一种机器人轨迹规划方法及系统
By generating an initial trajectory through multimodal sensor data fusion and an improved APF-RRT algorithm, and pre-verifying it using a digital twin platform, the problems of low multimodal perception fusion and insufficient virtual-real interaction in the end-effector trajectory planning of wheeled humanoid robots are solved, achieving efficient and safe trajectory planning and execution.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTH CHINA UNIV OF TECH
- Filing Date
- 2026-05-25
- Publication Date
- 2026-07-17
AI Technical Summary
Existing trajectory planning technologies for wheeled humanoid robots suffer from low multimodal perception fusion and insufficient pose estimation accuracy. Traditional trajectory planning methods have significant limitations, lack effective pre-execution verification, and the simulation system lacks closed-loop control, making it impossible to achieve bidirectional virtual-real interaction. This results in low trajectory planning efficiency and insufficient safety.
Multimodal sensor data is denoised in the same layer and fused across layers to generate a unified task-level state vector. An improved APF-RRT fusion algorithm is used to generate an initial trajectory. A digital twin platform is used for collision detection and reachability pre-verification. Dynamic response control is combined to perform local minimum escape or global replanning.
It achieves high-precision trajectory planning, improves the efficiency and safety of trajectory planning, ensures that virtual optimization results can be directly converted into control commands for the actual production line, reduces physical debugging costs, and enhances the operational capabilities of flexible production lines.
Smart Images
Figure CN122401414A_ABST