Driving Trajectory Generation Using Lane Gates and Vehicle Traces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating driving trajectories for autonomous vehicles are time-intensive and unreliable due to manual curation based on assumptions of zero traffic, which do not account for real-world road conditions and traffic variations.
Innovation Solution
A trajectory generation system that utilizes vehicle traces from manually operated vehicles to create driving trajectories by clustering paths through lane gates, calculating lateral positions, and generating trajectories based on actual driving behaviors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual curation methods are used to generate driving trajectories, then the trajectories can be created with basic functionality, but the process becomes time-intensive and unreliable due to assumptions of zero traffic
Solution Approach 1:
The system copies actual driving trajectories from manually operated vehicles and uses them as templates for autonomous vehicle navigation. Instead of manually creating trajectories from scratch, the system collects and replays real-world driving data, significantly reducing generation time while improving reliability by basing trajectories on actual road conditions rather than zero-traffic assumptions
Solution Approach 2:
The system performs preliminary data collection and processing by gathering trajectory data from manually operated vehicles before autonomous operation. This pre-collected data is stored and processed into usable trajectory formats, so that when autonomous vehicles need navigation paths, they can quickly access pre-validated trajectories rather than generating them in real-time
2Adaptability or versatility
If manual curation based on zero traffic assumptions is used, then the trajectory generation process is simplified, but the trajectories do not account for real-world road conditions and traffic variations
Solution Approach 1:
The system uses data from manually operated vehicles to automatically generate trajectories for autonomous vehicles. The manually operated vehicles essentially service the data collection function, providing real-world trajectory data that automatically adapts to actual road conditions, traffic patterns, and environmental factors without requiring manual adjustment of each trajectory
Solution Approach 2:
The system changes the fundamental parameter basis for trajectory generation from theoretical zero-traffic conditions to actual observed traffic and road conditions. By collecting data from multiple vehicles under various real-world conditions and processing this data to extract representative trajectories, the system adapts trajectories to reflect actual operational parameters rather than idealized assumptions
Data Source
AI summary
Systems, methods, and other embodiments described herein relate to generating driving trajectories based on vehicle traces through lane gates. In one embodiment, a method includes generating a network of lane gates for a multi-lane road. The method also includes, for a target lane gate in the network, 1) identifying paths of connected lane gates passing through the target lane gate, 2) clustering vehicle traces passing through the target lane gate that have a same trace path, and 3) calculating a lateral position along the target lane gate for a cluster of vehicle traces. The method also includes generating a driving trajectory for the path based on lateral positions for the cluster at multiple lane gates along a path.