Project B1 - Decision Support System for Dynamic Re-Scheduling of Trains Under Disturbance
University of Birmingham
Univeristy of Southampton
Birmingham - Dr. C. J. Goodman and Dr. R. Takagi
Southampton - Prof. M. McDonald and Mr. J. Armstrong
Atkins; AEA Technology Rail; Mott McDonald
Railways operating near their theoretical capacity, which is routinely true of metros and increasingly so of substantial parts of the UK mainline network, are particularly vulnerable to disruption by human behaviour or engineering failures that lead to delays to or even cancellation of services. These problems may, and, in the UK, undoubtedly should, be addressed by increasing network capacity and improving operating reliability. However, these are long-term goals, and require enormous investment. In the short term (and generally), it is clearly advantageous to respond to disruptive incidents in such a away as to minimise the resulting problems.
Withour intelligent regulation of the consequences of such disruptive incidents, the consequential effects on passenger and freight services can be much greater than necessary. Traditionally, such regulation is provided by the control centre dispatchers. While concentrating the control function in smaller numbers of control centres provides an improved overview of the 'big picture', this very concentration, in conjunction with increasing traffic, potentially results in over-load of the control centre staff when things do go wrong. Whilst part of the answer may be in improved information presentation to the operators, it is equally certain that providing automatically generated 'best guess' strategies for operator approval should assist in alleviating their task-load.
The evolution of a viable real-time decidion support system to assist operators cope with schedule deviations on a railway can be based on a number of techniques and it is not clear abinitio which methods are likely to succeed. The work programme thus involves at least four stages:
- A review of existing industry practices, by means of (i) a literature search and review, and (ii) observation and analysis of a variety of typical operating railways practices. This will establish what is currently done by human operators when responding to disruptive incidents, and which are the major dynamic features and constraints that define what control actions are possible and/or desirable;
- Incorporation of this knowledge in an operational model or models that may be anywhere between the extremes of (i) a general mutli-train network simulator involving detailed modelling of traction performance, signalling systems, etc. and (ii) a set of abstracted rules, by combining techniques appropriate to railway link, junction, and through and terminal station requirements into a generic model or rule set, which takes into consideration the respective performance characteristics of different train types;
- Encapsulation of the knowledge regarding factors to be targeted and the desirability of different scenarios for recovery, compared to one another and to the 'do-nothing' scenario. In work already conducted by the members of the team, for example, this involved assigning penalties to excess waiting time and journey time (inter alia) experienced by metro passengers, Suitable 'weightings' may be applied to different passenger and freight train types to reflect tehir relative priorities; and
- Assessment of different methods for finding the best solution in accelerated time. Candidate techniques vary from classical gradient-based search procedures to simple sets of expert system rules that may bypass the search process altogether by remembering the outcome of off-line explorations. Either approach needs a general purpose multi-train network simulator to trial solutions and to act as a test-bed for evaluating different techniques. Such a program exists at The University of Birmingham.
- Development of Multi-train simulator;
- Incorporation of margin tables into optimisation process;
- System analysis of optimisation criteria.
Progress to Date:
- Development of multi-train simulator;
- Construction of network model based on Abbottswood Junction, near Cheltenham;
- Construction of model timetable based on working timetable for Abbottswood;
- Development of simple algorithm for optimisation based on train delay minutes.
You can download the report sent to the Railway Safety and Standards Board on the progress of project B1 up to September 2003 by clicking HERE