Trains are super efficient. My uncle is a locomotive engineer. He has one or two conductors with a train that is a couple of miles long. They pay a premium to keep staffing at a minimum -- he can be called with 30 minutes notice to come in and drive a train for a 24 hour period.
You'd probably be able to run more smaller trains with automation, but competing with trucks is tough, because the efficiency gains associated with getting rid of people and adding automation are blown away by the need to cross-dock and do the last mile via local truck.
Yep! This is where the western / action movie trope of racing to cross the train tracks comes from. If you can get the train between yourself and your pursuer, it’s going to be a long time before they can do anything to catch up again. Often times in these movies, whoever is fleeing will take a few casual moments to look back at their blocked pursuer because there’s so much time to spare.
You need a lot more people to move the same amount of cargo in trucks, so there is a lot more money to be saved.
The consequences of individual crashes is also higher, which changes the perception and consequences of mistakes, even if the number of trucks means the actual scale of the problem might be similar.
The consequence of crashes is what led to the earliest implementations of automation on railways, starting in the 19th century: https://en.wikipedia.org/wiki/Train_stop
Arguable a mechanical interlocking also counts, as it prevented unsafe routes being set for a train, and made it faster to set the intended route: https://en.wikipedia.org/wiki/Interlocking
The quoted costs for positive train control just seem astronomical for what is actually being achieved;
> Under RSIA08, PTC is required on about 60,000 miles of railroad track. The Federal Railroad Administration (FRA) estimates full PTC implementation will cost approximately $14 billion. [1]
I understand that everything looks simple from the outside, and these have to be rugged, reliable, redundant, and failsafe, but that’s a lot of dollars per mile for sensors.
> PTC uses signals and sensors along the track to communicate train location, speed restrictions, and moving authority. If the locomotive is violating a speed restriction or moving authority, on- board equipment will automatically slow or stop the train. A more expansive version of PTC, called communications-based train control (CBTC), would bring additional safety benefits plus business benefits for railroad operators, such as increased capacity and reduced fuel consumption. However, CBTC is not currently being installed by any U.S. railroad, due to the additional cost and the challenge of meeting implementation deadlines.
So we’re talking about basically AEB (automatic emergency braking) for trains, and it costs $200k per mile to deploy? I don’t get it.
> Based on analysis of past PTC-preventable incidents, the FRA estimated in 2009 that $90 million in annual safety benefits will be realized after full implementation of PTC.
Oh good, breakeven without accounting for any time value of money in only 155 years. The only word that comes to mind is boondoggle.
Yeah, in a 2D problem space, with every vehicle already reporting position back to the network!
The two dimensions being track offset (distance from start of line), and time.
This isn’t even obstacle detection. This is a) obeying the speed limit, and b) not asking two known entities to occupy the same length of track at the same time.
It does not, for example, detect a car stuck on the tracks at the next railroad crossing.
I'm curious what the return looks like for self driving trains.
I'm going to bet for both safety and employment, trains and transport ships are orders of magnitude more efficient already. Trucks are a pretty big source of human inefficiency.
