Hub direct-drive electric vehicles are driving with hub motors, which cause negative vibration directly on wheels and suspensions. To better analyze the problems of dynamic deterioration resulted by coupling excitations from random road and unbalanced magnetic pull (UMP), a cooperative hub direct drive-air suspension (HDD-AS) system is developed, and evaluation indices based on vertical and longitudinal dynamic coupling mechanism are proposed. The vertical and longitudinal characteristics in the time domain under different damping coefficients and motor speeds are demonstrated, and comparison between traditional electric wheel and suspension-integrating models and HDD-AS system model is conducted. Then, the responses of random road excitation and UMP are analyzed, and explanations of vertical and longitudinal characteristics in the frequency domain under different damping coefficients are given. These results provide important insights into the influence of coupling excitations and damping coefficient on the HDD-AS system model. Finally, the effectiveness of the proposed model is verified by the experiment on the test bench.

For the hub motor electric vehicle (HM-EV), the drive motor is directly integrated with the wheel. The unbalanced magnetic pull (UMP) of hub motor would be generated by magnet gap deformation under road surface roughness excitation. The longitudinal and vertical dynamic performances of the HM-EV system are therefore deteriorated. Firstly, to analyze and optimize the longitudinal and vertical dynamic performance of the HM-EV system, a new ten-degree-of-freedom mathematical quarter HM-EV system model equipped with air suspension model, permanent magnet brushless direct current (PM BLDC) hub motor model and rigid ring tire model is proposed. The UMP of PM BLDC hub motor is taken into consideration in this model. A HM-EV system model validation test bench is constructed. The accuracy of the model is verified by experiment. Secondly, based on quarter HM-EV system model, the BP neural network is adopted to calculate the longitudinal and vertical UMP. The relative error between results calculated by BP neural networks and electromagnetic formula is less than 5% and root-mean-square error (RMSE) is less than 2. With proposed BP neural networks calculation method, UMP calculation time is shortened by 70.3%. Finally, the adjustable force is introduced and model predictive control (MPC) method is used to suppress the longitudinal and vertical vibration of HMEV system. Two control methods, namely model predictive control (MPC) and constrained optimal control (COC) are proposed. The simulation results show that by applying MPC, the RMS value of evaluation indexes are decreased by 17.21%–44.10% respectively, which is better than COC (−14.42%–17.21%). With MPC, longitudinal and vertical vibration are suppressed. Comparison of two UMP calculation methods with MPC controller is conducted. The relative errors of evaluation indexes are within 3.85%. Therefore, the driving safety and riding comfort of the HM-EV are improved compared to the passive suspension and COC active suspension.