Autonomous Vehicle Intersection Path Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately predicting vehicle trajectories through intersections due to inaccuracies in mapped lane segments, especially at high speeds where actual paths diverge from mapped representations.
Innovation Solution
A system that identifies intersections, objects, and outlets, determines constant curvature paths, generates reference paths, and adjusts driving operations based on predicted trajectories, using scoring functions and metadata application to ensure safe navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mapped lane segments are used for trajectory prediction, then the system has a simple and reliable baseline approach, but the prediction accuracy deteriorates at intersections especially at high speeds
Solution Approach 1:
The system segments the path generation process into distinct components: identifying outlets from the intersection, generating constant curvature paths to each outlet, scoring paths based on multiple criteria (heading discrepancy, curvature, lateral acceleration), and pruning to select feasible paths. This segmentation allows complex intersection navigation to be broken down into manageable, independent steps that can be processed systematically.
Solution Approach 2:
The system dynamically adapts path generation based on current vehicle state (speed, heading) and intersection geometry. The constant curvature paths are generated on-demand rather than pre-defined, and the scoring function dynamically weights different path characteristics based on current conditions. This dynamic approach allows the system to handle diverse driving scenarios without requiring exhaustive pre-programming.
2Measurement precision
If constant curvature paths are generated for multiple outlets, then trajectory prediction accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The system generates constant curvature paths to all identified outlets (excessive action) but then applies a scoring and pruning mechanism to eliminate infeasible paths. This approach ensures that no potentially valid path is missed during generation, while the subsequent pruning step removes paths that fail to meet feasibility criteria (excessive computational effort on invalid paths). The balance between completeness and efficiency is achieved through this two-stage process.
Solution Approach 2:
The scoring function provides feedback on path feasibility by evaluating multiple criteria including heading discrepancy at the target point, curvature constraints, and lateral acceleration limits. Paths that score below thresholds are pruned from consideration. This feedback mechanism guides the selection of feasible paths without requiring exhaustive evaluation of all possible trajectories, reducing computational burden while maintaining accuracy.
3Measurement precision
If reference paths are generated that differ from mapped lane segments, then accuracy at high speeds improves, but deviation from standard mapped data increases system complexity
Solution Approach 1:
The system changes key parameters of the path representation by using constant curvature paths with dynamically calculated target points rather than relying on pre-defined mapped lane segments. The target point is positioned at a specific distance from the outlet along the polyline, and the constant curvature constraint ensures physically realistic paths that account for vehicle dynamics. This parameter change enables accurate representation of high-speed turning behavior that mapped segments cannot capture.
Data Source
AI summary
An autonomous vehicle identifies an intersection, identifies an object in proximity to the intersection, identifies a plurality of outlets of the intersection, and, for each outlet, identifies a polyline associated with the outlet, identifies a target point along the polyline, and determines a constant curvature path from the object to the target point. The system determines a score associated with each outlet based at least in part on the constant curvature path of the outlet, generates a pruned set of outlets that includes one or more of the outlets from the plurality of outlets based on its score, and for each outlet in the pruned set, generates a reference path from the object to the target point of the outlet.