2D LiDAR-Camera SLAM for Indoor Localization in Similar Spaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving SLAM systems face challenges in accurately localizing and mapping indoor environments due to the sensitivity of vision and LiDAR sensors to illuminance and limited resolution, which can lead to incorrect map creation and potential collisions.
Innovation Solution
A method and system that bind observation feature points from a camera to 2D LiDAR points and compare them with a map to accurately localize the mobility and create a map, using a controller to load maps, acquire LiDAR and camera data, reduce dimensionality, and detect absolute positions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vision or LiDAR sensors are used to construct space information, then the mobility can navigate the environment, but the sensor sensitivity to illuminance and limited resolution cause incorrect map creation
Solution Approach 1:
The patent combines multiple sensors (stereo vision, LiDAR, and position detection sensors) to create a fused perception system. The stereo vision provides depth information, LiDAR provides precise distance measurements, and position detection sensors provide movement data. By merging these sensor inputs, the system compensates for individual sensor weaknesses regarding illuminance sensitivity and resolution limitations, thereby improving map creation accuracy while reducing reliance on any single sensor's performance under varying lighting conditions
Solution Approach 2:
The patent introduces an intermediary processing layer that fuses sensor data before map creation. The controller integrates information from stereo vision, LiDAR, and position detection sensors through data fusion algorithms. This intermediary processing step filters out noise and compensates for individual sensor deficiencies, particularly mitigating the harmful effects of illuminance sensitivity on vision sensors and resolution limitations on LiDAR, thereby producing more accurate and reliable maps
2Measurement precision
If LiDAR detects thin legs of furniture, then the mobility can perceive obstacles, but the thin legs are treated as noise and deleted during mapping process
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously compares detected features with the generated map and adjusts the mapping process accordingly. When thin structures like furniture legs are detected by LiDAR, the feedback loop prevents their deletion by validating their consistency across multiple sensor inputs and temporal frames. The system uses position detection sensor data to track movement relative to these thin features, providing feedback that confirms their significance as topographical elements rather than noise, thereby preserving them in the final map
Solution Approach 2:
The patent creates a composite representation of the environment by fusing data from multiple sensors with different characteristics. The stereo vision system provides visual context that helps identify thin structures as meaningful features, while LiDAR provides precise geometric data. This composite sensing approach creates a redundant representation where thin legs are confirmed by multiple independent measurements, preventing their misclassification as noise during the mapping process and preserving important topographical information
3Adaptability or versatility
If SLAM is performed in indoor environment with similar space structures, then the mobility can navigate, but it is difficult to perform SLAM due to lack of distinctive topographical features
Solution Approach 1:
The patent employs a multi-functional sensor system that performs multiple functions simultaneously. The stereo vision system not only provides depth information for obstacle detection but also captures visual features for place recognition. The LiDAR system simultaneously measures distances for navigation and creates geometric maps for localization. The position detection sensors provide movement data for both odometry and map building. This multi-functionality allows the system to navigate similar space structures while maintaining localization accuracy by cross-validating information across multiple functional streams
Solution Approach 2:
The patent transitions from relying solely on 2D visual features to incorporating 3D spatial information from LiDAR and stereo vision. By adding the depth dimension, the system can distinguish between similar-looking spaces through subtle geometric differences that are invisible in 2D images alone. The fused 3D point cloud data from LiDAR and stereo vision provides additional spatial dimensions for comparison, enabling accurate localization in environments with repetitive structures by detecting minute variations in three-dimensional geometry that would be imperceptible in two-dimensional visual data
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 enables accurate localization and map creation in indoor environments with complex structures, reducing the risk of collisions and improving the reliability of SLAM systems without requiring advanced computing capabilities.
Implementation Method 1
acquiring 2D LiDAR point data from a LiDAR
Implementation Method 2
acquire a front image of the mobility from a camera
Data Source
AI summary
A system and method of simultaneous localization and mapping can include loading a map to which 2D feature point data stored in a memory is bound, acquiring 2D LiDAR point data from a LiDAR equipped in a mobility, acquiring 3D observation feature point data by receiving a front image from a camera equipped on the mobility, reducing a dimension of the 3D observation feature point data to 2D observation feature point data, binding the 2D observation feature point data to the 2D LiDAR point data, and detecting an absolute position of the mobility based on the 2D observation feature point data bound to the 2D LiDAR point data and the map to which the 2D feature point data is bound.