Accurate tracking of vehicle steering trajectories is crucial to the safety of traveling in tunnels. In this paper, a multi-vehicle steering trajectory tracking method based on improved residual network is proposed for tunnel scenarios, which combines the attention mechanism and model predictive control technology to realize accurate tracking. Aiming at the problem that the traditional twin network tracking algorithm is not satisfactory enough in tunnel scenarios, the ECA channel attention mechanism is introduced to improve the structure of the residual network and enhance the feature extraction capability; the feature fusion module is designed to effectively integrate different levels of feature information; and the model predictive controller based on the spatial deviation model is constructed to realize accurate tracking. The experimental results show that in the simple occlusion scenario, the algorithm in this paper improves the tracking accuracy MOTA by 3.6% to 83.42% compared with SiamCAR algorithm, and the tracking precision MOTP improves by 3% to 88.19%, and the number of identity switching is reduced to 5 times; in the complex traffic scenario, the tracking accuracy improves by 2.4% to 78.77%, and the tracking precision improves by 4.2% to 85.69%. The active steering experiment based on data recharge verifies the effectiveness of the control method, and the system is able to adjust the trajectory deviation to ensure the smooth driving of the vehicle. The method can realize accurate tracking of multi-vehicle steering trajectories in tunnel scenarios and improve driving safety.
