Multi-sensor Fusion SLAM Voxel Subgraph Pose Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SLAM systems face challenges in maintaining robustness and accuracy in diverse and dynamic environments, particularly in environments with repeated structures such as tunnels or corridors, and in long continuous operations.
Innovation Solution
A multi-sensor fusion SLAM system that integrates a visual inertia module, laser scanning matching module, and loop closure detection module to enhance robustness and accuracy by using voxel subgraphs for pose calculation and error correction, employing algorithms like ICP and PnP for feature matching, and utilizing a deep neural network for loop closure detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pure lidar-based SLAM is used, then robustness to environmental changes is improved, but accuracy in environments with repeated structures deteriorates
Solution Approach 1:
The patent combines multiple sensor types (lidar, visual sensors, IMU) into a unified SLAM system. The visual-inertial-lidar fusion integrates features from different sensors to compensate for individual sensor limitations, particularly using visual features to disambiguate repeated structures while maintaining lidar's robustness to environmental changes.
Solution Approach 2:
The system creates a composite sensing approach by fusing data from heterogeneous sensors (lidar point clouds, image features, IMU measurements). This composite approach leverages the complementary strengths of each sensor type to achieve both robustness and precision in challenging environments.
2Reliability
If more sensors are integrated for multi-sensor fusion, then robustness and accuracy are improved, but device complexity increases
Solution Approach 1:
The patent implements a unified optimization framework that handles multiple sensor types and multiple error sources (intrinsic parameters, extrinsic parameters, sensor biases) within a single mathematical model. This multi-functional approach consolidates what would otherwise require separate processing pipelines, reducing overall system complexity while maintaining the benefits of multi-sensor fusion.
3Measurement precision
If comprehensive sensor fusion is performed, then positioning accuracy is improved, but calculation amount increases
Solution Approach 1:
The patent extracts and optimizes only the essential parameters needed for accurate positioning and mapping, separating them from redundant information. The voxel subgraph construction extracts key spatial relationships from complete point clouds, reducing the data volume requiring intensive optimization while preserving positioning accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A multi-sensor fusion SLAM system (100) and a robot. The system (100) operates on a mobile robot and comprises: a visual inertia module (10), a laser scanning matching module (20), a closed-loop detection module (30), and a visual laser image optimization module (40). According to the multi-sensor fusion SLAM system (100) and the robot, the calculation amount of laser matching constraint optimization can be reduced by using a voxel subgraph so that the pose calculation is more accurate, accumulated errors of long-time operation of the system can be corrected in time by means of sufficient fusion of modules, and the robustness of the system and the accuracy of positioning and mapping are integrally improved.