Second-order sliding-mode control as an alternative to static and dynamic optimisation in human gait simulation

Abstract

Non-linear control techniques can provide an alternative to computationally intensive optimisation strategies in the simulation of human locomotive gait. This paper demonstrates the application of a second-order sliding-mode controller to a three-dimensional neuro-musculo-skeletal lower limb model in order to motion track kinematic data from ten normal male subjects. The controller was able to maintain stable tracking over two complete gait cycles of every sample with mean errors almost exclusively within one-standard deviation of the measured trajectories. The insensitivity to plant/model mismatches and unmodelled dynamics of the sliding-mode technique permits stable control of the highly non-linear muscles in a non-uniform arrangement without the need for inverse-dynamics analysis of muscle behaviour or body dynamics. As a true forward-dynamics system a unique solution is produced, rendering optimisation redundant.