architectural results to the organization of the motor system.