Vision-Aided Inertial Navigation Feature Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vision-aided inertial navigation systems face challenges in efficiently processing image and inertial data to accurately estimate the position and orientation of a sensing platform, especially in GPS-denied environments, due to limitations in handling large feature sets and computational resources.
Innovation Solution
The implementation of an inverse sliding-window filter (ISWF) that classifies features into SLAM and MSCKF categories, where SLAM features are maintained within the state vector and MSCKF features generate constraints, allowing for efficient computation of state estimates by excluding MSCKF features from the state vector and using them to constrain other poses within the sliding window.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all observed features are maintained in the state vector for accurate pose estimation, then measurement precision is improved, but device complexity increases due to larger state vectors and computational burden
Solution Approach 1:
The patent segments the set of observed features into two distinct categories: SLAM features (maintained in the state vector) and MSCKF features (used for constraints but not maintained). This segmentation allows the system to process a large total number of features while keeping the state vector size manageable, thereby resolving the contradiction between measurement precision and device complexity
Solution Approach 2:
The patent extracts MSCKF features from the state vector while still utilizing them for pose estimation through constraint computation. By taking out these features from the state vector maintenance burden but keeping them active for measurements, the system achieves accurate pose estimation without the computational overhead of maintaining all features in the state vector
2Measurement precision
If a large number of features are processed for each pose estimation, then measurement precision is improved, but productivity decreases due to increased computational time
Solution Approach 1:
The patent segments features into SLAM and MSCKF categories with different processing treatments. SLAM features are maintained in the state vector and processed through standard estimation algorithms, while MSCKF features are used to compute constraints between poses without being maintained in the state vector. This segmentation enables efficient real-time processing while maintaining high estimation accuracy by leveraging both feature types appropriately
Solution Approach 2:
The patent applies partial action by selectively maintaining only SLAM features in the state vector while using MSCKF features purely for constraint computation. This partial maintenance approach reduces computational time per frame while still utilizing information from all observed features, thereby achieving real-time processing capability without sacrificing measurement precision
3Measurement precision
If the state vector includes positions for all features, then measurement precision is improved, but loss of time increases due to longer computation cycles
Solution Approach 1:
The patent segments the feature set into SLAM features (with positions maintained in the state vector) and MSCKF features (used for constraints but without maintaining position states). This segmentation allows the system to compute constraints between poses using information from all features while avoiding the time cost of maintaining and updating position states for all features, thus reducing computation time per update while preserving localization accuracy
Solution Approach 2:
The patent extracts the position state maintenance requirement from MSCKF features while retaining their utility for constraint computation. By taking out the state maintenance burden from MSCKF features, the system achieves accurate localization using all feature observations without the time penalty of maintaining position states for every observed feature
Data Source
AI summary
This disclosure describes inverse filtering and square root inverse filtering techniques for optimizing the performance of a vision-aided inertial navigation system (VINS). In one example, instead of keeping all features in the system's state vector as SLAM features, which can be inefficient when the number of features per frame is large or their track length is short, an estimator of the VINS may classify the features into either SLAM or MSCKF features. The SLAM features are used for SLAM-based state estimation, while the MSCKF features are used to further constrain the poses in the sliding window. In one example, a square root inverse sliding window filter (SQRT-ISWF) is used for state estimation.