Type two fuzzy logic control system for powered ankle-foot prosthesis: robust performance against terrain and speed variations

The development of powered prostheses' control systems is drifting away from discontinuous control systems based on the finite state machine (FSM), due to the issues of misclassification. Recent research focused on model-based control systems, most widely used is the hybrid zero dynamic (HZD). However, the HZD system depends on the model accuracy and number of feedback signals. Navigating different terrains is vital for independent mobility for people with lower-limb amputation. In this paper, we propose a control system based on Takagi-Sugeno-Kang (TSK) inference. The fuzzy system is based on Type 2 membership functions to accommodate the gait cycle uncertainties. To fulfil the design procedure, experiments were conducted to capture the ambulation data in different terrains. Twelve individuals participated in four experiments, a set of seven inertial measurement units (IMUs) were placed on the lower body. The data was used to assess the control system behaviour terrains and subjects' data. A model was built to represent the powered ankle-foot prosthesis, where the ground reaction force (GRF) and the target angular position were experimentally extracted and fed to the model. The control system was evaluated using three evaluation parameters (root mean square error (RMSE), mean absolute error (MAE) and normalized cross-correlation). The average RMSE is 3.22 ± 2.7 degree, and a high correlation of 96.73% can be observed. The performance matrix is uniform based on Kruskal-Wallis test (p = 0.9954), and effect size (Cramér's V = 0.02) indicated negligible practical significance of the change of speed and terrain on the control performance.