Vehicle Self-Localization Using Object-to-Map Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for road geometry and environment modeling are resource-intensive and time-consuming, making them impractical for modern applications requiring efficient localization in environments without satellite navigation, especially for autonomous vehicle control.
Innovation Solution
A modularized localization framework using on-board sensors and high-definition maps, combined with a decision module to dynamically allocate resources and optimize workload, allowing for efficient and accurate vehicle localization through object recognition and comparison with ground truth objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional road geometry modeling and environment modeling methods are used, then accurate localization can be achieved, but the process becomes resource-intensive and time-consuming
Solution Approach 1:
The patent segments the localization process into distinct modules: sensor data acquisition, object detection, map matching, and localization calculation. Each module handles specific tasks independently, improving overall processing efficiency while maintaining accuracy. The road geometry modeling and environment modeling are separated into distinct processing streams that can be optimized independently.
Solution Approach 2:
The patent performs preliminary actions by pre-processing sensor data and pre-detecting objects before the actual localization calculation. High-definition maps are pre-generated with road geometry and environment features extracted in advance, so that during runtime, the system only needs to match detected objects with pre-processed map data, significantly reducing computation time.
2Measurement precision
If comprehensive sensor data processing and object detection are performed, then localization accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The patent applies partial action by selectively processing only the most relevant sensor data and objects for localization. The system identifies and focuses on key objects (road signs, poles, lane markings) that provide the most significant localization information, rather than processing all detected objects equally. This reduces computational load while maintaining accuracy.
Solution Approach 2:
The patent applies local quality by allocating different processing resources to different types of objects and data based on their importance. Critical objects like road signs and lane markings receive higher processing priority and more computational resources, while less critical objects receive minimal processing. This selective approach reduces overall processing time while maintaining localization accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods described herein relate to self-localization. Methods may include: receiving sensor data from a vehicle traveling along a road segment; identifying one or more objects of the environment from the sensor data; determining a coarse location of the vehicle; comparing the identified one or more objects with corresponding ground truth objects; calculating an observation in response to the comparison of the identified one or more objects with the corresponding ground truth objects; determining a location of the vehicle with a higher precision than the coarse location based, at least in part, on the observation; and allocating resources of the apparatus based, at least in part, on the observation.