Reflex-tutored motor control

Reflex-tutored motor control

The mammalian nervous system is developed in such a way that it achieves motor performance that artificial systems strive for. An important feature of this process is the differential development of the system’s components. That is, long before the cortex takes over control of the body, spinally-generated movements such as reflexes dominate early behaviour and operate with little if any cortical supervision. During this period the cortex is already activated by sensory inputs resulting from the early, spinal movements. The mechanism through which the cortex uses these inputs for its motor development, and how a similar process could be used for the benefit of artificial learning motor systems remains unexplored. The aim of this project is to develop a theoretical framework that will unveil the computational mechanisms that sensory inputs enable in the biological motor system through its differential development, and to translate these mechanisms into an artificial neural network implemented on a silicon neuromorphic motor controller of a robot.