This paper presents mathematical methods to develop a high-efficiency and real-time driving energy management for a front-and-rear-motor-drive electric vehicle (FRMDEV), which is equipped with an induction motor (IM) and a permanent magnet synchronous motor (PMSM). First of all, in order to develop motor-loss models for energy optimization, database of with three factors, which are speed, torque and temperature, was created to characterize motor operation based on HALTON sequence method. The response surface model of motor loss, as the function of the motor-operation database, was developed with the use of Gauss radial basis function (RBF). The accuracy of the motor-loss model was verified according to statistical analysis. Then, in order to create a two-factor energy management strategy, the modification models of the torque required by driver (Td) and the torque distribution coefficient (β) were constructed based on the state of charge (SOC) of battery and the motor temperature, respectively. According to the motor-loss models, the fitness function for optimization was designed, where the influence of the non-work on system consumption was analyzed and calculated. The optimal β was confirmed with the use of the off-line particle swarm optimization (PSO). Moreover, to achieve both high accuracy and real-time performance under random vehicle operation, the predictive model of the optimal β was developed based on the hybrid RBF. The modeling and predictive accuracies of the predictive model were analyzed and verified. Finally, a hardware-in-loop (HIL) test platform was developed and the predictive model was tested. Test results show that, the developed predictive model of β based on hybrid RBF can achieve both real-time and economic performances, which is applicable to engineering application. More importantly, in comparison with the original torque distribution based on rule algorithm, the torque distribution based on hybrid RBF is able to reduce driving energy consumption by 9.51% under urban cycle.