Hyper-Graph Map Partitioning for Large-Area SLAM Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As autonomous vehicles use SLAM-based mapping, the computational burden increases with larger graphs, making it inefficient to optimize maps of large areas such as continents or countries, as the graph optimization requires significant computational resources.
Innovation Solution
The use of hyper-graph data structures, where nodes can be subgraphs representing smaller areas, allows for partitioning of larger areas into manageable subgraphs, reducing computational requirements by optimizing only the subgraphs relevant to the vehicle's operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If SLAM-based mapping is used to generate maps of large areas, then the map coverage area increases, but the computational burden increases significantly
Solution Approach 1:
The patent divides a large-area map into multiple smaller submaps, each covering a specific geographic region. Each submap is independently optimized and stored, allowing the system to manage large coverage areas without optimizing the entire map at once. This segmentation reduces the computational burden by limiting optimization scope to individual submaps rather than the complete large-area map.
2Loss of information
If the graph size increases to cover larger areas, then the map completeness improves, but the optimization time and resources increase
Solution Approach 1:
The map is segmented into multiple submaps that collectively provide complete coverage of the large area. Each submap maintains completeness for its specific region through independent optimization, while the overall system achieves comprehensive coverage by combining multiple submaps. This approach ensures map completeness without requiring simultaneous optimization of the entire large-area graph.
Solution Approach 2:
The system performs optimization on individual submaps rather than the complete large-area map. This partial action approach optimizes only the necessary portions (individual submaps) at any given time, significantly reducing optimization time and computational resources while still achieving complete map coverage through the aggregation of optimized submaps.
3Measurement precision
If the graph includes all nodes and edges for large areas, then the navigation accuracy improves, but the computational resources required increase significantly
Solution Approach 1:
The navigation system operates on segmented submaps rather than a single large-area graph. Each submap contains nodes and edges relevant to its specific geographic region, maintaining navigation accuracy within that region while reducing the total number of nodes and edges that need to be processed simultaneously. This segmentation allows the system to preserve navigation precision without the computational overhead of managing a complete large-area graph.
Data Source
AI summary
Among other things, methods for generating maps using hyper-graph data structures are disclosed. The method can include receiving and storing data from at least one sensor of a vehicle in an environment. The method can include generating, based on the received data, a graph, having at least one node corresponding to at least one subgraph. The at least one subgraph can include subgraph nodes corresponding to geographical and/or logical positions. The subgraph nodes can be connected by subgraph edges representing spatial constraints and/or logical connections. The at least one subgraph can include contextual data classifying each of the subgraph nodes according to a property of the environment associated with the subgraph.