Dear: vehicle mobility prediction using diffusion-expanded attention network based on IoV trajectory data

In recent years, the rapid development of vehicle networking technology makes the acquisition of Internet of vehicles (IoV) trajectory data easier and more extensive. Mobility prediction based on these IoV trajectory data has become one of the research hotspots in the field of transportation. However, due to the randomness and nonlinear characteristics of IoV trajectory data, vehicle mobility prediction is a challenging task. In addition to traditional statistical methods, neural networks are promising methods for predicting vehicle mobility. Therefore, we design a deep neural network model and introduce the attention mechanism to achieve accurate prediction of vehicle mobility by learning and modeling the timing characteristics of IoV trajectory data. Diffusion-Expanded Attention Recurrent Neural Network (DEAR), a position prediction model for preference perception, is proposed. The introduction of diffusion model can better consider the dynamics of vehicle movement, adapt to the ever-changing traffic environment, and improve the accuracy of trajectory prediction. The experimental results show that our method can accurately predict the moving behavior of vehicles such as position, speed, and acceleration, which provides important support for real-time traffic flow optimization and driving safety, and has a wide application prospect.