A New Self-Adaptive Fuzzy PID Tracking Control for Autonomous Wheelchairs

Autonomous wheelchairs require robust control strategies to ensure accurate and safe navigation under system uncertainties and environmental disturbances. This paper presents a systematic design framework for trajectory tracking control based on a neuro-fuzzy identified model of an experimental wheelchair platform. Real input–output data were collected by exciting the system with a pseudo-random binary sequence (PRBS) signal, and the corresponding linear and angular velocities were recorded. A nonlinear model was then developed using an adaptive neuro-fuzzy inference system and validated against experimental measurements. Based on the identified model, a self-adaptive fuzzy PID (SAFPID) controller was designed to improve tracking performance and suppress vibration around the reference path. The controller employs an online adaptation mechanism to tune the membership function parameters, while the main control gains were optimized using the COVID-19 optimization algorithm through a multi-objective performance index. The proposed approach was implemented on an embedded platform consisting of an Arduino Mega 2560 and a Raspberry Pi 3. Experimental results demonstrate that the proposed controller significantly outperforms a conventional PID controller, reducing the root mean square tracking error from 1.258 m to 0.598 m and from 5.359 m to 1.365 m for two different trajectories.