基于时空联合分层先验知识的智驾系统数据增强方法
By generating synthetic perception data and predicting vehicle behavior, and utilizing spatiotemporal joint hierarchical prior knowledge for data augmentation of intelligent driving systems, the problems of insufficient data diversity and accuracy in traditional methods are solved, thereby improving the performance and generalization ability of the model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH OF CHINA
- Filing Date
- 2026-06-22
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional data augmentation methods have failed to fully utilize spatiotemporal joint hierarchical prior knowledge in the field of intelligent driving, resulting in insufficient data diversity and accuracy, which affects the performance and generalization ability of intelligent driving systems.
A data augmentation method based on spatiotemporal joint hierarchical prior knowledge is adopted. Synthetic perception data is generated by integrating spatiotemporal attention mechanism under the constraints of physical simulation results and spatial invariance. Combined with kinematic and dynamic models, vehicle behavior is predicted, and control data is generated according to vehicle control strategy to achieve end-to-end driving model training.
It improves the quality of training data and model performance of intelligent driving systems, and enhances decision-making accuracy and generalization ability in complex driving scenarios.
Smart Images
Figure CN122414271A_ABST