Lane-Level Map Generation from Crowdsourced Vehicle Position Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mapping systems for autonomous and semi-autonomous vehicles are time-consuming to generate and quickly become outdated due to road changes, as they rely on survey vehicles to capture data, which is not efficient for real-time lane-level information updates.
Innovation Solution
A method and system that processes vehicle position data from multiple vehicles to determine lane topology and traffic conditions, using pre-processing, dilation, aggregation, clustering, and road geometry correction based on yaw rate and constraints from sensors like cameras and lidars, to generate dynamic and accurate lane-level maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If survey vehicles are used to capture road data, then map accuracy is improved, but time consumption and update frequency are worsened
Solution Approach 1:
The system enables vehicles to automatically contribute their own position data to map generation without requiring dedicated survey vehicles. Each vehicle serves itself and simultaneously contributes to the collective map data, eliminating the need for specialized data collection vehicles and reducing time consumption while maintaining accuracy through multiple data sources.
Solution Approach 2:
Regular vehicles performing their normal transportation function simultaneously serve as data collection devices for map generation. The system transforms ordinary vehicles into multi-functional units that both transport passengers/goods and contribute to map accuracy, eliminating the need for separate survey vehicle operations.
2Measurement precision
If survey vehicles traverse the road network to capture data, then lane-level information is obtained, but the maps become outdated quickly due to road changes
Solution Approach 1:
The system enables continuous, real-time collection of position data from multiple vehicles as they continuously traverse the road network during normal operations. This continuous data stream ensures maps are constantly updated with current road conditions, preventing obsolescence while maintaining high lane-level accuracy through ongoing verification.
Solution Approach 2:
The system proactively collects and processes position data from multiple vehicles to anticipate and prepare for road changes before they affect map accuracy. By continuously monitoring position data trends and comparing against existing maps, the system can detect road changes early and update maps preemptively, maintaining currency without waiting for survey vehicle re-traversal.
3Productivity
If vehicle position data from multiple vehicles is processed to determine lane topology, then map update speed is improved, but data processing complexity increases
Solution Approach 1:
The system divides the complex task of map generation into distinct processing stages: data collection from multiple vehicles, pre-processing of position data, lane topology determination, and map generation. This segmentation allows each stage to be optimized independently and processed in parallel from multiple vehicle data streams, increasing update speed while managing complexity through modular processing.
Solution Approach 2:
The system combines position data from multiple vehicles into a unified dataset for collective processing. By merging data from numerous sources simultaneously and processing them together through the pipeline, the system achieves faster map updates through parallel information contribution while the unified processing approach manages complexity through consistent handling of aggregated data.
Data Source
AI summary
Methods and apparatus are provided for controlling a vehicle. In one embodiment, a method includes: receiving, by a processor, vehicle position data from the vehicle; processing, by the processor, vehicle position data with vehicle position data from other vehicles to determine a lane topology; processing, by the processor, the vehicle position data to determine traffic conditions within a lane of the lane topology; generating, by the processor, a map for controlling the vehicle based on the lane topology and the traffic conditions.