For the deficiency of current electric vehicle compound regenerative braking, a new design of electromagnetic-mechanical coupled regenerative braking system was proposed (referred to as EMCB), and the coupled mechanism and dynamic model were analyzed. For the deficiency of conventional ABS discrete on-off control, a continuous-state control of anti-lock braking control strategy was proposed based on EMCB system and fuzzy-adaptive sliding mode control for electric vehicle. With Matlab/Simulink the dynamic model of EMCB and controller model were established, and the vehicle model was built by CarSim. With the co-simulation platform of Matlab/Simulink and CarSim, the braking energy recovery, braking stability and braking efficiency were studied and simulated in case of straight braking condition on step adhesive coefficient road and cornering braking condition on low adhesive coefficient road. The results showed that the ABS control strategy had good responsiveness, robustness and slip control performance. Fuzzy-adaptive variable structure control was insensitive to the road conditions and system model and control variable was flat and smooth, it reduced the demand of executive system. The control strategy ensured braking stability and braking efficiency at the limit of adhesion and large angle steering, which can maintain high recovery efficiency and increase the driving range of electric vehicles effectively.