Improving Station Dwell Times

27/01/2021

RSSB funded research projects are using data to understand, predict and prevent variations in the time a train spends at the platform.

A train stopping for longer than timetabled at a station is a common cause of service delay. Most of these delays are less than three minutes. They are ‘sub-threshold’, meaning they do not go through the delay attribution process. But even delays of tens of seconds can, when aggregated, adversely impact the overall performance and capacity of the network.

Measures introduced because of Covid-19, such as social distancing and more frequent timetable changes, mean that the industry is facing new operational challenges. A better understanding of the root causes of dwell delay and the steps that can be taken to mitigate its impact will play an important role in ensuring punctual and reliable services for customers.

Text 5

RateSetter+ has combined outputs from different projects funded through the Data Sandbox+ and ‘Faster, safer and better boarding and alighting’ competitions.

Project underway: RateSetter+ and the impact of social distancing

RateSetter+ aims to generate insights into how Covid-19 related social distancing policies might impact dwell times and affect overall service performance. It builds on the RateSetter+ Covid-19 application, which examined how social distancing influences passenger behaviour at the platform-train interface.

RateSetter+ is looking at overall service performance risk, identifying pinch points where performance issues cascade across the network, and services that might cause higher levels of disruption if their dwell times are extended. This information should help passenger operators design interventions to alleviate the effects of Covid-19 social distancing on dwell times. Mitigations might include focusing effort on specific locations or trains, or adjusting timetables to avoid congestion caused by extended dwell times.

Read more about the RateSetter+ in our research catalogue. The final report is due to be published in Spring 2021.

Text 6

The Data Sandbox+ competition, launched in 2019, for research into innovative data-driven solutions to key operational performance challenges, also funded two projects focussed on managing dwell time.

Completed project: IntelliDwellTime
The IntelliDwellTime (IDT) demonstrator project developed a software solution that generates and analyses dwell data to a high level of granularity. The tool displays dwell time at every station stop on the network to the second, and breaks it down into three phases:

Wheels stop to doors open

Doors are open

Doors closed to wheels start

Users of the tool can drill down to see dwell variation across a wide range of parameters, such as by station or service.

The development of the IDT tool was led by Porterbrook, in collaboration with Abellio ScotRail, data science company Elastacloud, and the University of Southampton. Analysis of the Abellio ScotRail data showed that over 40 per cent of dwell arises in Phase 3, which means there is more time between doors closing and wheel start than between doors opening and closing.

IDT also correlates dwell data with other factors, to help operators identify the root causes of delay. The tool could be used to compare options and justify operational or infrastructure alterations. Plans to enhance IDT include integrating it with existing industry tools and incorporating other data feeds, such as passenger loading.

Read more about IDT in our research catalogue.
Completed project: A feasibility study towards the conceptual development of a real-time digital twin to reduce dwell time variations on the Thameslink route

This project used simulations to establish correlations between delays and the flow and density of people at key nodes, such as ‘bottleneck’ stations. It has opened a technical path for industry to improve dwell time performance through real-time passenger flow prediction and crowd management.

OpenSpace Group, in collaboration with the University of Birmingham, developed a proof-of-concept prediction engine. This explores how data fusion and machine learning can be used to generate passenger boarding and alighting predictions from rail datasets, including computer vision and gateline systems.

Read more about this project in our research catalogue.

Text 7

Finally, another of the outputs from the ‘Faster, safer and better boarding and alighting’ competition, which was launched in 2016, is available to use in an operational environment.

Completed project: Intelligent Computer Vision agents

This project looked at how on-board and station-mounted cameras could be used to optimise the flow rates of passengers. The research, led by the University of Lancaster, demonstrated how autonomous and data-driven techniques can alert drivers of unusual behaviour near the platform-train interface. Operators could use this information to direct passengers to less busy parts of the platform and train, to help avoid or reduce potential crowding.

The product is commercially available from Digital Rail as Rail Sight Yellowline.

Haven’t found what you’re looking for?

Get in touch with our expert for more information.

Marcus Carmichael

Tel: 020 3142 5506

https://customer-portal.rssb.co.uk