Automation in Railway Operations: Effects on Signaller and Train Driver Workload

Abstract

Throughout the railway domain, increasing levels of automation are employed to ensure safety and increase efficiency on the tracks. This impacts the task characteristics of the signaller and train driver. In a scenario of German railway automation, an automated interlocking system routes trains automatically and trains equipped with “ATO over ETCS” (“Automatic Train Operation over the European Train Control System”) automatically drive along the predefined routes adhering to speed restrictions. Thus, the task load of signallers and train drivers decreases, as manual inputs decline in favour of monitoring the functioning of automatic systems. Yet, it remains unclear whether decreasing task load directly lowers subjective workload. Additionally, the question of optimal workload levels has yet to be addressed. In order to ensure optimal performance, under- and overload are to be avoided. In previous studies on workload in railway operations, workload was assessed without considering an individual optimum of workload. In the simulator studies described in the paper, subjective workload was assessed with the goal of analysing the impact of automation on achieving an optimal workload level. In two separate studies with train drivers and signallers, subjective workload was assessed by two self-report measures (NASA-TLX and DLR-WAT) after participants completed periods of manual driving and driving with automated systems. Results consistently indicate lower subjective workload in the automated work settings for both signallers and train drivers. Interestingly, signaller workload was close to optimal subjective levels while train driver workload scores were considerably lower than optimal. This highlights the need for thoughtful introduction of automation into the train driver environment. Furthermore, the DLR-WAT differentiates between mental workload caused by the different stages of information processing. In train drivers and signallers, workload stemming from information perception seems to be more pronounced than workload stemming from mental operations occurring at later stages of information processing, especially in automatic work settings. Assessing workload relative to an individual optimum and differentiating causes of mental workload along the different stages of information processing offers unique insights into signaller and train driver workload. The results make it possible to ascertain, which specific aspects of the introduction of automation in the signaller and train driver tasks lead to lowered overall workload.