Distributed Pose Graph Optimization for Fresh HD Vehicle Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional maps for autonomous vehicles lack the accuracy and timeliness required for safe navigation, as they are often outdated and rely on expensive and time-consuming survey processes, with GNSS systems providing insufficient precision and coverage.
Innovation Solution
The development of high-definition (HD) maps that utilize pose graphs and distributed computing to generate and maintain accurate, up-to-date maps by combining data from multiple vehicles, allowing for real-time updates and precise localization of autonomous vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional survey teams use drivers with specially outfitted cars and high resolution sensors to create maps, then map accuracy can be improved, but the process becomes expensive and time consuming
Solution Approach 1:
The patent segments the map creation process into distributed tasks performed by multiple vehicles. Instead of using a single survey team, the system divides the geographic region into multiple areas, each covered by different vehicles equipped with sensors. This allows parallel processing of map data collection, significantly reducing the time required while maintaining accuracy through multiple measurement points.
Solution Approach 2:
The patent transforms regular vehicles into multi-functional units that can both transport passengers/cargo and simultaneously collect mapping data. The vehicles serve dual purposes: their primary transportation function and their secondary function as mobile mapping stations. This eliminates the need for dedicated survey vehicles and enables continuous map updates using the existing vehicle fleet.
2Reliability
If survey fleets include a thousand cars to capture road updates, then map freshness can be improved, but the cost and complexity of the system increases
Solution Approach 1:
The patent implements a self-service mapping system where vehicles automatically collect, process, and contribute their own sensor data to the map without requiring centralized coordination of a large survey fleet. Each vehicle independently performs mapping tasks in its operational area and uploads data to the central system, which automatically integrates it. This eliminates the need for complex fleet management while maintaining map freshness through continuous autonomous data collection.
Solution Approach 2:
The system establishes a feedback loop where map data collected from vehicles is continuously analyzed, updated, and redistributed to the fleet. The central processing system receives sensor data from vehicles, processes it to detect road changes, updates the map accordingly, and makes the updated map available to all vehicles. This closed-loop feedback mechanism ensures map freshness without requiring a large dedicated survey fleet.
3Area of stationary object
If GNSS based systems are used for vehicle localization, then coverage area can be improved, but measurement precision deteriorates with large error conditions
Solution Approach 1:
The patent merges multiple localization data sources to compensate for GNSS limitations. It combines GNSS global coverage capability with local sensor data (LIDAR, cameras, inertial sensors) and pre-generated HD maps to create a hybrid localization system. The GNSS provides broad area coverage and initial position estimation, while local sensors and map matching provide precise location refinement, achieving both wide coverage and high accuracy simultaneously.
Solution Approach 2:
The patent introduces HD maps as an intermediary layer between GNSS and the vehicle's actual position. Instead of relying directly on GNSS coordinates, the system uses GNSS to estimate vehicle location, then refines this estimate by matching sensor data against the detailed HD map features. The HD map acts as a mediator that translates coarse GNSS positions into precise vehicle locations by comparing observed environmental features with map data, thereby improving accuracy while maintaining wide coverage.
Data Source
AI summary
According to an aspect of an embodiment, operations may comprise obtaining a pose graph that comprises a plurality of nodes. The operations may also comprise dividing the pose graph into a plurality of pose subgraphs, each pose subgraph comprising one or more respective pose subgraph interior nodes and one or more respective pose subgraph boundary nodes. The operations may also comprise generating one or more boundary subgraphs based on the plurality of pose subgraphs, each of the one or more boundary subgraphs comprising one or more respective boundary subgraph boundary nodes and comprising one or more respective boundary subgraph interior nodes. The operations may also comprise obtaining an optimized pose graph by performing a pose graph optimization. The pose graph optimization may comprise performing a pose subgraph optimization of the plurality of pose subgraphs and performing a boundary subgraph optimization of the plurality of boundary subgraphs.