Statistical evaluation of a fuzzy control system implemented on a low-cost microcontroller for wheelchair applications

The implementation of Fuzzy control systems on microcontrollers provides a low-cost, efficient, and innovative approach for assistive technologies, such as wheelchairs controlled by head and neck motion. However, their practical implementation is sensitive to several design parameters that are often overlooked. This study presents a statistical evaluation of a Fuzzy control system implemented on Arduino Mega and Due microcontrollers, focusing on the effects of the Iteration Step (IS) during defuzzification, microcontroller architecture, and overlap between membership functions, using the Center of Gravity (COG) method. System performance was evaluated using Root Mean Squared Error (RMSE) and processing time. The influence of an overlap between membership functions on the RMSE was also investigated. It was found that a smaller IS, such as 0.01, increased the precision of the system but led to long processing times (over 700 ms). Intermediate IS values (0.1 and 1) offered the most favourable trade-off between speed and accuracy, with RMSEs as low as 0.13% of the motor's total output and processing times under 260 ms. The Arduino Due significantly outperformed the Mega in processing time at low IS values, with little variation in accuracy. At higher ISs, the difference in processing time was negligible for practical use, making the Mega a viable cost-effective option. These findings provide empirical guidance for optimising Fuzzy control parameters in low-cost control systems, especially in applications where both real-time responsiveness and accuracy are critical.