Using N-Version Architectures for Railway Segmentation with Deep Neural Networks

Autonomous trains require reliable and accurate environmental perception to take over safety-critical tasks from the driver. This paper investigates the application of N-version architectures to rail track detection using Deep Neural Networks (DNNs) as a means to improve the safety of machine learni...

Full description

Saved in:
Bibliographic Details
Main Authors: Philipp Jaß, Carsten Thomas
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Machine Learning and Knowledge Extraction
Subjects:
Online Access:https://www.mdpi.com/2504-4990/7/2/49
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849431857995710464
author Philipp Jaß
Carsten Thomas
author_facet Philipp Jaß
Carsten Thomas
author_sort Philipp Jaß
collection DOAJ
description Autonomous trains require reliable and accurate environmental perception to take over safety-critical tasks from the driver. This paper investigates the application of N-version architectures to rail track detection using Deep Neural Networks (DNNs) as a means to improve the safety of machine learning (ML)-enabled perception systems. We combine three different neural network architectures (WCID, VGG16-UNet, MobileNet–SegNet) in a 3M1I configuration. In this configuration, we apply two fusion methods to increase accuracy and to enable error detection: Maximum Confidence Voting (MCV), combining the DNN predictions at the image level, and Pixel Majority Voting (PMV), a novel approach for combining the predictions at the pixel level. In addition, we implement a new method for evaluating and combining prediction confidence values in the N-version architecture during runtime. We adjust the overall prediction confidence according to the conformity of all individual predictions, which is not possible with an individual network. Our results show that the N-version architecture not only enables a detection of erroneous predictions by utilizing those adjusted confidence values, but it can also partially improve the predictions by using the PMV combination algorithm. This work emphasizes the importance of model diversity and appropriate thresholds for an accurate assessment of prediction safety. These approaches can significantly improve the practical applicability of ML-based systems in safety-critical domains such as rail transportation.
format Article
id doaj-art-c64e5705b4364d12b98f4bb67940ff0d
institution Kabale University
issn 2504-4990
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Machine Learning and Knowledge Extraction
spelling doaj-art-c64e5705b4364d12b98f4bb67940ff0d2025-08-20T03:27:30ZengMDPI AGMachine Learning and Knowledge Extraction2504-49902025-05-01724910.3390/make7020049Using N-Version Architectures for Railway Segmentation with Deep Neural NetworksPhilipp Jaß0Carsten Thomas1HTW Berlin, University of Applied Sciences, Wilhelminenhofstraße 75A, 12459 Berlin, GermanyHTW Berlin, University of Applied Sciences, Wilhelminenhofstraße 75A, 12459 Berlin, GermanyAutonomous trains require reliable and accurate environmental perception to take over safety-critical tasks from the driver. This paper investigates the application of N-version architectures to rail track detection using Deep Neural Networks (DNNs) as a means to improve the safety of machine learning (ML)-enabled perception systems. We combine three different neural network architectures (WCID, VGG16-UNet, MobileNet–SegNet) in a 3M1I configuration. In this configuration, we apply two fusion methods to increase accuracy and to enable error detection: Maximum Confidence Voting (MCV), combining the DNN predictions at the image level, and Pixel Majority Voting (PMV), a novel approach for combining the predictions at the pixel level. In addition, we implement a new method for evaluating and combining prediction confidence values in the N-version architecture during runtime. We adjust the overall prediction confidence according to the conformity of all individual predictions, which is not possible with an individual network. Our results show that the N-version architecture not only enables a detection of erroneous predictions by utilizing those adjusted confidence values, but it can also partially improve the predictions by using the PMV combination algorithm. This work emphasizes the importance of model diversity and appropriate thresholds for an accurate assessment of prediction safety. These approaches can significantly improve the practical applicability of ML-based systems in safety-critical domains such as rail transportation.https://www.mdpi.com/2504-4990/7/2/49machine learningN-versionsafetyautonomous rail systemssemantic segmentationsafety-critical AI
spellingShingle Philipp Jaß
Carsten Thomas
Using N-Version Architectures for Railway Segmentation with Deep Neural Networks
Machine Learning and Knowledge Extraction
machine learning
N-version
safety
autonomous rail systems
semantic segmentation
safety-critical AI
title Using N-Version Architectures for Railway Segmentation with Deep Neural Networks
title_full Using N-Version Architectures for Railway Segmentation with Deep Neural Networks
title_fullStr Using N-Version Architectures for Railway Segmentation with Deep Neural Networks
title_full_unstemmed Using N-Version Architectures for Railway Segmentation with Deep Neural Networks
title_short Using N-Version Architectures for Railway Segmentation with Deep Neural Networks
title_sort using n version architectures for railway segmentation with deep neural networks
topic machine learning
N-version
safety
autonomous rail systems
semantic segmentation
safety-critical AI
url https://www.mdpi.com/2504-4990/7/2/49
work_keys_str_mv AT philippjaß usingnversionarchitecturesforrailwaysegmentationwithdeepneuralnetworks
AT carstenthomas usingnversionarchitecturesforrailwaysegmentationwithdeepneuralnetworks