Trajectory Segmentation for GNSS-Limited HD Map Recording Runs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for creating and updating high-definition (HD) maps for autonomous driving face challenges in data collection granularity and frequent updates, particularly when using swarm data, due to inaccuracies in GNSS positioning and odometry drift, leading to high data processing efforts.
Innovation Solution
A method for segmenting recording runs based on GNSS accuracy and odometry data to form segments with defined minimum lengths, ensuring start and end regions have accurate GNSS data, and aligning these segments with map sections for efficient SLAM processing, using a kD tree for data management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If swarm data is used for HD map creation, then data coverage and update frequency are improved, but data processing complexity and computational effort increase due to GNSS inaccuracies and odometry drift
Solution Approach 1:
The patent segments the trajectory into multiple sections based on GNSS accuracy thresholds and odometry drift detection. Each segment is processed independently, allowing parallel computation and reducing overall processing complexity. This segmentation enables the system to handle swarm data more efficiently by breaking down the complex global problem into manageable local sections.
Solution Approach 2:
The patent applies different processing quality levels to different segments based on their GNSS accuracy and odometry characteristics. High-accuracy segments receive standard processing, while low-accuracy segments undergo more rigorous validation and correction. This local quality approach optimizes computational resources by applying intensive processing only where necessary.
2Speed
If trajectory segmentation is performed without considering GNSS accuracy, then processing speed is improved, but positioning precision deteriorates due to inclusion of inaccurate segments
Solution Approach 1:
The patent performs preliminary segmentation of the trajectory based on GNSS accuracy thresholds and odometry drift detection before the main processing step. This preliminary action identifies and isolates low-accuracy segments, allowing them to be handled separately with appropriate correction algorithms, thus preventing them from degrading the overall positioning precision while maintaining processing efficiency.
Solution Approach 2:
Different processing approaches are applied to different segments based on their quality characteristics. High-accuracy segments are processed quickly with standard algorithms, while low-accuracy segments receive enhanced processing including odometry correction and cross-validation. This local quality differentiation maintains positioning precision without sacrificing overall processing speed.
3Stability of the object's composition
If minimum segment length is increased, then data consistency within segments is improved, but the number of segments decreases reducing parallel processing capability
Solution Approach 1:
The patent implements a hierarchical segmentation strategy where the trajectory is first divided into larger consistency-based segments, which are then further subdivided into smaller parallelizable units. This multi-level segmentation maintains data consistency within each segment while creating sufficient independent units for effective parallel processing, thus resolving the contradiction between consistency and parallelism.
Solution Approach 2:
The minimum segment length is not fixed but dynamically adjusted based on the characteristics of the trajectory being processed. In regions with high GNSS accuracy and low odometry drift, longer segments are used to maximize consistency. In regions with lower accuracy, shorter segments are used to maintain parallel processing efficiency. This dynamic adaptation allows the system to optimize both consistency and parallelism for each specific case.
Data Source
AI summary
Method for segmenting a trajectory (1) of a recording run (2) for the swarm-data-based compilation of HD maps, sensor data being acquired along the trajectory (1) during the recording run (2), and the sensor data being able to include GNSS measurement data, comprising the following steps:a) detecting points (4) on the trajectory (1) at which position determination is not possible using GNSS measurement data alone or the accuracy of the position determined using GNSS measurement data alone does not attain a predetermined positioning accuracy, andb) segmenting the trajectory (1) into multiple segments (3) in such a way that each segment (3) attains a predetermined minimum length (5) and that at least one point (4) detected in step a) is located neither in the start region (6) nor in the end region (7) of each segment (3).