Rail Obstacle Detection via Route Atlas Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current obstacle recognition systems in the automotive industry, particularly in the railway sector, face challenges in accurately identifying obstacles due to the complexity of light space profiles and the lack of understanding of AI algorithms, leading to potential safety risks and uncertainties in collision avoidance.
Innovation Solution
The use of a route atlas to evaluate recognized objects, where location information is compared against pre-defined positions, reversing the obstacle recognition process to ensure only known objects are recognized, and employing machine learning algorithms to predict obstacle presence within a safety distance, initiating safety measures like emergency braking when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If AI algorithms are used for obstacle detection, then object recognition capability is improved, but reliability deteriorates due to black box nature and inability to predict detection distances
Solution Approach 1:
The patent introduces route atlases as an intermediary between the AI obstacle detection system and the safety certification process. The route atlas contains pre-registered infrastructure objects with known positions and characteristics, serving as a reference framework that mediates between the black-box AI detector and the safety requirements, enabling verifiable and predictable obstacle detection
Solution Approach 2:
The patent applies preliminary action by pre-registering all infrastructure objects (signals, switches, track elements, overhead lines, tunnels, stations) in route atlases before actual operation. This pre-established knowledge base allows the system to predictably identify obstacles by comparing detected objects against the pre-defined route atlas, rather than relying solely on black-box AI classification during critical moments
2Area of stationary object
If traditional obstacle detection methods are used, then detection coverage is improved, but manufacturing precision deteriorates due to inability to clearly determine clearance gauge freedom
Solution Approach 1:
The patent applies local quality by creating detailed, location-specific route atlases that contain precise information about infrastructure objects at specific positions along the route. Each route atlas entry includes exact position coordinates, object types, and clearance gauge specifications, enabling precise determination of whether the clearance gauge is free of obstacles at each local position rather than providing generic detection coverage
3Adaptability or versatility
If AI-based obstacle detection is deployed, then adaptability is improved for various objects, but device complexity increases due to black box AI systems requiring monitoring
Solution Approach 1:
The patent applies copying by creating digital replicas of the physical rail infrastructure in the form of route atlases. These digital copies contain all necessary information about infrastructure objects (signals, switches, track elements, overhead lines, tunnels, stations) including their positions, types, and characteristics, allowing the system to adapt to various objects through data lookup rather than complex AI classification
Data Source
Figure 1~3
Figure 4
Figure 5
AI summary
The invention relates to a method for detecting obstacles (HD1 ... HD2) on a track traveled by a vehicle (FZ), in which the vehicle (FZ) is equipped with a sensor device that detects objects (OB1 ... OB8) located in front of the vehicle (FZ) in the direction of travel (FR). The detected objects (OB1 ... OB8) are recognized by a computer and evaluated by a computer to identify obstacles (HD1 ... HD2). A track atlas containing a multitude of detectable objects (OB1 ... OB8) and their positions on or along the track is used for evaluating the objects (OB1 ... OB8). Location data is determined for the detected objects (OB1 ... OB8) and compared with the track atlas. Each detected object (OB1 ... OB8) whose location data matches the position of a corresponding detectable object (OB1 ...If the system matches OB8), it triggers an assessment that the track is clear, and this continues until a detectable object (OB1 ... OB8) is present in the track atlas within a predetermined safety distance (SCA) in front of the vehicle (FZ) for which no detectable object (OB1 ... OB8) has been assigned. Then, regardless of previously triggered assessments, the system triggers an assessment that an obstacle (HD1 ... HD2) is located on the track. Furthermore, the invention comprises an arrangement, a vehicle (FZ), a computer program product, and a delivery device for the computer program product.