一种基于历史时序特征的分车型跟驰行为决策方法
By using a vehicle-type-based car-following behavior decision-making method based on historical time-series features and employing the SE-TCL neural network to predict the acceleration and speed of following vehicles, the shortcomings of existing models in simulating driving behavior of different vehicle combinations are addressed, thereby improving the intelligence level of autonomous vehicles.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HEFEI UNIV OF TECH
- Filing Date
- 2022-07-08
- Publication Date
- 2026-07-17
AI Technical Summary
Existing car-following models struggle to accurately simulate driving behavior across different vehicle combinations, especially in mixed traffic flows. They fail to adopt appropriate car-following behaviors based on the type of vehicle ahead, resulting in insufficient intelligence levels in autonomous vehicles.
A vehicle-type following behavior decision-making method based on historical time-series features is adopted. By acquiring vehicle trajectory data, smoothing it using the symmetric exponential moving average method, and constructing an SE-TCL neural network model, a time attention module, a time-series convolution module, and a time-series learning module are combined to predict the acceleration and speed of the following vehicle.
It improves the accuracy and precision of predicting following behavior for different vehicle combinations, and can select appropriate driving behaviors based on the types of vehicles in front and behind, thus enhancing the intelligence level of autonomous vehicles.
Smart Images
Figure CN115186583B_ABST