Train Driver Visual Performance: A Parametric Survival Model of Reaction Time

Cogan, Baris; Milius, Birgit · 2024 · Crossref

DOI: 10.54941/ahfe1005255

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 investigates the visual performance of train drivers, specifically focusing on reaction times to visual stimuli under varying driving conditions. The research is motivated by the increasing implementation of Automatic Train Operation (ATO) systems in the railway sector. As automation levels rise, defining the functional requirements for these systems requires a realistic assessment of human performance capabilities. Understanding how train drivers perceive and react to trackside hazards is crucial for ensuring safety and determining the practical limits of automated systems. The study aims to quantify this perceptual performance using simulator experiments and parametric survival analysis. The experimental design utilized a driving simulator at the Technical University of Berlin, involving 18 active train drivers with an average age of 33.4 years and 7.3 years of professional experience. Participants were tasked with detecting stationary cube-shaped stimuli appearing approximately 800 meters from the train and responding by sounding the horn. The study manipulated several independent variables: object size (180 cm vs. 90 cm), contrast (high orange vs. low brown), train speed (40, 100, and 160 km/h), and train protection systems (PZB, ETCS Level 2, and on-sight driving). Reaction time was defined as the duration from stimulus appearance to detection. Data analysis employed a Weibull Accelerated Failure Time (AFT) parametric survival model, selected for its superior fit compared to Normal and Log-normal distributions, to examine the relationship between covariates and reaction times. The results indicate that object size, contrast, and train speed significantly influenced driver reaction times. Larger objects and higher contrast were associated with shorter reaction times. Specifically, reducing the stimulus size from 180 cm to 90 cm increased reaction time by 49%, while lower contrast increased it by 22%. Counterintuitively, higher speeds led to faster detection; reaction times at 40 km/h were 43% longer than at 100 km/h, and increasing speed to 160 km/h further reduced reaction times by 31%. The study found no statistically significant difference in reaction times attributable to the type of train protection system (PZB, ETCS, or on-sight), suggesting that observed variations were likely due to speed differences rather than the control systems themselves. Survival probability curves confirmed that the likelihood of failing to detect small objects was significantly higher than for large objects. The study concludes that visual perception performance in train driving is heavily dependent on stimulus characteristics and operational speed. These findings provide valuable data for deriving functional requirements for ATO systems, offering a baseline for human performance against which automated systems can be evaluated. The research also demonstrates the utility of parametric survival analysis in the railway domain for modeling reaction time data. However, the authors note limitations, including the simulated environment's differences from real-world conditions and the lack of consideration for cognitive factors like expectation or risk assessment. Future research is recommended to investigate the complex relationship between train protection systems and visual attention, as well as the impact of additional driving tasks on perception.

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-nvidia on 2026-08-10; verification: verified.

Topics

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

• rail grade crossings
• perception reaction time
• hazard perception

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data
• Theoretical Contribution: computational model