This can really take off once self-driving matures. The _main_ problem with freight train is not their speed or the rail track throughput, but the time it takes to sort the train cars ("dwell time"). It's so bad, that the average car "speed" can be around 10 km/h. Or even slower.
And the railroads do not particularly care about optimizing their network, they are content to milk the bulk hauls for as much profit as they can. My friend worked at a startup that tried to pitch fully automated couplers to rail companies. They didn't care, even though it could have cut the dwell time significantly.
But if the improvements can be made on the _cargo_ side, then it's a different story.
Self driving is much less important for the sorting process than new couplings (i.e. https://en.wikipedia.org/wiki/Digital_automatic_coupling)
Yep. That startup tried to do something similar, except in a backwards-compatible way, and for American railroad standards.
In 2005. Train companies didn't care.
> They didn't care
I find that hard to believe, anything that could reduce time in transit and switching yard labor would be attractive. The process of assembling a train is far more automated today than it was in the past, so evidence does not support that they are content to just "milk" their current business.
> The process of assembling a train is far more automated today than it was in the past, so evidence does not support that they are content to just "milk" their current business.
Not really. If you take a railroad worker from the 1980-s, they would be able to work, with only minor training.
The dwell time actually _increased_. Rail companies are focusing on hauling bulk goods (coal, construction materials, oil, etc.) rather than trying to compete with trucks for fast delivery.
It's far easier to optimize for throughput than latency, after all. And rail companies are local monopolies, so they're doing whatever brings more money next quarter.
> If you take a railroad worker from the 1980-s, they would be able to work, with only minor training.
The same could be said about computer programmers.
Trains and trucks serve two different markets. Trains are better for long hauls of bulk goods or containerized cargo where you have a lot of stuff all going from one place (e.g. a port) to another (inland distribution hub).
Trucks are good for "last-mile" local delivery or small loads/single containers going from one place to another.
> The same could be said about computer programmers.
Not really. Programming has fundamentally changed since the 80-s: version control systems, connectivity, new and more efficient languages, etc. Train yards have not changed a bit. Dispatchers might have computers now, and individual train cars can be tracked in a central DB, but the physical work of coupling/decoupling cars and shuffling them around has not changed AT ALL.
> Trucks are good for "last-mile" local delivery or small loads/single containers going from one place to another.
The US is special, it's geographically HUGE, so trucks end up playing an outsized role in long-distance transport.
Trains are much cheaper and more efficient, so they can potentially help to reduce CO2 pollution _and_ the transportation cost. But train companies are just not interested in that.
Having self-driving trucks transported on the interchangeable flatbeds can potentially change that. Trucks can just drive onto the waiting traincars, ride to the destination location, and then just drive off the flatbeds.
>I find that hard to believe
History is littered with complacent businesses that failed to innovate.