The project involved the decentralisation of a computer controlled model railway. Put simply, this meant that processing had to be undertaken locally on each component (train or signal).

Traditional model trains are simplistic, dumb, and controlled from a central point. Clearly, this had to change, because the central point represented a weakness - if it stopped working, so would the whole network.

Although the delivered product was a model railway, the idea behind the project was to develop a platform to allow experimentation with train networks using unstructured networking and decentralised control. This means that the components had to undertake decision making on-board, operate safely in an uncertain network environment and not rely on the presence of other components.

Distributed processing could be useful in a real railway, doing away with the need for high level processing (over an entire region, for example) and provide resiliance from widespread disruption in the event of failures in small areas.

Interestingly, my final project supervisor (who is an expert in Distributed Systems) argued that this would not be possible. Recently, another expert, claimed that this idea might have merit, and noted that localised planning used to be common before privatisation (for example, freight was never timetabled, it was just laid up in sidings until the line was clear). I think that this indicates that this is an area which is perhaps unexplored. Are train timetables actually needed?

Ultimately, the actual experimentation with distributed automation was over ambitious and was not investigated.

NB:	This content was quickly written, and should not be taken as academic quality. Some errors may be present. I do not accept responsiblity for your use of this information.
	I am grateful for the support received from RadioMetrix, supplier of radio modules.
	This material is Copyright and not endorsed by the University of Essex or RadioMetrix.