Autonomous Vehicle Reference Line Stitching via Parallel Thread Precalculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving systems face inefficiencies in generating smooth reference lines for motion planning, as existing optimization methods are time-consuming and can occupy a significant portion of the planning cycle, limiting the length of the reference line that can be calculated within the available time.
Innovation Solution
The system utilizes multiple processing threads to precalculate and smoothly extend the reference line ahead of the vehicle, allowing for parallel processing while the vehicle is moving, thereby optimizing the reference line generation without delaying the planning cycle.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If smooth optimization is performed on the reference line to make it smoother and easier for the ADV to follow, then the reference line quality is improved, but the processing time increases significantly (0.1 second for 100m reference line)
Solution Approach 1:
The system pre-calculates and smooths the reference line in advance before the planning cycle begins. By performing the computationally intensive smooth optimization beforehand, the system prepares the reference line so that it is ready for use during the actual planning cycle, eliminating the time bottleneck during critical real-time operation.
Solution Approach 2:
The reference line generation process is divided into separate stages: initial reference line creation and subsequent smooth optimization. This segmentation allows the smoothing operation to be performed independently in advance, separating the time-consuming optimization from the time-critical planning cycle.
2Measurement precision
If the reference line length is increased to improve navigation accuracy and anticipate turns, then the planning quality is improved, but the optimization time increases proportionally
Solution Approach 1:
The system performs smooth optimization on extended reference line segments in advance, before they are needed for planning. This allows longer reference lines to be processed offline, maintaining high navigation accuracy while preventing the optimization time from impacting real-time planning performance.
Data Source
AI summary
In one embodiment, when an ADV is driving on a trajectory generated based on a reference line, a separate processing thread is executed to precalculate a new reference line as a future reference line for a future planning cycle in parallel. The future reference line is being created while the ADV is moving along a trajectory generated based on the original reference line and before reaching a location corresponding to the starting point of the future reference line. The future reference line is overlapped with an end section of the original reference line, such that the future reference line can be connected to the end section of the original reference line. The future reference line serves an extension of the original reference line before the ADV reaches the end section of the original reference line.