SLAM Bundle Adjustment with Selective Landmark Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SLAM methods struggle to balance accuracy, consistency, and computational cost, often sacrificing global consistency for speed, leading to potential drifting and inaccuracies in the map or modified map, especially in resource-constrained environments.
Innovation Solution
A method that selectively modifies landmarks based on the recency of observations, optimizing recently observed landmarks and preserving those unlikely to benefit from optimization, thereby reducing the number of parameters in bundle adjustment and computation time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bundle adjustment is performed on all landmarks to maintain global consistency, then mapping accuracy is improved, but computation time and resource utilization increase significantly
Solution Approach 1:
The patent segments the set of all landmarks into two distinct groups: recently observed landmarks and previously observed landmarks. This segmentation allows the system to apply different processing strategies to different subsets, performing bundle adjustment only on the recently observed portion while freezing the previously observed landmarks, thereby reducing computational complexity while maintaining accuracy for the most relevant features
Solution Approach 2:
The patent applies local quality by making recently observed landmarks mutable (subject to optimization) while making previously observed landmarks immutable (frozen). This creates a spatial and temporal differentiation in how landmarks are treated, allowing the system to focus computational resources on landmarks that are most likely to improve current mapping accuracy without the overhead of processing all landmarks equally
2Reliability
If bundle adjustment is performed on all landmarks to ensure map consistency, then mapping reliability is improved, but device resource utilization increases
Solution Approach 1:
The patent segments landmarks based on observation recency, creating a mutable set (recently observed) and an immutable set (previously observed). This segmentation enables the system to maintain map consistency through selective optimization of only the recently observed landmarks, significantly reducing the computational burden and energy consumption compared to optimizing all landmarks while preserving the reliability benefits for the most relevant features
Solution Approach 2:
The patent changes the state parameter of landmarks from a uniform mutable state to a differentiated state based on observation recency. By transitioning previously observed landmarks to an immutable state and keeping only recently observed landmarks as mutable, the system reduces the number of optimization parameters, thereby lowering device resource utilization while maintaining sufficient map consistency
3Measurement precision
If all landmarks are optimized in bundle adjustment to reduce drift, then localization accuracy is improved, but computational cost increases
Solution Approach 1:
The patent segments the landmark set into recently observed (mutable) and previously observed (immutable) groups. This segmentation reduces the scope of bundle adjustment to only the recently observed landmarks, decreasing the number of optimization variables and computational operations required while maintaining localization accuracy for the most relevant features that are most likely to exhibit drift
Solution Approach 2:
The patent applies local quality by making only recently observed landmarks subject to optimization while freezing previously observed landmarks. This creates a differentiated processing approach where computational resources are concentrated on landmarks that are most likely to benefit from optimization and contribute most to reducing drift, rather than uniformly processing all landmarks
Data Source
AI summary
A method including identifying a landmark in image data captured by a device, receiving pose data associated with the image data, determining that the landmark is included in a map of an environment, and in response to determining that the landmark is included in the map, determining that a timestamp associated with the image data meets a criterion, and in response to the device determining that the timestamp meets the criterion, modifying the map based on the landmark and the pose data.