A QCGNN-Based Predictive Framework for a Common Sliding Mode Control for Enhanced Fault-Tolerant Performance of a Four-Wheel Independently Driven Electric Vehicle

Abstract

The increasing popularity of electric vehicles is inevitable due to their lower dependence on conventional fuel and reduced air pollution. Among various drivetrain architectures, four-wheel independently driven electric vehicles (4WID-EVs) have gained significant attention owing to their superior load-carrying and dynamic performance. However, the distributed four-motor architecture makes them vulnerable to unpredictable motor failures, necessitating an effective fault-tolerant control strategy. This work proposes a common sliding mode controller (CSMC)-integrated quantum complete graph neural network (QCGNN) for adaptive tuning under one-, two-, and three-motor failure conditions at reference speeds of 20 and 40 m/s, ensuring stable operation through continuous state feedback. Simulation results demonstrate fault recovery within 3 s, a rise time of 1.2–1.3 s, a settling time below 5.5 s, a peak overshoot below 8%, and a steady-state error below 0.2%. Compared with the QCGNN-Optimal LQR, the proposed QCGNN-CSMC reduces the Mean Absolute Error (MAE) from 34 to 18, Root Mean Square Error (RMSE) from 41 to 23, and the steady-state error from 1.07 to 0.56, while maintaining R2 values above 0.90. Rapid controller prototyping further validates the robustness, reliability, and real-time applicability of the proposed fault-tolerant control framework for 4WID-EVs.