Obstacle Recognition Along the Railway Tracks - An Investigation of Parameters of Reaction Speed in German Train Drivers

Kurzfassung

A challenge for automation on open-track railway systems, especially those operating long distance, is a lack of safety standards for obstacle detection, including benchmarks for performance of automated systems. This thesis aims to measure how train drivers perform when they encounter obstacles on the tracks to provide such a benchmark. A simulator experiment with professional train drivers and a larger-sample online study with non-professionals were conducted. In the simulator experiment, participants performed a driving task and reacted to obstacles on the tracks. In the online study, participants reacted to obstacle appearances in videos. Size and contrast of obstacles, as well as driving speed and the use of train control systems, were varied in a within-subjects design, and their impact on reaction time (RT) was analyzed with a linear regression model on log-transformed data. Size and contrast were expected to influence stimulus salience, leading to increased RT for lower contrast or smaller stimuli. Higher driving speed was expected to decrease RT due to gaze verticalization. Train control system variation was expected to lead to decreased RT in conditions with lower attentional demands at the driver’s desk. The simulator experiment showed significant differences in RT for all tested factors in the expected directions, except the use of specific train control systems (ETCS/PZB). Predicted RT differences were especially large for variations in size, speed, and use of train control systems (yes/no). In the online study, all tested variables led to significant differences in RT. The findings of this research show the influence of stimulus properties and operational parameters on train driver performance in an obstacle recognition task and provide a baseline for safety standard definition for future automation.