Automatic Train Operation

In the wake of this summer’s on-again, off-again BART strike, some of us wonder why a transit union can shut down what ought to be an automated train. Why can’t management “operate” (perhaps at a reduced schedule) the vehicles, as happens in many other organizations on strike? It is surprising that, in 2013, modern, grade-separated rail systems don’t all have automatic train operation (ATO). While there are a number of ATO systems internationally, it is rare in the US. At least part of the problem is lack of trust (deserved or otherwise). There have been at least two notable crashes of test-trains using automatic operation.

Fremont Flyer crash (October 2, 1972) – a train under testing with automatic control overshot the end of the track (link). 8143196966 cbee24a4ef b
A Precariously Positioned DLR Train On (March 10, 1987) DLR-overhang

However in one case (BART), drivers were re-inserted into the system (perhaps enabling this summer’s strikes), in the other (DLR), driverless trains remained standard (and safe).

The 2009 Washington Metro train collision was also attributed to failures in a track circuit part of the Automatic Train Control systems, and since the collision ATO has been disabled.

It should in principle be far easier to automate trains (which run on fixed guideways) than cars, which have to track many more moving parts in their environment. The technologies are very different, since the trains are centrally controlled, while any successful automated vehicles will be autonomous. The issue of course is not only design, but implementation, and robustness to faulty designs.