Abstract In this paper, the predictive control strategy is employed to improve the current tracking performance of hub motor in 4WD electric vehicle due to its fast dynamic response. But the performance of the conventional predictive deadbeat current control suffers greatly from the parameter variations and other disturbances. Toward this, this paper presents a new predictive control strategy for hub motor; this control scheme combines an improved predictive control law with a state-observer to estimate the future motor currents and system disturbances based on a decoupled model. It provides a decoupled control of hub motor and offers stability against the variations in motor inductance and robustness against system uncertainties. The feasibility and validity of the proposed predictive current control strategy is verified through the simulation results. Introduction Electric vehicles have become a research focus in the automotive industry under pressure from energy shortage and environmental pollution. Among various architectures of electric vehicles, four in-wheel independent drive (4WD) electric vehicles are considered as promising vehicle architectures due to its zero emission, high energy efficiency and individual control of each wheel [ 1]. The hub motor serves as the actuator of the 4WD electric vehicle, its own characteristics and its control performance directly affect the driving stability and dynamics performance of the 4WD electric vehicle. Permanent Magnet Synchronous Motor (PMSM) are gradually being applied to the 4WD electric vehicle as its hub motors, due to its high energy density, high torque inertia ratio, low torque ripple and compact design, etc [2]. The PMSM hub motor drive system is a torque servo system, the basic requirements for its controller are the fast dynamic response during the transient state, lower current ripple in the steady-state and robustness against the plant uncertainties. The vector control based on rotor flux orientation is a widely used approach for high- performance control of a PMSM [3], which allows the electromagnetic torque and stator flux of a PMSM can be controlled by q- and d- axis currents separately. However, the cross-coupling terms in voltage equations result in an interaction effect in two axis currents control, especially under the high speed, the significant increase in interaction effect will degrade current control performance [ 4]. The feedforward decoupling and feedback decoupling can realize complete linearization decoupling, but their decoupling accuracy depend on precise parameters of motor. Various current control strategies for PMSM drive have been developed over the last few years, such as hysteresis control, ramp comparison control, synchronous frame proportionalintegral (PI) control and predictive controls [ 5,6], etc. Compared with other current control schemes, the predictive deadbeat current control (PCC) is currently attracting growing interest as a promising mean of improving the performance of PMSM drive system. The PCC is able to predict the future behaviors of regulated currents using a motor model, when combined the SVPWM technique, it provides the fastest transient response, more precise current control and the lowest current ripple [ 7]. But as a model-based control strategy, the conventional predictive current controls are fragile in presence of plant uncertainties and other unknown disturbances. They become unstable when the actual stator inductance is half of its programmed value, and presents a steady-state current error under the parameter variations and other disturbances. To overcome above defects of the conventional PCC, a robust predictive current control (RPCC) is proposed in [ 8,9], it employs a luenberger observer to predict the future current to compensate the control delay and reduce the effects of the inductance variation on the system stability, but it only improves the system stability,