Human lower limb movement prediction based on optimal control

Wang, Weida (2024). Human lower limb movement prediction based on optimal control. University of Birmingham. Ph.D.

Wang2024PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (4MB)

Abstract

The movement behaviour of the human body can be predicted by solving an optimal control problem, however, many limitations exist, for example, only certain movement behaviour such as level walking has been undertaken, and these predictions are often limited to able-bodied individuals. This thesis aims to advance existing predictive simulation studies, using optimal control theory to achieve prediction for both disabled individuals and able-bodied individuals.
There are significant technical challenges in predicting sit-to-stand motion for people with lower limb amputations, making it rare for researchers to delve into this field. The predictive study of sit-to-stand motion for the unilateral transtibial amputees completed in this thesis is the first in this field. This study not only modelled seat contact, but also created a novel "free fall method" to determine the initial and final postures of the motion, which is key for successful predictions. On this basis, using the established simulation framework, the motion behaviour under nominal prosthetic stiffness was first predicted, and then the effects of changing prosthetic stiffness were studied. Finally, several scientific conclusions were drawn.
Limited ankle dorsiflexion is a common problem due to ankle injuries and surgical operations, greatly affects our mobility. This thesis is the first to complete the predictive study on the sit-to-stand movement behaviour for people with ankle dorsiflexion limitations. The prediction of sit-to-stand motion was formulated as an optimal control problem and solved by the direct collocation method. Subsequently, the effects of different levels of ankle dorsiflexion limitation were studied, revealing the corresponding biomechanical compensation strategies. The nature of no experimental data required as input makes our predictive study eliminate the lengthy tests in the laboratory or clinical settings, with great potential to benefit the development of new interventional treatment for people with ankle joint impairments.
The objective function is the important component of the optimal control problem. However, the weights of different terms in the objective function are often adjusted manually, imposing a substantial burden in terms of time and labour. To address this problem, this thesis integrated the inverse optimal control method with the muscle-driven musculoskeletal dynamics model to investigate muscle coordination during the transition from squatting to standing. A nested bilevel optimisation algorithm was proposed to solve the inverse optimal control problem. The proposed bilevel optimisation algorithm effectively addressed the complex muscle coordination problem during human motion, providing a new approach for studying muscle coordination patterns.
The new model and algorithm proposed in this thesis effectively fill various gaps in the existing human lower limb predictive simulation studies and have great potential in the customisation of rehabilitation programs, design of assistive devices and development of bionic robots. In addition, the achievements in this thesis have important reference significance for predicting movements in other scenarios.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):