Vision-Cued Random-Access LIDAR for Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle technologies face challenges in accurate localization and navigation due to the limitations of magnetometers, GPS systems, and conventional LIDAR, which are often inaccurate, EM-sensitive, bulky, or inefficient, especially in GPS-denied environments and on small platforms with high disturbance.
Innovation Solution
A vision-cued random-access LIDAR system that uses a smart-vision algorithm to identify regions of interest, classify objects, and direct LIDAR pings for continuous three-dimensional location and orientation determination, enabling efficient navigation and obstacle avoidance even at high speeds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional scanning LIDAR is used for localization and navigation, then complete environmental mapping can be achieved, but data generation becomes excessively large and processing efficiency decreases
Solution Approach 1:
The patent extracts only the essential information needed for localization by using vision-cued random-access LIDAR to target specific features identified by the vision system, rather than capturing complete environmental maps. This selective extraction reduces data volume while maintaining localization accuracy.
Solution Approach 2:
The vision system performs preliminary identification of features and regions of interest before LIDAR measurement. This preliminary action guides the LIDAR to only measure relevant features, avoiding unnecessary data collection and improving processing efficiency.
2Ease of operation
If magnetometers are used for localization, then navigation capability is provided, but accuracy is insufficient and EM sensitivity increases
Solution Approach 1:
The patent replaces electromagnetic sensing (magnetometers) with optical sensing (vision system and LIDAR). This substitution eliminates EM sensitivity issues while providing both navigation capability and improved localization accuracy through geometric feature matching.
3Measurement precision
If GPS systems are used for localization, then global positioning is achieved, but vulnerability to jamming increases and indoor performance deteriorates
Solution Approach 1:
The patent uses visual features in the environment as intermediary reference points for localization, replacing direct satellite signal dependency. This intermediary approach allows the system to achieve global positioning accuracy through feature matching while being immune to GPS jamming and functional in indoor environments.
4Length of moving object
If time-of-flight cameras are used for range measurement, then depth information is obtained, but lateral resolution becomes poor and pixel limitations occur
Solution Approach 1:
The patent merges vision system data with LIDAR range measurements. The vision system provides high-resolution lateral information while LIDAR provides accurate depth data. This combination achieves both good lateral resolution and accurate range measurement, overcoming the limitations of time-of-flight cameras.
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 solution provides accurate and efficient localization and navigation in real-time, reducing data overload and platform disturbance, while being compact and flexible, and does not require external references or surface contact, making it suitable for small platforms and GPS-denied environments.
Implementation Method 1
directing random-access LIDAR to ping one or more of the classified objects
Implementation Method 2
The LIDAR may be a coherent LIDAR, which can be employed to determine the velocities of the classified objects relative to the moving platform
Data Source
AI summary
A vision-cued random-access LIDAR system and method which determines the location and/or navigation path of a moving platform. A vision system on a moving platform identifies a region of interest. The system classifies objects within the region of interest, and directs random-access LIDAR to ping one or more of the classified objects. The platform is located in three dimensions using data from the vision system and LIDAR. The steps of classifying, directing, and locating are preferably performed continuously while the platform is moving and/or the vision system's field-of-view (FOV) is changing. Objects are preferably classified using at least one smart-vision algorithm, such as a machine-learning algorithm.