The vehicle driveline suffers low frequency torsional vibration due to the abrupt change of input torque and torque fluctuation under variable frequency. This problem can be solved by model based control, so building a control oriented driveline model is extremely important. In this paper, an on-line recursive identification method is proposed for control oriented model and validated based on an electric car. First of all, the control oriented driveline model is simplified into a six-parameter model with double inertia. Secondly, based on stability analysis, motor torque and motor speed are chosen as input signal for on-line model identification. A recursive identification algorithm is designed and implemented based on Simulink. Meanwhile a detail model of the vehicle which considering driveline parameter variation is built based on ADAMS. Thirdly, on-line identification is conducted by using co-simulation of ADAMS and Simulink. Compared with off-line identification model, the online identification model can reflect dynamic stiffness which will be changing under different excitation frequency and variable vehicle parameters including tire damping and driveshaft damping. Finally, the validation of on-line identification model is conducted under tip-in condition. Results show that outputs of on-line identification model is consistent with the outputs of vehicle model in ADAMS. So, using online identification model, more accurate control will be achieved.