Semantic Map Sharding for Autonomous Vehicle Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles require large, high-definition maps that are inefficiently stored and processed as monolithic data files, leading to high network capacity and processing requirements, which can be cumbersome for both vehicles and simulation environments.
Innovation Solution
Semantic map sharding splits high-definition maps into logical chunks called shards, allowing for efficient storage and processing by generating shard data based on semantic objects and geographical sections, reducing data size and processing load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high-definition maps are stored as monolithic data files, then complete geographical coverage is achieved, but network capacity requirements and processing load increase significantly
Solution Approach 1:
The patent divides monolithic map data into multiple shard files organized in a hierarchical directory structure. Each shard contains a subset of semantic objects (traffic lights, signs, lane boundaries, etc.) for a specific geographical area. This segmentation allows autonomous vehicles to download and process only the map shards relevant to their current location and route, dramatically reducing network capacity requirements while maintaining complete geographical coverage when needed.
2Reliability
If high-definition maps are stored as monolithic data files, then complete geographical coverage is achieved, but processing load increases significantly
Solution Approach 1:
The map data is segmented into shard files with manageable sizes, each containing semantic objects for specific geographical sections. The autonomous vehicle's processing system loads only the necessary shards into memory based on current location and navigation needs, rather than loading entire monolithic map files. This reduces processor load and memory requirements while maintaining access to complete map data when needed.
3Use of energy by moving object
If semantic map sharding is implemented, then network capacity requirements and processing load are minimized, but data structure complexity increases
Solution Approach 1:
The patent implements a hierarchical directory structure where shard files are organized by geographical regions and semantic object types. Each shard file follows a standardized format with consistent metadata schemas, making the complex data structure manageable through systematic organization. Index files maintain mappings between geographical coordinates and relevant shard locations, enabling efficient data retrieval without requiring the system to process the entire complex structure.
Solution Approach 2:
The patent introduces index files and metadata structures that act as intermediaries between the autonomous vehicle's navigation system and the actual map shard data. These intermediary layers translate geographical queries into specific shard file requests, simplifying the interaction with the underlying complex segmented data structure while maintaining efficient access patterns.
Data Source
AI summary
A system and method for generating shard data based on semantic sharding is proposed. Shard data is generated by a map sharder that is configured to generate the shard data based on semantic objects in a geographical area and a definition of sections. The sections cover parts of the geographical area. The map sharder searches, for each of the sections, for semantic objects that are located at least partly in a section and stores found objects in a shard data entry of the shard data.