Since the driving behavior of traditional human-driven vehicles(HVs) is affected by the uncertainty of the driver's psychological and physiological activities, which will cause frequent acceleration and deceleration of vehicles, it is difficult to ensure that the subsequent connected and automated vehicles(CAVs) can quickly track this behavior under mixed traffic conditions. In response to this issue, this paper first proposes a combined neural network for predicting the driving behavior of traditional HVs in advance. On this basis, considering the communication time delay and vehicle kinematic characteristics, a cooperative control strategy among vehicles within a mixed vehicle group based on transportation cyber physical systems(T-CPS) is designed to quickly track on the traditional HV behavior. Meanwhile, the string stability between the CAVs in the mixed vehicle group is analyzed. Finally, under mixed traffic conditions, a mixed vehicle group consisting of 1 traditional HV, 1 leader CAV, and 4 following CAVs is set up. The next generation simulation(NGSIM) vehicle trajectory data is used to select high-quality traditional HV states. The effectiveness and feasibility of the proposed cooperative control strategy are verified through simulation experiments. The simulation results show that the proposed cooperative control strategy can ensure that all the CAVs can quickly track the traditional HVs behavior, which provides certain theoretical guidance and reference for solving the new problems brought by the new mixed traffic.