一种面向控制可执行性的自动驾驶轨迹预测控制方法
By generating multimodal predicted trajectories and parameterized continuous reference trajectories, and combining model predictive control and residual control correction, the problems of trajectory continuity and control executability in autonomous vehicles are solved, and the continuity of trajectory prediction and control stability are improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHANGCHUN UNIV OF TECH
- Filing Date
- 2026-06-17
- Publication Date
- 2026-07-17
AI Technical Summary
Existing trajectory prediction methods in autonomous vehicles suffer from insufficient trajectory continuity, physical feasibility, and control executability, leading to frequent adjustments of control inputs by the controller and reduced closed-loop tracking stability.
By constructing a vectorized scene representation, multimodal predicted trajectories and parameterized continuous reference trajectories are generated. Combined with model predictive control and residual control correction, a unified loss function is used for modular end-to-end joint optimization to generate continuous, smooth, and executable control commands.
It improves the continuity of trajectory prediction and the executability of control, reduces the rate of control change, and enhances the closed-loop tracking stability of autonomous vehicles.
Smart Images
Figure CN122402584A_ABST