Dynamic Submap Localization for Changing Environments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation systems face challenges in long-term autonomy in changing environments due to issues like structural changes, lighting conditions, weather, and seasonal variations, which traditional mapping approaches struggle to adapt to effectively.
Innovation Solution
A method that dynamically stores and updates experiences based on visual features, allowing vehicles to recognize and localize in changing environments by creating a 'plastic map' that adapts to new conditions, storing new experiences only when necessary and linking them topologically to maintain accurate localization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a monolithic map is created to enable localization in changing environments, then localization accuracy is improved, but storage requirements and processing complexity increase
Solution Approach 1:
The patent segments the monolithic map into multiple submaps, each representing a specific time period or environmental condition. Instead of storing one large comprehensive map, the system creates several smaller submaps that can be independently stored and processed. When localization is needed, the system selects and combines relevant submaps based on current environmental conditions, thereby reducing individual storage requirements while maintaining localization accuracy.
Solution Approach 2:
The patent implements a dynamic map management system where submaps are created, updated, and discarded based on environmental changes and localization needs. Rather than maintaining a static monolithic map, the system dynamically adjusts which submaps are active and how they are combined, allowing the map structure to adapt to changing conditions while optimizing storage utilization.
2Measurement precision
If a monolithic map is created to enable localization in changing environments, then localization accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent divides the processing task of localizing in a large monolithic map into multiple smaller tasks of localizing in individual submaps. Each submap can be processed independently with simpler algorithms, and the results are then combined. This segmentation reduces the computational complexity of each individual processing step while maintaining overall localization accuracy.
Solution Approach 2:
The system dynamically selects which submaps to process based on current environmental conditions and localization requirements. Rather than continuously processing all map data, the system activates only the relevant submaps needed for current localization tasks, thereby reducing processing complexity and computational resource requirements.
3Adaptability or versatility
If all sensed scenes are stored to handle environmental changes, then adaptability is improved, but storage requirements increase
Solution Approach 1:
The patent segments the storage of sensed scenes into multiple submaps organized by time periods or environmental conditions. Instead of storing all scenes in a single large repository, the system divides them into manageable submaps that can be selectively stored and retrieved. This segmentation maintains adaptability to environmental changes while reducing the storage burden on any single storage unit.
Solution Approach 2:
The patent implements a strategy where older or less relevant submaps are discarded or archived when they are no longer needed for current localization tasks, while the ability to recover and reuse them is maintained when needed. This approach allows the system to maintain adaptability to environmental changes over time while managing storage requirements by selectively discarding and recovering submaps based on current needs.
Data Source
Figure 1
Figure 3
Figure 4~5
AI summary
A method of localising a vehicle hosting a sensor comprising the steps of capturing data from the sensor providing a sensed scene around the vehicle at a current time, processing the sensed scene to extract a set of features from the sensed scene and to determine, using the extracted features, a position of the vehicle from the sensed scene in relation to the position of the sensor at an earlier time, comparing the extracted set of features from the sensed scene to one or more stored experiences, wherein each stored experience comprises a plurality of sets of features, where each set of features has been determined from a previously sensed scene, to ascertain whether the sensed scene can be recognised within any of the stored experiences, and if the extracted set of features of the sensed scene are recognised within any one of the stored experiences then that stored experience is used to produce an estimate of the position of the vehicle at the current time with respect to the stored experience.