ATO testing on the District

There's an article in Modern Railways of May 1963 titled Development of un-manned trains by London Transport. Includes a picture of a train testing automatic control on the South Ealing Test Track (I think the EB Fast in those days). Also actually operated in service with one train for a short period from Stamford Brook to Ravenscourt Park EB only.

Computers are much better than humans at most routine tasks, humans are much better than computers at most exceptions to the routine. This has been true since the very early days of computing, and until we get a working artificial general intelligence it's not going to change any time soon. Improvements to automation of things like train driving have been largely driven by increasing the number of "routine" situations the software knows how to handle and improved recognition of which routine situation it is in.

South Ealing test track was initially used in December 1962 with a later passenger service trial between Stamford Brook and Ravenscourt Park (eastbound only) in April 1963.

Whilst what you've said is absolutely correct*, the big problem for unmanned trains, (and driverless cars), is that the number of possible exception conditions is enormous, and it would be a foolish person who claimed to have a complete list.


* Or, perhaps, my opinion is the same as yours.

Can I point out that those early 1960s District ATO tests, and the 1967 Victoria Line*** , NO COMPUTER was responsible any train protection and movement authority, that was all done by relays and or V-frames. True there are computers between man signaller and train, and holding timetables, and relaying signallers push button commands etc requesting interlockings (relays or V-frames) or junction machines to do things, but actually processing movement authority NO.

Even the Central Line is not all processor based - part of the line (mostly, but not uniquely, the western end) is relay interlocking, and part (mostly, ditto comment, the eastern end) Westrace, but even then at some of the Westrace interlockings one or two (actually I think it is 3 ?) sites the ATO code generatos are relays. Yes there are control room PC command workstations and central kit (CSPs etc) at Wood Lane, and yes there are 16 Local Site Computers [LSC] along the line - but not one of them is responsible for train protection or movement authority, they merely request relays or Westraces to do things, or report back to Wood Lane what the Westrace or relays have done.

The reason I mention this it makes it not 50-60 years of computers completely running ATO, but a much much shorter time, like 2000s, before you do have "computer" processors doing it all. When you start comparing time lines with fuel cells and nuclear fusion they are not so different. We are a lot further from this than many would have you believe.


*** yes, before anyone refers to it, the often seen Victoria Line 1960s documentary does say they were computers, but no they were not, this was media BS (refer to an IRSE lecture by the late Eddie Goddard who showed a clip from it in that lecture with a similar pithy comment).

I've been off the Central 5 years now, memory lapsing perhaps.

I don't think it's fair to say that saying the Victoria line was run by computer is BS.

It was run by a programmable system (machine), and, certainly in the past, that would have been good enough to warrant the epithet 'Computer Controlled'.

I would dare to suggest that train protection systems will never be solely computer controlled (in the sense that we understand a computer today), because they cannot be made fail safe. (No matter how good the programming it is always possible for the hardware to fail in unpredictable ways.)

If you re-read what I wrote re. Vic. line, I said no computer in train protection or movement authority; I did not say no computers full stop.

Moving to the C21, Jubilee Northern and 4LM are all computer train protection systems.

And therein lies the rub.

There will be circuitry built into the system that will examine the output of the three computers and that is where the ultimate failsafe will reside. (I should have said 'solely' computer controlled.)

e.g. Suppose you have a simple crossing of two single tracks, a and b.

Although computers may be involved at many levels I would bet serious money that there will be an extremely simple and extremely robust system in place that says: if line a is 'open' the line be will be 'closed', and vice versa. The same will apply, mutatis mutandis, at every contention point in the entire system.

2 out of 3 of the computers need to agree on every input before it will accept it

That logic extends to the computer - it is still ultimately humans that set the parameters and do the programming and testing.

CBTC does not use track circuits

The primary location system for CBTC is the VOBC on the train (computers) reporting their position to the VCC (computers), using the RFID tags on the track for position information.
The back up uses axle counters which again feature a microprocessor based system (and yes, if you want to interrogate it, you need a computer).

One of the weaknesses with an axle counter system is that you can leave a detection head off the rails and it will still show the section as unoccupied (at least until the first train).

Yes, good point. What I meant is you could remove the rail it is mounted on and leave it in the cess, and the system would still function, as has happened.