Research on vehicle trajectory fusion prediction based on physical model and driving intention recognition

At present, accurately predicting the long-term trajectory of traffic vehicles for autonomous vehicles remains a challenging task. Dynamic scenarios often necessitate frequent replanning, which can waste computing resources and increase the risk of traffic accidents. To address this issue, this paper proposes a vehicle trajectory fusion prediction method based on a physical model and driving intention recognition. Firstly, trajectory prediction is based on the Constant Turn Rate and Acceleration (CTRA) model, which is combined with the vehicle's motion state to obtain Trajectory1. Next, a Hidden Markov Model (HMM) is employed to identify driving intentions. Building upon this, a Gaussian Mixture Model (GMM) is used to perform probability density statistical analysis on driving data, yielding feature parameters Dx and Dy. These parameters are then combined with a Quintic polynomial to predict the trajectory, resulting in Trajectory2. Finally, Trajectory1 and Trajectory2 are fused to obtain the ultimate predicted trajectory, referred to as Trajectory3. To validate the effectiveness of the trajectory prediction method proposed in this paper, the algorithm is tested in both left lane change (LCL) and right lane change (LCR) scenarios. The test results demonstrate that the root mean square error (RMSE), mean absolute error (MAE), and maximum absolute error (MXAE) for Trajectory3, generated using the fusion algorithm, are significantly smaller than those for Trajectory1 and Trajectory2. This indicates the efficacy of the proposed model, which contributes to making high-quality decisions and plans for autonomous vehicles, reducing the probability of traffic accidents, and enhancing public confidence in autonomous vehicle technology.