It seems that Monday’s fatal metro train crash in Washington DC was caused by a failure of a block control signal that should have indicated the presence of a train in a block of track. It is incredible to me that in the 21st century that such a single failure point could result in a crash. Apparently Metro has two other train location information systems but neither has direct authority over the trains.
In other words, so far as positive train control is concerned, trains can just disappear from the system if a block sensor is defective. Instead, Metro should have a system that maintains continuous knowledge of where a train could be based on the totality of information from various sensors (including GPS when trains are in the open and other positive location techniques when trains are in tunnels), and this picture should be extended by physics to provide a range of uncertainty about where trains could be based on the information in the system. The block control information should serve as a back-up to this positive control picture, stopping trains as necessary but only once the primary system has failed to provide adequate control.
All trains that could possibly overlap in space should come to a stop when the block control information diverges from the positive control model. The block control information and the identified train location information should be continually checked against each other, triggering train stoppages and immediate repair to resolve any discrepancy.
Disclaimer: I am not a transit engineer and am relying solely on information published in the Washington Post.