Control of Induction Motor Drive Using Fuzzy Sarsa(λ)-Learning and Meta-Heuristic Algorithm for Electric Vehicle Application

Abstract Electric vehicles demand drive systems that combine fast dynamic response with high efficiency and reduced torque ripple. Direct torque control (DTC) of induction motors is widely used in this context, but conventional schemes suffer from variable switching frequency, flux distortion, and poor performance at low speeds. To address these limitations, this paper introduces a learning-based strategy that integrates a fuzzy inference system with SARSA(λ) reinforcement learning. A genetic algorithm is employed offline to tune the membership functions, providing a reliable initialization for the fuzzy system. During operation, the SARSA(λ) agent adapts the control policy in real time by balancing exploration and exploitation through eligibility traces. The proposed controller is evaluated in MATLAB/Simulink under multiple step changes in speed and load conditions and is benchmarked against classical DTC and fuzzy logic–based DTC. Simulation results show that the method achieves faster transient response, lower overshoot, and a significant reduction in torque ripple compared with the reference schemes. These findings suggest that reinforcement learning combined with fuzzy optimization can improve the adaptability and robustness of induction motor drives for electric vehicle applications.