Investigation of human gait control using simulation and sliding mode techniques
This thesis deals with the novel application of non-linear sliding-mode control techniques to the fields of gait analysis and locomotion control. The aim of the study was to create a platform for the development of sliding-mode controlled FES (functional electrical stimulation) systems for subjects suffering neural gait disorders. A model of the human locomotive system featuring 10 pin-jointed segments and 50 functioning muscles was created. A sliding-mode controller was applied to force the simulation to follow kinematic trajectories collected from subjects using gait analysis techniques and optimisation algorithms were developed to predict possible muscle activity patterns during the sampled gait sequences. With accurate parameters, motion tracking by the system took place with an error of less than 0.25 degrees and joint moments were generated within 1 standard deviation of the expected reference curves. Significant features of the measured EMG (electromyogram) readings typically matched similar features in the simulated muscle activations. Antagonism occurred in the simulated signals in the same periods as in the reference EMG readings, i.e. where the joint angles are most sensitive to moments acting on them. This is due to the switching nature of the sliding-mode controller. Sliding-mode techniques also provide insensitivity to model-plant mismatches reducing the need for accurate parameters, of which hundreds would be required and the majority of which would be difficult to obtain. As a forward-dynamics neuro-musculo-skeletal model with an integral sliding-mode controller and optimised muscle activity estimator, the model can function both as a powerful tool for gait analysis and non-invasive EMG estimation and as a platform for the development of FES controllers for subjects with neuro-muscular gait anomalies, thus fulfilling the stated aim of the study.