Camera Localization Using HD Map Edgels Without LiDAR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurate localization due to limitations in conventional maps and GPS systems, which lack the precision needed for safe navigation, especially when sensors are obscured or lack sufficient depth information.
Innovation Solution
The use of high-definition maps and camera-based localization methods that employ imaging systems to determine the vehicle's pose without relying on depth sensors like LiDAR, by projecting edgels onto image frames and optimizing transformations to minimize distance between 3D edge locations and their corresponding pixels, allowing for precise vehicle positioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS systems are used for vehicle localization, then the system is simple and provides global coverage, but the localization accuracy deteriorates to over 100 meters
Solution Approach 1:
The patent introduces HD maps as an intermediary between GPS and the vehicle localization system. The HD maps provide high-precision pre-stored road geometry and feature data that serve as a reference framework, enabling the vehicle to determine its position with 10 cm or better accuracy by matching sensor data against the detailed map features, thus resolving the accuracy limitation of GPS without requiring a completely new system.
Solution Approach 2:
The patent transitions from relying solely on satellite-based 3D positioning (GPS) to incorporating 2D map-based reference data (HD maps with road geometry and features). This dimensional shift allows the system to leverage pre-captured high-precision map information from aerial or ground-based platforms to achieve superior localization accuracy compared to GPS alone.
2Measurement precision
If conventional maps are used for navigation, then the system is simple to implement, but the map accuracy deteriorates and cannot provide the 10 cm or less precision required for safe navigation
Solution Approach 1:
The patent applies preliminary action by pre-capturing and storing high-precision map data before vehicle operation. HD maps are created in advance using aerial photography, LiDAR scanning, or other high-accuracy surveying methods, and this pre-acquired data is stored for continuous use during vehicle navigation, eliminating the need for real-time high-precision mapping while achieving 10 cm or better accuracy.
Solution Approach 2:
The patent uses copying by creating detailed replicas of the physical road environment in digital HD map form. These maps contain precise copies of road geometry, lane markings, curbs, and other features at 10 cm or better accuracy, which vehicles can reference for localization without needing to physically measure or recreate the environment during operation.
3Reliability
If sensor systems are used for real-time observation, then the vehicle can make driving decisions on the fly, but the sensors cannot observe everything all the time due to obstructions by corners, rolling hills, and other vehicles
Solution Approach 1:
The patent applies beforehand cushioning by pre-storing comprehensive environmental information in HD maps during periods when the vehicle does not need to react immediately. These maps capture road geometry, features, and surroundings in advance, providing a buffer of known information that compensates for moments when real-time sensors are obscured or cannot observe critical features, thus enhancing observation reliability.
Solution Approach 2:
The patent merges real-time sensor data with pre-acquired HD map information to create a more reliable and complete environmental model. By combining live camera and LiDAR observations with the detailed reference framework from HD maps, the system achieves more reliable localization and perception than either source could provide alone, especially when sensors are temporarily obstructed.
Data Source
AI summary
Camera based localization performed to determine a current pose of an autonomous vehicle without the aid of depth sensors such as LiDAR. The vehicle comprises an imaging system configured to capture image frames depicting portions of the surrounding area. Based on an initial pose of the vehicle, edgels corresponding to three-dimensional locations are loaded and mapped to corresponding edge pixels of the captured image frame. A pose of the vehicle is optimized based upon the determined correspondences by identifying a transformation that minimizes a distance between the edgels and their corresponding edge pixels. The determined transformation can be applied to the initial pose to determine an updated pose of the vehicle.