Laser Scanner Ego-Motion Estimation with Visual-Inertial SLAM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mapping systems for autonomous devices face challenges in accurately generating robust maps in real-time while in motion due to issues such as sensor distortions, reliance on GPS in GPS-denied environments, and inefficiencies in merging data from various sensors like IMU, cameras, and laser scanners.
Innovation Solution
A modularized mapping system that integrates an IMU, camera, and laser scanner with a computing system to process data sequentially from coarse to fine resolutions, incorporating fault tolerance and bidirectional feedback to correct biases, enabling real-time, high-accuracy map generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a laser scanner is used for motion estimation in a mobile mapping system, then depth information and scale can be obtained, but scan distortions occur due to the relative movement of the scanner during scanning
Solution Approach 1:
The system performs preliminary motion compensation by predicting scanner movement during the scanning process and pre-correcting the laser point positions before map generation, thereby preventing scan distortions from occurring in the final output
Solution Approach 2:
The system uses feedback from motion sensors (IMU, odometry) to continuously monitor scanner position and orientation changes, and applies real-time corrections to laser scan data to maintain accuracy despite scanner movement
2Loss of information
If GPS is used for location information, then external location data can be obtained, but the system becomes dependent on GPS signals which may not be available in GPS-denied environments
Solution Approach 1:
The system extracts location and motion information from onboard sensors (IMU, wheel encoders, visual odometry) rather than relying on external GPS signals, making the system self-sufficient and operational in GPS-denied environments
Solution Approach 2:
The system introduces intermediate localization methods such as visual odometry and inertial navigation as mediators between the absence of GPS and the need for location information, enabling continuous operation across different environmental conditions
3Quantity of substance
If multiple sensors (IMU, camera, laser scanner) are integrated for mapping, then comprehensive data can be collected, but data merging becomes complex and computationally intensive
Solution Approach 1:
The system segments the data processing pipeline into distinct modules: IMU preprocessing, visual odometry computation, laser scan registration, and map building. Each module handles specific sensor data independently before integration, reducing overall processing complexity
Solution Approach 2:
The system merges sensor data at strategically chosen points in the processing pipeline, combining IMU predictions with visual odometry results, then integrating laser scan data with the emerging map structure, rather than attempting to process all sensors simultaneously
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
The system achieves real-time, high-accuracy map generation capable of handling sensor degradation and dynamic environments, providing accurate navigation and obstacle avoidance without dependence on external location technologies.
Implementation Method 1
a 3D laser scanner... to provide depth information associated with individual pixels
Implementation Method 2
an inertial measurement unit... to provide inertial motion data
Implementation Method 3
a camera... to provide visual measurement data
Data Source
AI summary
A system configured to derive a motion estimate for a SLAM device using an IMU forming a part of the SLAM system. The system may be configured to refine the motion estimate via a visual-inertial odometry optimization process to produce a refined estimate and refine the refined estimate via a laser odometry optimization process by minimizing at least one residual squared error between at least one feature in a current scan and at least one previously scanned feature.