As the market for motor steering components and parts becomes increasingly mature, the requirement for vehicle steering stability becomes strict. Steering wheel torque may fluctuate when the vehicle is running on an uneven road and under steering reversal condition. A change in the return torque of the steering wheel may result in insufficient return or return overshoot, which can affect the vehicle's steering stability. To overcome the adverse effect of random pavement excitation on vehicle return-ability, an electric power steering (EPS) system return control algorithm that integrates disturbance observer and back-stepping sliding-mode algorithm was proposed in this study. First, a return control strategy based on the return torque needed by the vehicle was established. Under the return condition, the observed value of disturbance observer was considered the theoretical basis. Then, the power-assisting motor return current of the EPS system was compensated, and the target current value of return control was determined. A full-vehicle model was established in Adams dynamic simulation software, and a controller model of the EPS system was built in MATLAB software. A random pavement excitation signal was then added in the Adams full-vehicle model, and return-ability was evaluated using vehicle yaw velocity and lateral acceleration. Finally, the stability of the designed system was verified on the test bed. Results show that the return torque provided by the power-assisting motor is verified through a comprehensive performance bench of 2 N.m, and the maximum return residual angle of the steering wheel is kept within 5°,thereby improving the vehicle steering stability. This study provides a theoretical reference for the return response of EPS systems under external excitations of different frequencies.