Optimizing the operation of rail networks using simulations is an on-going task where heuristic methods such as Genetic Algorithms have been applied. However, these simulations are often expensive to compute and consequently, because the optimization methods require many (typically >104) repeat simulations, the computational cost of optimization is dominated by them. This paper examines Bayesian Optimization and benchmarks it against the Genetic Algorithm method. By applying both methods to test-tasks seeking to maximize passenger satisfaction by optimum resource allocation, it is experimentally determined that a Bayesian Optimization implementation finds 'good' solutions in an order of magnitude fewer simulations than a Genetic Algorithm. Similar improvement for real-world problems will allow the predictive power of detailed simulation models to be used for a wider range of network optimization tasks. To the best of the authors' knowledge, this paper documents the first application of Bayesian Optimization within the field of rail network optimization.
Optimising the allocation of limited resources, be they existing assets or
investment, is an ongoing challenge for rail network managers. Recently,
methodologies have been developed for optimising the timetable from the
passenger perspective. However, there is a gap for a decision support tool
which optimises rail networks for maximum passenger satisfaction, captures
the experience of individual passengers and can be adapted to different
networks and challenges. Towards building such a tool, this thesis develops a
novel methodology referred to as the Sheffield University Passenger Rail
Experience Maximiser (SUPREME) framework. First, a network assessment
metric is developed which captures the multi-stage nature of individual
passenger journeys as well as the effect of crowding upon passenger
satisfaction. Second, an agent-based simulation is developed to capture
individual passenger journeys in enough detail for the network assessment
metric to be calculated. Third, for the optimisation algorithm within SUPREME,
the Bayesian Optimisation method is selected following an experimental
investigation which indicates that it is well suited for ‘expensive-to-compute’
objective functions, such as the one found in SUPREME. Finally, in case studies
that include optimising the value engineering strategy of the proposed UK High
Speed Two network when saving £5 billion initial investment costs, the
SUPREME framework is found to improve network performance by the order
of 10%. This thesis shows that the SUPREME framework can find ‘good’
resource allocations for a ‘reasonable’ computational cost, and is sufficiently
adaptable for application to many rail network challenges. This indicates that a
decision support tool developed on the SUPREME framework could be widely
applied by network managers to improve passenger experience and increase
ticket revenue. Novel contributions made by this thesis are: the SUPREME
methodology, an international comparison between the Journey Time Metric
and Disutility Metric, and the application of the Bayesian Optimisation method
for maximising the performance of a rail network.
This paper considers the design and construction of new passenger rail networks – a common and expensive task worldwide. Optimising the allocation of initial investment between components of a proposed network is a challenge for planners who wish to reduce the initial investment required whilst maintaining key objectives, such as providing high levels of passenger satisfaction. Previous decision support tools to assist with this challenge typically do not consider the influence of the high-level network specification on passenger satisfaction. Here, a novel optimisation methodology is presented that includes the effects of various factors, e.g. permissible line-speed, train performance, train comfort, and station comfort. The methodology combines Agent-Based Modelling, Bayesian Optimisation, and a model which quantifies passenger experience. To demonstrate its effectiveness, the methodology is illustrated in a hypothetical case-study where it identifies a network specification which substantially improves the satisfaction of the virtual passengers. Furthermore, the network specification reduces the modelled cost of network construction by £5 billion from £64 billion. The case-study demonstrates that the methodology is computationally tractable for realistically sized tasks, and captures the trade-offs between investment and component performance, making it potentially useful to network planners concerned with the satisfaction of passengers.
'%3e%3cpath fill='%23012169' d='M0 0v30h60V0z'/%3e%3cpath stroke='%23fff' stroke-width='6' d='m0 0 60 30m0-30L0 30'/%3e%3cpath stroke='%23C8102E' stroke-width='4' d='m0 0 60 30m0-30L0 30' clip-path='url(%23b)'/%3e%3cpath stroke='%23fff' stroke-width='10' d='M30 0v30M0 15h60'/%3e%3cpath stroke='%23C8102E' stroke-width='6' d='M30 0v30M0 15h60'/%3e%3c/g%3e%3c/svg%3e)
'/%3e%3cpath d='M200 752.362h200v300H200z' style='fill:%23fff;fill-opacity:1;stroke:none' transform='translate(0 -752.362)'/%3e%3cpath d='M400 752.362h200v300H400z' style='fill:%23ff883e;fill-opacity:1;stroke:none' transform='translate(0 -752.362)'/%3e%3c/svg%3e)