Output-Feedback Adaptive Neural Control of a Compliant Differential SMA Actuator

This brief focuses on modeling and neural-network based control of a novel compliant differential shape memory alloy (SMA) actuator characterized by reduced total stiffness and increased compliance. A fourth-order strict-feedback nonlinear model with an internal dynamics is derived to fully describe the SMA actuator. Due to nonlinearity, parametric uncertainty, and state-measurement difficulty of the SMA actuator, an adaptive observer-based output-feedback adaptive neural control method is developed to rigorously guarantee closed-loop stability. An experimental device is constructed to test the performance of the SMA actuation control system, where load changes and control tasks with various frequencies are considered during experiments. Experimental results have demonstrated effectiveness and superiority of the proposed approach.