Vehicle Trace Alignment Using Road Feature Context
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle location tracking systems suffer from inaccuracies due to errors in GNSS measurements, odometry drift, and sensor errors, leading to reduced accuracy in determining vehicle positions and relative locations, which affects the precision of autonomous driving systems.
Innovation Solution
A method for aligning vehicle traces by obtaining an alignment context with observed feature positions and modifying vehicle trace data based on correspondence and error margins, using sensor fusion and optimization techniques to improve spatial accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GNSS measurements are used for vehicle location tracking, then vehicle positions can be determined globally, but measurement accuracy deteriorates due to GNSS errors
Solution Approach 1:
The patent combines multiple independent measurement systems (GNSS, odometry, visual odometry, map matching) into a unified location tracking system. Each system has different error characteristics, and their combination through sensor fusion algorithms produces a more reliable and accurate vehicle position estimate than any single system alone, resolving the contradiction between global coverage and measurement reliability.
Solution Approach 2:
The system implements feedback mechanisms where vehicle position estimates are continuously refined by comparing predicted positions (from odometry) with actual positions (from GNSS and map matching). Error margins are dynamically adjusted based on the consistency between different measurement sources, allowing the system to compensate for GNSS errors and maintain reliable location tracking.
2Measurement precision
If odometry sensors are used to measure local motion, then relative movement accuracy is improved, but measurement reliability deteriorates due to accumulated error and drift over time
Solution Approach 1:
The system periodically resets or recalibrates the odometry measurements by aligning with GNSS positions and map features at regular intervals or when confidence thresholds are met. This periodic correction prevents error accumulation and maintains long-term reliability while preserving the high short-term accuracy of odometry for smooth motion tracking.
Solution Approach 2:
Odometry measurements are continuously monitored and corrected based on feedback from GNSS positions and map matching results. When drift exceeds acceptable thresholds, the system adjusts the odometry trajectory to align with reliable reference points, maintaining both accuracy and reliability over extended periods.
3Measurement precision
If multiple vehicle traces are processed simultaneously, then spatial accuracy of individual traces is improved through alignment, but system complexity increases
Solution Approach 1:
The alignment process is segmented into hierarchical levels: first aligning individual traces to a reference coordinate system using map features, then aligning multiple traces to each other through mutual optimization. This segmentation allows the complex multi-trace alignment problem to be broken down into manageable sub-problems, reducing computational complexity while maintaining accuracy.
Solution Approach 2:
Before performing full multi-trace alignment, the system performs preliminary alignment of each trace to the map coordinate system using prominent features and error margins. This preliminary positioning provides a good initial state for the subsequent mutual alignment process, significantly reducing the computational search space and complexity of the optimization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of aligning one or more vehicle traces for one or more vehicles traversing a geographical area, the method comprising: obtaining an initial alignment context for the geographical area, wherein the initial alignment context comprises a plurality of observed positions of road features; obtaining one or more vehicle traces for the one or more vehicles, each vehicle trace defining a path travelled by the respective vehicle across the geographical area, and comprising one or more observations of respective road features observed relative to the respective path; selecting from the alignment context, for each of the vehicle traces, a respective set of observed positions corresponding to observations in said vehicle trace; and adjusting each vehicle trace based on discrepancies between the observations of the trace and the corresponding observed positions in the selected set of observed positions, to align each vehicle trace to the alignment context.