Development of a Smart Prosthetic Hand Using Fuzzy Logic for Muscle Signal Interpretation

Abstract

The thumb contributes roughly half of the total functional capacity of the human hand, since opposition and pinch are governed by its motion across the carpometacarpal (CMC) saddle joint. Restoring credible thumb motion to a powered prosthesis is therefore central to restoring credible hand function, and the difficulty lies in the controller rather than the mechanism. The inertial, damping and elastic parameters of the CMC joint are not known with certainty, and the controller driving it must remain accurate under that uncertainty. A fixed gain controller tuned to nominal values loses performance when the true parameters drift, and a conventional fuzzy controller with a static rule base cannot adapt to a plant it did not see during design. This thesis examines whether an adaptive fuzzy logic controller with a Lyapunov based online update law can track a commanded angular trajectory of the thumb CMC joint on a plant whose dynamics are not exactly known in advance. The study is carried out entirely in simulation in MATLAB R2025b and Simulink. The joint is modelled as a single degree of freedom in flexion extension, driven by a smooth trajectory generated from a sigmoid command through a critically damped reference model. An adaptive Takagi Sugeno Kang fuzzy controller with twelve Gaussian basis functions on the joint state produces an online estimate of the plant nonlinearity. That estimate is combined with sliding mode feedback in a certainty equivalent control law and updated by a Lyapunov derived adaptation law with bounded parameters, with convergence of the tracking error following from Barbalat’s lemma. The controller is evaluated across twelve simulation cases spanning six reference amplitudes from 0.20 to 1.05 radians, four levels of plant parameter uncertainty, four load disturbance magnitudes and four reference trajectory types. Each case is run against a non adaptive equivalent differing only in the adaptation gain, isolating what adaptation contributes. Without adaptation the joint settles 8.57 per cent short of its commanded angle at every setpoint and never enters the two per cent settling band. With adaptation the steady state error falls to 0.03 per cent, tracking error is reduced by 41.9 per cent on the nominal plant, and the advantage widens to 54.8 per cent when the plant parameters are doubled. On a continuously periodic reference, where no steady state offset exists for the adaptation to remove, the benefit falls to 2.2 percent. The contribution of the work is the design, stability analysis and simulation based validation of an adaptive TSK fuzzy controller for the thumb CMC joint, together with quantitative evidence that Lyapunov based online adaptation eliminates the steady state error that fixed gain sliding mode feedback cannot remove. The results establish the feasibility of the proposed approach for position control of an uncertain prosthetic joint. Keywords: Adaptive fuzzy control, Takagi Sugeno Kang (TSK) fuzzy system, Lyapunov based parameter adaptation, thumb carpometacarpal (CMC) joint, prosthetic hand, sliding surface, MATLAB simulation, uncertain nonlinear dynamics.