Optimising the Loading Diversity of Rail Passenger Crowding using On-Board Occupancy Data

Ball, Simon David (2016). Optimising the Loading Diversity of Rail Passenger Crowding using On-Board Occupancy Data. MPhil thesis The Open University and Transport Research Laboratory.

DOI: https://doi.org/10.21954/ou.ro.0000bbbf


Crowded conditions on trains can lead to lower passenger satisfaction, discourage rail travel, result in negative economic impacts and are a factor in a number of health and safety hazards. In the UK there is an annual survey of rail passenger crowding, although the measures used do not reflect coach-by-coach variations, nor do they reflect variations across the peak period.

In this MPhil thesis I investigated the application of weight-based automatic passenger counting data to deliver more even loadings on trains through the provision of new real-time and static solutions. In addition I investigated the potential benefits of such solutions in terms of reduced dwell times and reduced crowding. The overall concept proposed was to make the most of the existing available capacity; for example, so that no-one is standing when seats are available. Through analysing a large sample of air suspension data, I identified station-specific trends where some coaches were over capacity while others had spare capacity. I also conducted a critical review of academic research into on-train crowding and solutions that seek to optimise ‘loading diversity’.

This study contributes to this emerging subject area in several ways: I propose two new metrics to describe inter-coach loading diversity that, unlike existing metrics, contain information relative to the capacity; I have revealed a link between the inter-coach loading diversity metrics and estimated boarding times, with trains classified as ‘very uneven’ on departure typically having dwell times of approximately five to ten seconds greater than services that were classified as being ‘even’ with a similar total number of passengers on board; and finally I have applied classification supervised learning techniques to predict the load factor for a given service and these predictors were an improvement over taking the historical average.

Viewing alternatives

Download history


Public Attention

Altmetrics from Altmetric

Number of Citations

Citations from Dimensions

Item Actions