A control theoretic perspective on human motor primitives

© 2013 Rhys Llewellyn DavisBased on experimental data, the concept of motor primitives has been introduced to understand how humans control movement. From a physiological point of view, a motor primitive is a predefined muscle activation sequence that is stored in the spinal cortex and that can be initiated by descending motor command signals. It is an elementary building block in generating movement. The motor primitive model posits that the brain controls movement through timed activation of multiple primitives. Further, the model suggests that human movement can be reduced in complexity by representing all movement as weighted activation of sets of motor primitives. This thesis explores a control perspective of motor primitives in order to deepen our understanding of the human motor system. In particular the notions of reachability and stability of motion are analysed. Two specific and competing implementations of the motor primitive model are revisited from a control theory perspective. It is shown that both can be presented in a single framework. The particular implementations are then compared through simulations in terms of human motor performance (accuracy, robustness, smoothness and resolution). This thesis proposes extensions to the two specific implementations that impact their simulated performance and then investigates a generalized method for reducing the computational complexity of the motor primitive model. This generalized method is called primitive abstraction and is a unique contribution of this thesis. Finally this thesis outlines future research opportunities and discusses extension of primitive abstraction into motor learning