TLDR: This study explores how railway professionals perceive safety and automation, challenging common assumptions. It highlights that railway safety is a complex “system of systems” rather than just the train itself, and that direct transfer of automotive autonomy is not feasible. Professionals prefer assistive technologies and believe a railway-specific, blame-free causation model is essential for future safety metrics and automation development.
A recent study delves into the intricate world of railway safety and the future of autonomous train operations, offering crucial insights directly from the professionals who live and breathe the rail environment. The research, titled “Insights from Railway Professionals: Rethinking Railway assumptions regarding safety and autonomy”, was conducted by Josh Hunter, John McDermid, and Simon Burton from the University of York’s Centre for Assuring Autonomy.
The study aimed to understand how railway professionals, including drivers, route planners, and administrative personnel, perceive ‘safety’ within the rail context. This understanding is vital for guiding future technological advancements in the industry, especially concerning automation.
Challenging Common Assumptions
The researchers identified and challenged three key assumptions often found in contemporary discussions about railway automation:
First, the assumption that autonomy is easily transferable between cars and railways. While automotive technology has inspired some rail innovations, professionals universally highlighted fundamental differences. Operating a train requires extensive, intimate knowledge of routes, signals, and environmental factors, often taking months of training for a single new route. While there’s little desire for full driver removal, there’s a positive sentiment towards driver-assistance technologies, such as head-up displays, that aid in parsing signs and signals.
Second, the idea that the railway is inherently ready for full autonomy. Despite the intuitive thought that trains, with their limited operating area and predictable traffic, would be easier to automate than cars, the study reveals a far more complex reality. The London Docklands Light Rail (DLR) is often cited as an example of full automation, but even Transport for London (TfL) does not classify it as a high grade of automation, and has no current interest in improving the Grades of Automation within the London underground ecosystem. True full automation of an entire railway ecosystem has not been formally attempted outside of very closed, single-track systems.
Third, the belief that the railway ecosystem can properly quantify safety. Safety metrics are notoriously difficult to generate in rail due to the low incidence of severe accidents. Current approaches often define safety retrospectively, based on past events. However, administrative professionals, in particular, suggested that safety is not a fixed, achievable goal but rather a continuous endeavor influenced by external factors like weather, traffic, and infrastructure age. They also highlighted the often-overlooked risk of closing a rail line, which can push people towards less safe alternatives like cars.
The Railway as a Complex Ecosystem
A significant finding was the broad understanding of the ‘railway ecosystem.’ While drivers initially focused on the immediate cab environment (driver, signal officer, guard), discussions with other professionals revealed a much wider scope. This ecosystem includes environmental agencies, meteorologists, track designers, and personnel involved in route planning, rolling stock development, and driver training. The Stonehaven derailment, caused by issues with old earthworks and drainage, served as a stark example of how external environmental factors and infrastructure maintenance are critical components of overall railway safety, far beyond the immediate control of the train crew.
Defining Safety: A Multifaceted View
When asked to define safety, different professional groups offered varied perspectives. Drivers often saw safety as almost predetermined – the ability to get from point A to point B without incident, trusting the existing system. Developers viewed safety as adherence to rules, with improvements often stemming from retrospective analysis of incidents. Administrative workers, however, presented a more nuanced view, acknowledging the influence of daily external factors and the ongoing responsibility to maintain operations while managing risk. This highlights that safety is not a static concept but a dynamic, evolving ideal.
Beyond Blame: A New Causation Model
The study also underscored the railway industry’s unique approach to incident investigation, particularly in the UK, where bodies like the Rail Accident Investigation Branch (RAIB) focus on preventing future accidents rather than assigning blame. This contrasts with traditional accident causation models, like the Swiss Cheese model, which can lead to identifying ‘human error’ as a root cause. The research suggests that these models are not well-suited for rail, advocating for a new, railway-specific causation model that investigates incidents without allocating blame, aligning with the industry’s collaborative safety philosophy.
Also Read:
- A New Framework for Scaling Automated Driving Across Global Differences
- HySAFE-AI: A New Framework for Ensuring Safety in AI-Powered Autonomous Systems
Looking Ahead: The SACRED Methodology
This research contributes to the development of the SACRED methodology, a seven-step approach designed to generate safety metrics for automated railway systems. SACRED proposes an Operational Domain Model (ODM) approach, which lists potential hazards within a given area, rather than the more common Operational Design Domain (ODD) that focuses on conditions under which a system is designed to function. This system-agnostic approach is crucial for developing effective safety metrics for emerging technologies like machine learning in rail.
In conclusion, the study emphasizes a disconnect between current technological advancements, practical applications, and the perspectives of railway professionals. It highlights the critical need for greater cohesion across disciplines and a deeper understanding of the railway as a complex ‘system of systems.’ Developing widely applicable railway safety metrics requires a new perspective, one that acknowledges the intricate interactions of all external stakeholders and moves beyond a vehicle-centric view of autonomy.
