The robustness of a second order sliding mode approach to human gait simulation

Abstract

Nonlinear control techniques such as the sliding mode approach can provide a useful alternative to computationally intensive optimisation strategies in the simulation of human gait. Application of the sliding mode approach has the added advantage that it can provide estimates of internal parameters/signals and can be used to measure and/or monitor the deviation from normal gait in patients who are suffering from conditions such as osteoarthritis, for example. Effectively the principle of the equivalent injection that has been long used for fault detection and diagnostics in engineering machinery can be seen to provide a useful new dimension to gait analysis. Any such model, however, requires apriori estimates to be made of the physical dimensions and muscle characteristics of an individual. The fidelity of any resulting gait analysis will be dependent upon the degree to which the system is sensitive to the selection of such parameters. This paper investigates the effect of inducing errors in the mass, scale and proportions of the individual when simulating the gait cycle of ten normal subjects using a second order sliding mode approach. Seven experiments were carried out to examine the influence of errors up to 50% in the assumed parameters of a three-dimensional musculoskeletal lower body model. The second order sliding mode based simulated gait process was concluded to be sufficiently robust to body segment parameter variation that the use of scaled default parameters can be justified in the gait simulator.