Compressed Octrees for HD Map Nearest Neighbor Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional maps for autonomous vehicles lack precision, accuracy, and timeliness, making them inadequate for safe navigation due to limitations in sensor detection and the expense and inefficiency of traditional mapping methods.
Innovation Solution
The use of compressed octrees to represent high-definition maps, allowing for efficient storage, transmission, and real-time updates, enabling precise localization and navigation with a nearest neighbor search algorithm that identifies relevant points within a search radius.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional maps are used for autonomous vehicle navigation, then the system is simple to implement, but the precision and accuracy are insufficient for safe navigation within 30cm threshold
Solution Approach 1:
The high definition map is segmented into multiple tiles covering different geographic regions. Each tile contains map data for a specific area, allowing the system to load only relevant tiles for the vehicle's current location. This segmentation enables precise navigation within 30cm threshold while managing system complexity through modular data organization and selective loading.
2Reliability
If survey teams create comprehensive maps using expensive survey cars, then map accuracy improves, but the time and cost increase significantly, and maps cannot be updated frequently enough to keep pace with road changes
Solution Approach 1:
Autonomous vehicles themselves contribute to map creation and updating by collecting sensor data during normal operation. The vehicles capture point cloud data, images, and other sensor information, which is then processed to update the high definition map. This self-service approach enables frequent map updates to keep pace with road changes without requiring expensive survey teams, improving both map freshness and update speed.
Solution Approach 2:
The system creates compressed representations (copies) of the high definition map data that can be efficiently stored and transmitted. Instead of storing complete uncompressed point cloud data, the system uses compressed formats that retain essential navigation information while reducing storage requirements and enabling faster updates.
3Measurement precision
If high resolution sensor data is collected continuously to maintain accurate maps, then map accuracy improves, but storage requirements and data transmission bandwidth increase significantly
Solution Approach 1:
The system changes the parameter of data representation by applying compression algorithms to the point cloud data. The compression transforms the raw high-resolution sensor data into a more efficient format that retains the essential geometric and semantic information needed for accurate navigation while significantly reducing the volume of data that needs to be stored and transmitted.
4Speed
If complete map data is loaded into memory for fast access, then navigation speed improves, but memory consumption increases beyond available vehicle memory capacity
Solution Approach 1:
The map data is segmented into tiles that can be loaded selectively into memory based on the vehicle's current location and navigation needs. Only the tiles relevant to the vehicle's current area and upcoming route are loaded into memory, enabling fast access to necessary map data while keeping memory consumption within vehicle capacity limits.
Solution Approach 2:
The system performs preliminary actions by pre-loading map tiles that are likely to be needed based on the vehicle's current position and planned route. This anticipatory loading ensures that necessary map data is already in memory when needed, maintaining fast navigation speed without requiring all map data to be loaded simultaneously.
Data Source
AI summary
According to an aspect of an embodiment, operations may comprise receiving a search query for points near a query-point, accessing a compressed octree representation of a point cloud comprising 3D points of a region, and traversing the compressed octree representation to identify regions that overlap a search space by, marking a current node as overlapping the search space responsive to determining that the current node is a leaf node, identifying a child node of the current node and performing a nearest neighbor search in the child node responsive to determining that a region represented by the current node overlaps the search space, and identifying a sibling node of the current node and performing the nearest neighbor search in the sibling node responsive to determining that a region represented by the current node does not overlap the search space.