Multi-Sensor Robot Mapping With Fusion SLAM and Pose Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robots face challenges in generating accurate maps of their environment using multiple sensors, as these sensors have distinct features and errors, which affect their ability to navigate and avoid obstacles effectively.
Innovation Solution
A robot system that utilizes a combination of LiDAR and camera sensors, along with artificial intelligence, to generate heterogeneous maps and perform fusion SLAM, allowing the separation of maps for single-type sensor usage, enabling accurate position estimation and obstacle avoidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple sensors (LiDAR and camera) are used to generate maps, then measurement precision and navigation accuracy are improved, but device complexity and computational requirements increase
Solution Approach 1:
The patent combines LiDAR and camera sensors into a unified mapping system where both sensors contribute to generating a single integrated map. The LiDAR sensor provides depth and spatial information while the camera sensor provides visual features, and both are fused through a common coordinate system and mapping algorithm to create a comprehensive environmental representation that improves position estimation accuracy.
Solution Approach 2:
The patent introduces a controller as an intermediary that processes and fuses data from multiple sensors. The controller integrates LiDAR distance measurements with camera visual features, performs coordinate transformations, and generates a unified map representation. This intermediary component manages the complexity of multi-sensor integration while delivering improved measurement precision.
2Speed
If sensors are used to generate maps while the robot is moving, then real-time navigation capability is improved, but measurement errors and map accuracy deteriorate
Solution Approach 1:
The patent implements feedback mechanisms where the robot's current position and orientation are continuously estimated using odometry information from wheel encoders and IMU sensors. This feedback is used to correct and update the map as the robot moves, compensating for motion-induced errors. The system continuously refines position estimates by comparing expected vs. actual sensor measurements, maintaining map accuracy during dynamic operation.
Solution Approach 2:
The patent performs preliminary odometry calculations and pose estimations before final map updates. By pre-processing sensor data and calculating expected positions based on motion models, the system prepares correction factors that are applied during map generation. This preliminary action reduces the computational burden during real-time operation and minimizes errors introduced by rapid motion.
3Adaptability or versatility
If heterogeneous maps from different sensors are generated, then adaptability to different sensor types is improved, but ease of operation and system simplicity worsen
Solution Approach 1:
The patent creates a universal map representation that can accommodate multiple sensor types through a common data structure and coordinate system. The mapping algorithm is designed to process both LiDAR range data and camera image data using unified mathematical transformations, allowing the same system to handle heterogeneous sensors without requiring separate processing pipelines or complex configuration.
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
Enables the generation of high-quality maps and accurate position estimation for robots using multiple sensors, enhancing navigation and obstacle avoidance capabilities.
Implementation Method 1
a LiDAR sensor sensing a distance between an object outside of the robot and the robot and generating a LiDAR frame
Implementation Method 2
a camera sensor capturing an image of an object placed outside of the robot and generating a visual frame
Data Source
AI summary
Disclosed herein is a robot generating a map based on multi sensors and artificial intelligence and moving based on the map, the robot according to an embodiment including a controller generating a pose graph that includes a LiDAR branch including one or more LiDAR frames, a visual branch including one or more visual frames, and a backbone including two or more frame nodes registered with any one or more of the LiDAR frames or the visual frames, and generating orodometry information that is generated while the robot is moving between the frame nodes.