LIDAR Plane Uncertainty Estimation for SLAM Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In GPS-denied environments, navigation systems that do not rely on Inertial Measurement Units (IMUs) face challenges in accurately estimating the uncertainty of features extracted from 3D imaging data, which affects the performance of simultaneous localization and mapping (SLAM) algorithms, leading to suboptimal navigation solutions.
Innovation Solution
A method and apparatus that generate three-dimensional imaging data using a LIDAR sensor, extract planes from this data, estimate the uncertainty of these planes, and use this information to produce navigation solutions, employing techniques such as calculating the dimensions of an envelope box and sample covariance to refine uncertainty estimates for the SLAM algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a 3D LIDAR sensor is used to extract features for navigation in GPS-denied environments, then the system does not need to rely on IMU, but the uncertainty estimation of extracted features becomes inaccurate
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing the covariance matrix of the extracted plane features before they are used in SLAM. The system performs eigenvalue decomposition on the covariance matrix in advance to obtain eigenvalues and eigenvectors, which are then used to compute the uncertainty estimate. This preliminary processing ensures that when the features are used for navigation, the uncertainty estimation is already prepared and accurate, resolving the contradiction between using LIDAR features and maintaining precise uncertainty estimation.
2Reliability
If standard SLAM methods are used with extracted features, then navigation can be performed without GPS, but the Kalman filter loses optimality when uncertainty is not properly estimated
Solution Approach 1:
The patent implements feedback by using the computed uncertainty estimate (based on eigenvalues of the covariance matrix) to adjust the Kalman filter's measurement update process. The uncertainty information feeds back into the SLAM algorithm, allowing the Kalman filter to properly weight the LIDAR feature measurements. This feedback mechanism ensures that the Kalman filter maintains its optimality property by giving appropriate gain to measurements based on their actual uncertainty, thus resolving the contradiction between navigation reliability and measurement precision.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the accuracy of navigation solutions by properly estimating feature uncertainty, maintaining the optimality of the SLAM algorithm and improving the quality of location measurements in GPS-denied environments.
Implementation Method 1
A 3D LIDAR produces a 3D range image of the environment
Implementation Method 2
A 3D LIDAR produces a 3D range image of the environment
Data Source
AI summary
In one embodiment, a method comprises generating three-dimensional (3D) imaging data for an environment using an imaging sensor, extracting an extracted plane from the 3D imaging data, and estimating an uncertainty of an attribute associated with the extracted plan. The method further comprises generating a navigation solution using the attribute associated with the extracted plane and the estimate of the uncertainty of the attribute associated with the extracted plane.