The delivered project consisted of a pair of trains, three signals and a PC (acting as controller). The trains operated on a loop of track, with a siding connected using a manual junction.

Each component consisted of the processing unit (a Cyan eCog1k evaluation board), a radio modem (RadioMetrix TDL2A) and a number of support peripherals (For example, the trains had motor controllers and the signals had LED output.).

The Layout

Demonstration

The demonstration consisted of showing:

• trains controlled from the PC, over the radio network.
• more than one train independantly controlled at the same time.
• PC status display of the trains and signals.
• signal-signal chaining, for three-aspect signalling.

Three Aspect Signalling Over-the-Air

Three aspect signalling essentially showed that the network could operate without the PC. In such a system, when the track ahead is clear, the signal will show green (if the next signal is not red) or yellow (if the next signal is red). Traditionally this is done using wires, but in the network it is done over the network.

Essentially, each signal broadcasts its state regularly, and each signal listens to the state of the next signal. The danger signal still comes from traditional track circuitry, and the state from the next signal expires after a fixed period (it must be refreshed). This makes the approach very simple, very safe and very effective. For more details see the presentation.

Why did I do it?

Forgetting all the stuff above, it became quite obvious during last summer that I needed to get experience working with embedded systems and networks. The usual "lets make a database interface" would not be enough for my final project and is not allowed on my degree scheme.

I used a model railway for two reasons. Firstly, there is a lot of activity going on in Europe with a new network-based train control system: obvious scope for employment. Secondly, to implement the above on robots would have required me to do something useful with them: this would have required guidance and localisation methods, which is very time consuming and would have distracted from the network element. This was not necessary with trains for rather obvious reasons.

What did I originally want to do?

Produce a completely automatic model railway network, with changing junctions and autonomous trains. Way too much for the time available...

What was missing?

Observant viewers will notice two rather obvious limitations: there are no automatic junctions and the trains cannot localise themselves. This makes practical use of this solution rather difficult at this stage!

It was not possible to implement Junctions in the available time, and such work involved problems that were either electronic or mechanical. This was a Computer Science project, and such work could not be justified.

Localisation of the trains would have been quite simple, barcodes would have been placed on the track and read by the trains. These could have consisted of the address of the upcoming signal, and the distance to the signal. With this information a train could listen to the state broadcasts of the signal, and stop in the correct distance if necessary. Unfortunately, finding a compatible barcode scanner proved too expensive, which is a great shame.

What next?

As a minimum, code readers should be added to the trains. This would allow unattended operation, demonstrating that two trains cannot collide. (A larger layout would be needed to allow sufficient 'head way'.)

Ideally, junctions and improvements to the network model could also be implemented.

Conclusions

A working radio based network was implemented, and a means of sharing information for information-driven activity was implemented. There were some shortcomings, but the modular nature of the system combined with the layered network architecture may permit these changes to be made relatively easily.

The demonstration could have gone a lot better. I managed to both collide and derail the trains, and I did not have enough time to demonstrate all the features implemented.

Finally...

I was invited to apply for a PhD analysing potential train control and protection methods (such as ERTMS) for the UK market, but I had to decline because it involved too much economics!

In the end, I went on to undertake research in wireless sensor networks - which are even more constrained than the equipment I used here.

Questions

I still welcome questions about this project, but I regret that it cannot be demonstrated easily because I no longer have access to the CPU boards.

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.