Semantic Map Sharding for High-Definition Geodata Loading

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Autonomous vehicles require large, high-definition maps that are typically stored as monolithic data files, which are cumbersome and resource-intensive, necessitating significant network capacity and processing power for uploading and processing.

Innovation Solution

Semantic map sharding splits high-definition map data into logical chunks called shards, allowing for efficient storage and processing by only loading relevant data sections, reducing network and memory requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If monolithic map data files are used to store high-definition maps for autonomous vehicles, then complete geographical coverage and semantic object information are provided, but network capacity and processing power requirements become excessively high

Engineering Contradiction:

Improvecompleteness of map dataVSAvoidnetwork capacity and processing power

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent divides the monolithic map data into multiple logical shards based on geographical sections. Each shard contains a subset of semantic objects and map features for a specific section, allowing selective loading and processing. This segmentation reduces the amount of data that needs to be transmitted over the network and processed by the autonomous vehicle's computing systems, directly addressing the contradiction between data completeness and resource consumption.

Inventive Principle:

Principle #1Segmentation

2Loss of information

If monolithic map data files are used, then all semantic objects and geographical information are available, but data transmission and processing become resource-intensive

Engineering Contradiction:

Improveavailability of semantic objectsVSAvoiddata size

Core Design Contradiction:

Loss of informationVSQuantity of substance

Solution Approach 1:

The patent extracts only the necessary semantic objects and map features relevant to the autonomous vehicle's current location and operational needs from the complete map data. By taking out only the required portions and storing them in separate shards, the system maintains availability of needed information while significantly reducing the total data size that must be managed and processed.

Inventive Principle:

Principle #2Taking out (Extraction)

3Area of stationary object

If complete high-definition map data is stored in a single file, then comprehensive geographical coverage is achieved, but storage and processing efficiency decrease

Engineering Contradiction:

Improvegeographical coverageVSAvoiddata processing efficiency

Core Design Contradiction:

Area of stationary objectVSProductivity

Solution Approach 1:

The patent segments the comprehensive geographical map data into multiple logical shards, each representing a specific geographical section. This allows the system to maintain complete geographical coverage while improving processing efficiency by loading and processing only the relevant shards needed for the vehicle's current operations, rather than processing the entire monolithic data set.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11526498B2Storage of shard data

Publication Date: 2022.12.13 GM CRUISE HOLDINGS LLC

AI summary

A method for storing shard data and data formats for storing shard data are proposed. Shard data entries are generated and stored, wherein each shard data entry comprises a definition of one or more semantic objects covered by the shard data entry. Shard metadata is generated and stored, wherein the metadata comprises references to the shard data entries and, for each of the shard data entries, data representative of a bounding box indicative of an area in a geographical area that is covered by the shard data entry.