Loop Closure Detection in SLAM Using Submap Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SLAM algorithms face challenges in accurately generating as-built floor plans for large and complex spaces without significantly increasing computational requirements, particularly in detecting loop closure to enhance pose estimates and global localization.
Innovation Solution
A computer-implemented method for detecting loop closure using range measurement data scans, which involves determining relative scan poses, aligning frames within submaps, periodically optimizing scan and submap poses, and identifying closed loops using probability grids and non-linear least squares optimization, with optional implementation of a branch and bound algorithm and depth-first search.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If SLAM algorithms are used to generate as-built floor plans swiftly and accurately, then productivity is improved, but measurement precision deteriorates for very large floor plans
Solution Approach 1:
The patent divides the large floor plan into multiple submaps, each representing a local area. The SLAM system processes and stores these submaps separately, allowing efficient local processing while maintaining the ability to detect global loop closures across submap boundaries. This segmentation enables swift processing of large spaces without sacrificing overall accuracy.
Solution Approach 2:
The patent performs preliminary loop closure detection by comparing current scan data against previously processed submaps and identifying potential loop closures before final map integration. This preliminary action allows the system to prepare correction data in advance, ensuring measurement precision is maintained even as productivity increases for very large floor plans.
2Measurement precision
If loop closure detection is implemented to enhance pose estimates, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent introduces submaps as intermediary structures that facilitate loop closure detection. Instead of directly comparing all scan data across the entire floor plan, the system uses submaps as intermediate representations, making the loop closure detection process more manageable and less complex while still achieving enhanced pose estimate accuracy.
Solution Approach 2:
By segmenting the environment into submaps, the patent reduces the computational complexity of loop closure detection. The system only needs to compare scans against relevant submaps rather than the entire map, significantly reducing algorithmic complexity while maintaining measurement precision for pose estimates.
3Measurement precision
If computational requirements are expanded to increase accuracy, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent segments the floor plan into submaps that can be processed independently and in parallel. This segmentation allows the system to apply computationally intensive accuracy-enhancing algorithms to smaller submap portions rather than the entire floor plan, maintaining high measurement precision while preserving productivity through efficient resource utilization.
Solution Approach 2:
The patent implements loop closure detection selectively based on confidence thresholds and environmental features. Rather than applying full computational processing to every scan, the system performs partial processing only where needed to achieve the required accuracy level, thus avoiding unnecessary computational overhead that would reduce productivity.
Data Source
AI summary
Computer-implemented methods and systems for detecting loop closure in SLAM applications can include accessing one or more range measurement data scans that each provide a collection of consecutively observed spatial data obtained at a given location. A relative scan pose for each range measurement data scan can be determined. Frames from each of the one or more range measurement data scans then can be aligned to frames within one or more submaps based at least in part by the determined relative scan pose. An enhanced scan pose for each range measurement data scan and an enhanced submap pose for each submap can be periodically determined and used to identify one or more closed loops within the one or more submaps. An output indicative of the identified one or more closed loops can be provided, along with optionally generated floorplan maps and/or given location determinations.