Semantic Map Update for Autonomous Vehicle Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual maps used by autonomous vehicles often contain undesirable artifacts and extraneous data from sensors, leading to inefficient processing, inaccurate information, and slow calculations, which can impact the vehicle's navigation and control.
Innovation Solution
A system that segments sensor data to remove dynamic objects and artifacts, associates data with voxel spaces, and updates maps by incrementing or decrementing counters based on sensor data, allowing for the omission of extraneous information and the weighting of static objects for improved localization and navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If sensor data is used to create virtual maps, then map coverage and information content are improved, but processing efficiency and calculation speed deteriorate due to extraneous information and artifacts
Solution Approach 1:
The patent segments sensor data into multiple categories (static objects, dynamic objects, potentially dynamic objects, and artifacts) and processes each segment differently. This segmentation allows the system to retain useful information while discarding extraneous data, resolving the contradiction between information content and processing efficiency
Solution Approach 2:
The patent extracts and removes artifacts and dynamic objects from the sensor data before creating the virtual map. By taking out the harmful extraneous information, the system maintains high information content for navigation while eliminating data that would slow down processing
2Measurement precision
If all sensor data is processed and stored in the map, then measurement precision is improved, but device complexity and memory requirements increase
Solution Approach 1:
The patent extracts only the essential static object data from sensor readings, removing dynamic and potentially dynamic objects. This extraction process reduces memory requirements while preserving the precision needed for localization by focusing on stable, reliable features
Solution Approach 2:
The patent applies different quality standards to different types of data: static objects are processed with high precision for localization, while dynamic objects and artifacts are either removed or processed with lower fidelity. This local quality approach optimizes memory usage while maintaining necessary measurement precision
3Loss of information
If dynamic objects are included in the map, then semantic information is improved, but reliability deteriorates due to moving objects impacting localization
Solution Approach 1:
The patent segments objects by their motion characteristics, separating static objects (reliable for localization) from dynamic and potentially dynamic objects. This segmentation allows the system to maintain high localization reliability by using only static objects, while still preserving semantic information about the environment
Solution Approach 2:
The patent changes the parameter of object persistence by only including objects that remain static across multiple sensor readings in the final map. This parameter change ensures that only reliable, non-moving features are used for localization, while dynamic objects are excluded despite their semantic value
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system may receive a sensor dataset representing an environment and use the dataset to create or update a map. In creating or updating the map, the system may determine an object classification of one or more detected objects and only selectively incorporate data into the map based at least in part on the classification. The map may be associated with the classification (or semantic) information of the objects, as well as weights based on the classification. Similarly, datasets with selected classes of data removed may be used for system localization. Further, the system may determine an object track of the objects. When updating the map, voxels in a voxel space may indicate an occupied voxel based on a threshold number of observances. The object track and clean map can then be used for controlling an autonomous vehicle.