Semantic Map Sharding for Autonomous Vehicle Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles require large and complex map data files for navigation, which lead to high network capacity and processing requirements, making them inefficient in terms of data size and memory usage.
Innovation Solution
The implementation of semantic map sharding, where a geographical area is split into sections, and semantic objects are stored in shard data entries, reducing the size of map data and processing load by generating shard data that can be used as an alternative to monolithic map data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If monolithic map data files are used for autonomous vehicle navigation, then complete geographical coverage is achieved, but network capacity requirements and processing load increase significantly
Solution Approach 1:
The patent divides the monolithic map data into multiple shard data files organized in a hierarchical structure. The geographical area is split into sections, with each section containing a subset of semantic objects. This segmentation reduces the size of individual data files that need to be transmitted and processed, while maintaining complete coverage when all shards are combined.
Solution Approach 2:
The patent extracts only the necessary semantic objects and geographical features required for autonomous navigation from the complete map data. By filtering and selecting relevant objects (such as roads, intersections, traffic signs) and storing them in a structured format, the system reduces data size while preserving navigation functionality.
2Adaptability or versatility
If large monolithic map data files are used, then comprehensive semantic objects coverage is achieved, but memory usage and processing efficiency deteriorate
Solution Approach 1:
The map data is segmented into multiple shard files organized in directories based on geographical sections. Each shard contains a manageable subset of semantic objects, allowing the system to process only the relevant portions needed for current navigation tasks rather than loading entire monolithic files into memory.
Solution Approach 2:
The patent introduces a hierarchical directory structure dimension to organize shard files. Instead of a flat monolithic structure, the system uses nested directories (e.g., region/section/shard) to arrange data spatially, enabling efficient access and processing by navigating through this dimensional hierarchy rather than searching through all data at once.
Data Source
AI summary
A method for generating shard data from a geographical area split into sections is proposed. The method uses data representative of a geographical area that is split into sections and data representative of semantic objects located in the geographical area. For each of the sections, semantic objects are searched that are located at least partly in a section. A shard data entry of the shard data is generated. The shard data entry includes one or more semantic objects found by the searching.