Calculation of Driving Parameters for GOA4 Signaling System using Machine Learning Methods

Akçay, Mehmet Taciddin; Akgundogdu, Abdurrahim · 2022 · Crossref

DOI: 10.35377/saucis...932969

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the optimization of driving profiles for driverless rail systems operating at Grade of Automation 4 (GOA4), where vehicles run without human drivers. The authors argue that accurate estimation of vehicle acceleration and braking parameters is critical for signaling system efficiency and operational performance. As rail systems increasingly rely on automated software and big data technologies, machine learning offers a method to replace manual factors with precise, algorithmic control. The research aims to estimate vehicle acceleration, braking acceleration, and travel time using machine learning methods to support the design of automatic driverless systems and improve reliability according to EN 50126–50129 and IEC 62290 standards. The methodology involved collecting 300 data arrays from operational metro systems, with inputs consisting of the distance between stations (1–2 km) and maximum operating speed (60–90 km/h). The target outputs were vehicle acceleration and deceleration values (ranging from 1 to 1.5 m/s²) and travel speed. The researchers applied four supervised machine learning models: Linear Regression, Random Forest, Multi-Layer Perceptron (MLP), and K-Neighbors Regressor (kNN). The models were trained and tested using a 10-group cross-validation method. Performance was evaluated using Root Mean Square Error (RMSE) and the coefficient of determination (R²), with regression curves analyzed to assess prediction accuracy against actual operational data. The results indicate that the Multi-Layer Perceptron (MLP) model achieved the highest performance among the tested methods, yielding an RMSE of 7.69 and an R² value of 0.56. Linear Regression followed as the second-best performer with an RMSE of 8.04 and R² of 0.52. The kNN model produced an RMSE of 8.88 and R² of 0.42, while the Random Forest model performed the worst, with an RMSE of 9.79 and R² of 0.30. Regression distribution graphs confirmed that the MLP results were closest to the ideal 45-degree slope line, indicating superior predictive capability for the driving parameters under the tested conditions. The significance of this work lies in demonstrating that machine learning, particularly neural networks like MLP, can effectively estimate critical driving parameters for GOA4 signaling systems. By integrating these algorithms into automatic control systems, operators can enhance operational performance, minimize malfunctions, and ensure compliance with high safety integrity levels (SIL 4). The study concludes that while current results are usable, future research should explore diverse data types and additional artificial intelligence algorithms to further refine the estimation of signaling system parameters and optimize rail traffic management.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction