Adaptive human reaching movements : modelling of sub-movement motor learning : sensory delays and curl field perturbations

Details

Supervisors

Crevecoeur, Frédéric

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil biomédical

Abstract

When we need to perform a goal-directed hand movement, our brain must compute the adequate command signals to send to the muscles of the arm in order to accurately reach the target. For this purpose, our brain uses an internal representation of the biomechanics of the upper limb, as well as of the external environment. However, in the event of an early exposure to an unexpected perturbation during the execution of the movement, the brain must quickly react by determining the nature of the perturbation and adapting the control signals accordingly. Motor learning is the ability of the human brain to learn new representations of the external environment when performing a motor task. Computationally, motor learning has been studied over movement repetitions, using the framework of optimal feedback control theory and in particular the Linear-Quadratic-Gaussian controller. In this work, we perform numerical simulations of such perturbed reaching movements with the hypothesis of "sub-movement" motor learning, namely that the central nervous system is able to compensate for the unexpected disturbance during the course of the movement. We start by reproducing the results of the recent work of Crevecoeur et al. \cite{crevecoeur2018sub} that uses a Least Squares identification technique in order to estimate the intensity of the disturbance during the execution of the movement. Afterwards, motivated by the biological likelihood, we introduce a temporal delay in the sensory feedback loop of the model. Our results show that the numerical implementation of the model used in this work can successfully reproduce results in agreement with the hypothesis of online motor learning, even when sensory motor noise and temporal feedback delays are taken into account. Furthermore, the model can also be flexibly adapted to more general force fields, such as a curl force field. Altogether, these advances will help to develop more realistic mathematical models of the motor system, allowing at the same time a better understanding of the motor function.