Rail Path Extraction Using 2D-3D Sensor Fusion Integrity Checks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing path extraction methods in the rail industry are either expensive and rely on survey-grade data gathering or lack accuracy and integrity in online onboard data processing, particularly when train position is unknown or uncertain, and they fail to adequately combine multiple sensors for improved performance and integrity.
Innovation Solution
A system using multiple onboard autonomy sensors, including Passive 2D and Active 3D sensors, combines data from LiDAR, cameras, and IMUs to extract paths with high integrity and accuracy, employing fusion and diversity checks to ensure robustness and completeness, even in conditions where train position is unknown.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If offline survey-grade data processing is used, then path extraction accuracy is improved, but cost and complexity increase significantly
Solution Approach 1:
The patent replaces expensive, complex surveying devices with inexpensive onboard autonomy sensors (cameras, LiDAR, IMUs) that are already present on modern trains. These sensors provide sufficient accuracy for path extraction without requiring specialized surveying equipment, manual operations, or extensive post-processing.
Solution Approach 2:
The patent substitutes mechanical surveying devices with electronic sensor systems. Instead of using Total Station, EM-SAT, or high-density LiDAR surveys requiring manual operations, the system uses onboard cameras, LiDAR, and IMUs with automated processing algorithms to extract paths in real-time.
2Measurement precision
If offline survey methods are used, then path accuracy is improved, but real-time applicability deteriorates when train position is unknown
Solution Approach 1:
The system performs path extraction online and in real-time using onboard sensors, rather than relying on pre-collected survey data. This allows the train to extract its own path information dynamically during operation, making the system adaptable to unknown or uncertain train positions without requiring prior surveying.
Solution Approach 2:
The train uses its own onboard autonomy sensors to extract path information independently, without relying on external surveying teams or pre-collected data. The system is self-sufficient, using cameras, LiDAR, and IMUs already present on the train to determine its position and path in real-time.
3Device complexity
If single sensor online extraction is used, then cost is reduced, but integrity and reliability deteriorate
Solution Approach 1:
The patent combines multiple onboard autonomy sensors (cameras, LiDAR, IMUs) into a unified path extraction system. By fusing data from these different sensor types, the system achieves higher reliability and integrity than single-sensor approaches, while still using inexpensive commercial-off-the-shelf components rather than specialized surveying equipment.
Solution Approach 2:
The system creates a composite sensing approach by integrating data from heterogeneous sensor types (optical cameras, active LiDAR, inertial IMUs). This multi-sensor fusion provides redundancy and cross-validation, improving the integrity and reliability of path extraction while maintaining cost-effectiveness.
4Speed
If online onboard data processing is used, then real-time capability is improved, but accuracy and integrity worsen without adequate sensor combination
Solution Approach 1:
The system uses feedback from multiple sensor sources to continuously refine and validate path extraction results in real-time. By cross-checking data from cameras, LiDAR, and IMUs, the system maintains high accuracy during online processing, using inter-sensor consistency checks to ensure reliability without sacrificing real-time performance.
Data Source
AI summary
Path extraction for a vehicle on a guideway, the vehicle having two or more sensors. Two or more sensor inputs are received from two or more sensors including at least one active 3D sensor input from at least one active 3D sensor and at least one passive 2D sensor input from at least one passive 2D sensor. At least one active 3D sensor path is extracted based on the at least one active 3D sensor input and at least one passive 2D sensor path based on the at least one passive 2D sensor input. At least one 3D sensor ground surface model is generated based on the at least one passive 2D sensor path. At least one passive 3D path is generated based on the at least one passive 2D sensor path and the at least one 3D sensor ground surface model. The at least one passive 3D path and the at least one active 3D sensor path are fused to produce a consolidated 3D path. In a path extraction pipeline, at least one supervision check is performed for providing integrity to the consolidated 3D path.