AMR Relocalization Using Milestone Ranking and Adaptive Particle Clouds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional relocalization techniques for autonomous mobile robots (AMRs) are computationally intensive, requiring high-performance hardware and memory, and lack accuracy when using camera modules, especially when both laser scanners and cameras are available, necessitating a more efficient and accurate pose estimation method.
Innovation Solution
The system employs a reduced search environment by constructing a path from milestones linked by waypoints, ranking them based on connectivity and historical data, and using Monte Carlo Localization (MCL) on selected milestones with adaptive particle generation to relocalize AMRs efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional relocalization techniques are used, then relocalization can be performed, but computational load and memory requirements are high
Solution Approach 1:
The patent segments the relocalization process into distinct phases: candidate pose identification using laser scan matching, followed by verification using particle filter. This segmentation allows each phase to use the most appropriate algorithm for its specific task, reducing overall computational load while maintaining reliability.
Solution Approach 2:
The patent performs preliminary action by using laser scan matching to identify candidate poses before applying the computationally intensive particle filter method. This preliminary filtering step significantly reduces the search space, allowing the particle filter to converge faster with fewer particles, thereby reducing computational load and memory requirements.
2Reliability
If conventional relocalization techniques are used, then relocalization can be performed, but accuracy is insufficient when using camera modules
Solution Approach 1:
The patent merges two different sensing modalities: laser scanners and camera modules. By combining laser scan matching for candidate pose identification with particle filter verification using camera images, the system achieves higher measurement precision than either modality alone could provide.
Solution Approach 2:
The patent uses laser scan matching as an intermediary step to identify candidate poses before applying the particle filter with camera images. This intermediary process provides accurate initial estimates that guide the particle filter, significantly improving the final pose estimation accuracy.
3Measurement precision
If comprehensive search is performed for relocalization, then accuracy may be improved, but computational load increases
Solution Approach 1:
The patent segments the search process into two stages: a coarse search using laser scan matching to identify candidate poses, and a fine search using particle filter to verify and refine the pose. This segmentation achieves comprehensive search accuracy while controlling computational load by limiting the particle filter to only candidate poses.
Solution Approach 2:
The patent performs preliminary laser scan matching to identify and filter candidate poses before applying the computationally intensive particle filter. This preliminary action reduces the search space to only promising candidates, allowing comprehensive verification without excessive computational load.
Data Source
AI summary
Various aspects of methods, systems, and use cases include techniques for robotic relocalization. A robot may be configured to perform relocalization using operations to determine a cause of a loss of pose; use a nearest neighbor process to select a set of milestones from a roadmap when the cause of the loss of pose is due to a malfunction, or use a ranking process to select the set of milestones from the roadmap when the cause of the loss of pose is not due to the malfunction, the roadmap including a plurality of milestones; generate particle clouds around each milestone in the set of milestones; and perform localization on each milestone in the set of milestones to attempt to relocalize the robot.