Dual Trajectory Planning for Low-Latency Autonomous Vehicle Response
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle trajectory planning techniques suffer from unacceptably long latency due to layered architectures, leading to delays in responding to sudden environmental changes, which can result in unsafe trajectories.
Innovation Solution
Implementing a dual trajectory planning system with a nominal planner and a fast-path planner that bypasses optimization layers, utilizing separate or shared compute resources to generate and compare trajectories based on up-to-date perception data, ensuring rapid response to changing conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a layered trajectory planning architecture is used to generate optimized trajectories, then trajectory optimization quality is improved, but processing time increases to 100 milliseconds or more per layer
Solution Approach 1:
The trajectory planning system is divided into two separate segments: a nominal trajectory planner that performs full optimization and a fast-path trajectory planner that performs rapid trajectory generation. This segmentation allows the system to maintain optimization quality when needed while achieving fast response times for urgent situations, resolving the contradiction between optimization quality and processing speed.
Solution Approach 2:
The fast-path trajectory planner performs a simplified, partial optimization process that generates trajectories quickly without full optimization. This partial action approach sacrifices some optimization depth in exchange for dramatically reduced processing time, while the nominal planner performs complete optimization when time permits.
2Manufacturing precision
If conventional layered trajectory planning is used, then comprehensive trajectory optimization is achieved, but response latency to sudden environmental changes becomes unacceptably long
Solution Approach 1:
The fast-path trajectory planner is pre-configured with simplified planning logic and pre-computed parameters that enable immediate trajectory generation when environmental changes are detected. This preliminary preparation allows the system to respond instantly to sudden changes without undergoing the full optimization process, thereby improving response speed while maintaining the capability for comprehensive optimization through the nominal planner.
3Measurement precision
If multiple processing layers are implemented for trajectory planning, then trajectory selection accuracy is improved, but total processing latency increases approaching half a second
Solution Approach 1:
The fast-path trajectory planner creates a simplified copy of the trajectory planning function that replicates essential planning capabilities without the full multi-layer processing overhead. This copy operates in parallel and can quickly generate candidate trajectories, reducing total processing latency while the nominal planner maintains comprehensive optimization and selection accuracy.
Data Source
AI summary
A method for selecting a trajectory for a vehicle includes receiving first perception data from one or more sensors positioned on the vehicle at a first time; generating at least one nominal trajectory based on the first perception data; receiving second perception data from the one or more sensors at a second time after the first time; generating at least one fast-path trajectory based on the second perception data; selecting a trajectory from the at least one nominal trajectory and the at least one fast-path trajectory.