Montreal has plans to expand idts automated light rail to 99Km by year 2029. Rather hopeful considering how Covid-19 has put a delay on many differnt system expansions.
To what extent is it automated?
It’s going to be interesting. The REM system was designed to be GoA level 4 (the highest) which does not involve any direct crew on the trains at all. I can provide a listing (from APTA) of the specific functions that a system at this level must provide; it is interestingly extensive.
Even more interesting is the decision to greatly increase build commitment based on the present state of construction – it shows faith in the ability of the systems integrators and Alstom to produce an effective and ‘likeable’ set of lines in the intended timeframe. I confess that I like what I’ve seen so far, and I hope they succeed fully as intended.
Presumably a fully automated system would either have to run under ground, with restricted access to the tunnels, or on a fenced-in right-of-way on the surface with cameras, etc. to detect intrusions.
My car has a radar and camera system that is able to detect vehicles in front of it, as well as pedestrians in some situations, and use the engine to slow down or actually apply the brakes, although the system will not bring the car to a complete stop. Is this the kind of technology that would be installed on an automated train?
Interestingly, the publicity material for the system makes much of the views that riders will have out of the windows as the train runs, so I think that great reliance is at least tacitly being made on stuff being kept out of the way of the train. If you look at the APTA functions list for transit GoA 4, you will see that the level-4 ‘system’ handles many of the tasks you’re concerned with – and this implies multiple sensor systems and effective logic to drive them and fuse the output sensibly.
My suspicion is that the ‘production’ railcars will be using some combination of radar/‘lidar’, camera sensing, and external devices like inductive loops to determine potential issues. There are probably some technical descriptions ‘out there’ already, as this is one of the things that would have to be well-defined before it would make sense to ‘extend’ the system before it is proven in testing; I just haven’t seen them.
The current anticollision systems I have seen work remarkably well, in a couple of what are interestingly distinct ways. I had a 2020 Prius on a recent business trip, and in awful New York traffic I found the anticollision brake actuating very sharply precisely as I got on the pedal judging from long experience that silliness was producing a developing need for emergency deceleration. That is an indication that very good algorithms are coupled with very good machine vision and ‘comprehension’ of events in space – the sort of thing required for railcars to work in their dedicated rights-of-way but watching for things that might obtrude on them.
The laser cruise is also a thing of beauty, with one little foible: it does not alert you when it slows