The invention discloses a decision-oriented intelligent automobile trajectory prediction method and prediction model. The method comprises the following two core contents: trajectory prediction based on a time sequence and trajectory prediction based on conditions. A prediction model serving a multi-cycle trajectory prediction task is constructed by introducing a time sequence, and time sequence consistency in each prediction cycle is realized, including information multiplexing of a historical cycle and prediction mode continuation of a future cycle; and scene consistency of trajectory prediction is realized by introducing a condition prediction model, namely scene-level multi-modal trajectory prediction is realized. In order to solve the problems that an existing mainstream trajectory prediction model constructs a prediction algorithm into an independent task which is single and closed and cannot be unified with a downstream decision planning task, a trajectory prediction model introducing a time sequence and condition prediction is constructed; uniformity of a trajectory prediction task, a downstream decision planning task and other tasks in an algorithm process is realized.
本发明公开了一种面向决策的智能汽车轨迹预测方法及预测模型,包括以下两个核心内容:基于时间序列的轨迹预测与基于条件的轨迹预测。通过引入时间序列构建服务于多周期轨迹预测任务的预测模型,实现各预测周期内的时序一致性,包括历史周期的信息复用与未来周期的预测模态延续;通过引入条件预测模型实现轨迹预测的场景一致性,即实现场景级别多模态轨迹预测。本发明针对现有主流轨迹预测模型将预测算法构建为独立任务,存在单一、封闭、不能与下游决策规划任务统一的问题,构建了一个引入时间序列与条件预测的轨迹预测模型,实现了轨迹预测任务与下游决策规划等任务在算法流程上的统一性。
Decision-oriented intelligent automobile track prediction method and prediction model
一种面向决策的智能汽车轨迹预测方法及预测模型
2024-08-16
Patent
Electronic Resource
Chinese
Intelligent automobile driving track planning method based on driving intention prediction model
European Patent Office | 2024
|Prediction method and device for driving track of intelligent automobile
European Patent Office | 2024
|Intelligent networked automobile real-time track prediction method based on improved neural network
European Patent Office | 2023
|