Driving Robot Depth Mapping for Multi-Height Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robots face challenges in generating accurate driving maps due to limitations of 2D LiDAR sensors in environments with black materials and the computational demands of 3D SLAM, leading to inaccurate position recognition and localization.
Innovation Solution
A driving robot equipped with a depth camera and processor that identifies scan data sets at multiple height levels, assigns feature scores, and generates main and sub-area maps based on these scores, downscaling sub-area maps to enhance map generation and localization accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If 2D LiDAR sensor is used for scanning, then the scanning process is simple, but the scanning accuracy deteriorates in environments with black materials absorbing light
Solution Approach 1:
The patent introduces a depth camera as an intermediary device to capture depth information in environments where 2D LiDAR fails. The depth camera serves as a mediator that can penetrate through challenges posed by black materials, providing complementary depth data that compensates for LiDAR's limitations in such environments.
Solution Approach 2:
The patent combines data from two different sensing modalities (2D LiDAR and depth camera) to create a composite mapping system. By fusing the simple scanning capability of LiDAR with the environmentally robust depth sensing of the camera, the system achieves both ease of manufacture and high measurement precision across diverse environments.
2Measurement precision
If 3D SLAM using all depth data is implemented, then the localization accuracy is improved, but the computing resource requirements increase
Solution Approach 1:
The patent extracts and processes depth data at multiple specific height levels rather than processing all depth data continuously. By selectively extracting scan data at predetermined height levels and generating separate area maps for each level, the system reduces the volume of data requiring intensive 3D SLAM computation while maintaining localization accuracy.
Solution Approach 2:
The patent segments the depth data processing task by dividing it into multiple independent height level processing streams. Each height level generates its own area map, allowing parallel processing and reducing the computational burden on any single processing unit. This segmentation enables accurate localization without requiring excessive computing resources.
3Productivity
If depth data at a specific height is used like 2D LiDAR, then the computational load is reduced, but the position recognition accuracy deteriorates when no feature data exists at that height
Solution Approach 1:
The patent transitions from processing depth data at a single height level to processing data across multiple height levels simultaneously. By adding the vertical dimension as an additional processing dimension, the system maintains computational efficiency through structured multi-level processing while ensuring position recognition accuracy by having multiple height levels as fallback options when features are absent at any single level.
Data Source
AI summary
A driving robot includes: a camera including a depth camera; and at least one processor configured to: control the camera to acquire depth data in one or more areas where the driving robot moves, identify, from the acquired depth data, a plurality of scan data sets corresponding to a plurality of predetermined height levels, identify, based on the plurality of scan data sets, a plurality of feature scores corresponding to the plurality of scan data sets, and generate at least one area map corresponding to at least one scan data set among the plurality of scan data sets, wherein a feature score, among the plurality of feature scores, corresponding to the at least one scan data set is greater than or equal to a predetermined critical value.