Georeferenced Trajectory Alignment for GPS-Degraded AV Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicle localization systems face challenges in generating accurate and coherent global trajectories due to unreliable GPS data, especially in urban environments, and the difficulty of aligning LIDAR data in dynamic settings, leading to inconsistent and fuzzy map creation.
Innovation Solution
A georeferenced trajectory estimation system that aligns sensor data from multiple vehicles into a common coordinate frame using LIDAR beam intensity values and pose graph optimization, enabling the creation of accurate geometric models and navigation maps by transforming trajectory data from different vehicles into a unified frame.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS data is used for vehicle localization, then the system can provide global position information, but the data quality becomes unreliable especially in urban environments
Solution Approach 1:
The patent uses LIDAR data as an intermediary to bridge the gap between unreliable GPS measurements and the need for accurate localization. By matching LIDAR scans against pre-built maps and using pose graph optimization, the system obtains reliable position estimates that compensate for GPS deficiencies in urban environments.
Solution Approach 2:
The system merges multiple data sources including GPS, LIDAR, and pose graph optimization results to create a unified localization solution. This fusion allows the system to leverage the global coverage of GPS while correcting its inaccuracies using LIDAR-based relative positioning.
2Area of stationary object
If LIDAR data from multiple vehicles is collected for map creation, then more comprehensive coverage is achieved, but aligning the data becomes difficult in dynamic settings
Solution Approach 1:
The patent introduces pose graph optimization as an intermediary computational framework that simplifies the alignment of multi-vehicle LIDAR data. By representing vehicle poses as nodes and relative measurements as edges in a graph, the system efficiently solves the complex alignment problem through graph optimization techniques.
Solution Approach 2:
The system segments the map creation process into distinct phases: individual vehicle trajectory estimation using pose graphs, followed by integration of multiple trajectories into a unified map. This segmentation allows each vehicle's data to be processed independently before combination, reducing overall complexity.
3Measurement precision
If trajectory data from multiple vehicles is integrated, then map accuracy improves, but systematic errors from low-quality GPS propagate through the system
Solution Approach 1:
The patent uses pose graph optimization as an intermediary layer that filters out systematic GPS errors before integrating multi-vehicle trajectories. The pose graph computes relative poses based on LIDAR data and motion models, creating corrected trajectory estimates that eliminate cumulative GPS drift and systematic biases.
Solution Approach 2:
The system converts the harmful effect of noisy GPS measurements into a benefit by using them as one of multiple inputs to the pose graph optimization. Rather than directly using GPS positions, the system processes GPS data through the pose graph framework which reconciles it with LIDAR-based relative positioning, transforming errors into corrected estimates.
Data Source
AI summary
A georeferenced trajectory estimation system receives data generated by vehicle sensors from a fleet of vehicles and identifies sets of sensor measurements for a geographic region taken at different times. The sets of sensor measurements include environment geometry data and unaligned trajectory data in local coordinate frames. The georeferenced trajectory estimation system aligns the environment geometry data between sets of sensor measures, and, based on the alignment of the environment geometry data, transforms the identified corresponding trajectory data into a common coordinate frame.