Autonomous Vehicle Trajectory Biasing in Wide Lanes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle trajectory prediction systems do not account for lane biasing in wide lanes, where vehicles may adjust their position due to future actions, leading to uncertainty in path planning, especially in scenarios like turns or lane splits.
Innovation Solution
Implement a method to identify a lane biasing interval based on lane width, calculate lateral offsets for each index point of a reference path within this interval, and adjust the vehicle's trajectory using these offsets to ensure safe and optimized lane travel.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vehicle trajectory prediction systems use standard lane-based path planning, then path planning is simple and straightforward, but accuracy deteriorates in wide lanes where vehicles bias to certain sides due to future actions
Solution Approach 1:
The patent segments the lane into multiple sub-lanes or regions based on future actions and vehicle biasing patterns. Instead of treating the entire lane as a single path, the system divides it into segments with different bias characteristics, allowing more accurate trajectory prediction in each segment while maintaining manageable complexity through localized processing.
Solution Approach 2:
The system performs preliminary analysis of future actions (turns, lane splits, merges) to identify biasing intervals before trajectory prediction. By pre-identifying where vehicles tend to bias to certain sides based on upcoming road geometry and traffic patterns, the system prepares adjusted reference paths in advance, improving prediction accuracy without adding real-time computational complexity.
2Productivity
If vehicles travel adjacent to each other in wide lanes creating pseudo two-lane roads, then lane utilization is optimized, but trajectory prediction reliability deteriorates due to lack of knowledge about vehicle biasing behavior
Solution Approach 1:
The system incorporates feedback about observed vehicle biasing patterns in wide lanes into trajectory prediction. By monitoring how vehicles actually position themselves in pseudo two-lane scenarios and feeding this information back into the prediction model, the system learns and adapts to biasing behaviors, improving reliability while maintaining efficient lane utilization.
Solution Approach 2:
The patent dynamically adjusts trajectory parameters based on lane width and detected biasing patterns. In wide lanes where vehicles create pseudo-lanes, the system modifies lateral position parameters and biasing factors to reflect actual vehicle behavior rather than assuming centered lane travel, thereby improving prediction reliability without reducing lane utilization efficiency.
3Measurement precision
If the system adjusts trajectory to account for lane biasing in wide lanes, then path planning accuracy is improved, but computational complexity increases
Solution Approach 1:
The system applies lane biasing adjustments locally only in wide lanes where future actions indicate biasing behavior, rather than uniformly across all lanes. By identifying specific intervals and regions where biasing occurs and applying adjustments only there, the system improves path planning accuracy in critical areas while minimizing additional computational complexity in standard lane scenarios.
Data Source
AI summary
Disclosed herein are systems, methods, and computer program products for biasing a trajectory of a vehicle. The methods comprise: identifying a lane biasing interval of a lane in which the vehicle is to travel based on a lane width; obtaining a first reference path of travel for the vehicle that passes through the lane biasing interval; generating a second reference path of travel for the vehicle using a plurality of lateral offsets and the first reference path of travel for the vehicle; and adjusting the trajectory of the vehicle using the second reference path of travel. The lateral offsets are based on a lateral offset from each index point of the first reference path that resides in the lane biasing interval.