Dynamic 3D Map Complexity Reduction for Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face complexity and inconsistency in 3D maps with multiple elevation planes, leading to increased computational effort and potential erroneous driving maneuvers due to irrelevant road features being categorized as relevant.
Innovation Solution
A method and system dynamically reduce 3D map complexity by using environment detection sensors to generate a reduced map that only includes relevant roads for the ego vehicle, employing a reduction algorithm that prioritizes roads based on proximity, navigability, and elevation, using graph search algorithms like Dijkstra's or breadth-first search to ensure non-overlapping roads are included.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a strict change of the map and entire environment model from one plane to the next is performed, then navigation between elevation planes is enabled, but consistency in the environment model is lost
Solution Approach 1:
The patent applies local quality by maintaining different map representations for different elevation levels. Each elevation plane has its own 2D map with specific road networks and features, while the 3D map integrates these planes with elevation information. This allows the system to navigate between planes while maintaining consistency within each plane's local environment model.
2Loss of information
If the multi-layer 3D map is directly passed to downstream AD components, then complete spatial information is provided, but complexity and computational effort in all subsequent AD components significantly increases
Solution Approach 1:
The patent segments the 3D map into multiple 2D elevation planes, each with its own simplified road network and features. Instead of processing a single complex multi-layer 3D map, downstream AD components can work with individual 2D maps corresponding to the current elevation level, significantly reducing computational complexity while maintaining necessary spatial information.
Solution Approach 2:
The patent transforms the 3D map data into multiple 2D representations by projecting each elevation plane onto a 2D plane. This dimensionality reduction allows downstream systems to process simpler 2D maps while the system maintains the ability to navigate between elevation levels through the integrated 3D map structure.
3Loss of information
If all roads in the multi-layer 3D map are categorized as relevant, then complete road coverage is ensured, but erroneous driving maneuvers occur due to irrelevant roads being included
Solution Approach 1:
The patent performs preliminary filtering of roads by elevation level before processing. By determining the current elevation level of the ego vehicle and pre-selecting only roads relevant to that elevation level, the system avoids including irrelevant roads from other elevation planes, thereby preventing erroneous driving maneuvers while maintaining complete coverage of relevant roads.
Data Source
AI summary
A method for dynamically reducing the complexity of a 3D map in an ego vehicle including recording an environment of the ego vehicle; providing a 3D map; localizing the ego vehicle in the 3D map; and reducing the 3D map by removing roads that overlap each other on different levels of elevation.