Abstract/Summary

One of the most debated processes of motor control is how biological systems solve the degrees of freedom problem, using the redundancy of the motor system advantageously. Optimisation strategies and control theory result in models that are task specific and limited at generalising. In this thesis, motor redundancy is addressed from a biologically inspired perspective, developing a model for spinal motor control in dynamic field theory (DFT). Empirical studies show that patterns of muscle activation contribute to coordination, emerging from the integration of cortical and spinal sensorimotor information. These patterns, called motor primitives, are at the core of the proposed model and are represented as attractors in two-dimensional neural fields. Weighted by cortical activations and combined with sensory feedback carrying the position of the end effector, primitives are combined in a resultant force field, associated to the motor plan. This process is formalised by a control law that, giving the forces at the joints of the manipulator, allows for the direct simulation of the forward dynamics. Other processes in the spinal cord are modelled, including a neural controller for the autonomous development of the task, an adaptive threshold enabling stable representations of motor features, and synaptic nodes converting neural representations into motor variables used to calculate the forces at the next time-step. Results show a generalised reaching repertoire, emerging from a few motor primitives and successful straight trajectories, with unimodal velocity profiles. Introducing two-dimensional traveling peak solutions as elemental behaviour shows how the developed methodology can be used to add physiologically inspired elements to cognitive robotics. The findings of this thesis connect existing models in DFT to biomechanical accounts based on motor primitives, resulting in a fully embodied account for motor control and opening the way for a unified framework to understand hand-eye coordination and develop bio-inspired robotics.