It’s dirty, it’s dull, it’s enormous, and it’s probably one of the most productive robots on the planet
It’s often the case that the more useful a robot is, the less exciting it is. The robots that do the hardest jobs tend to be straightforward solutions to straightforward problems, because that’s what works. The (self-declared) world’s largest robot is an efficient, grubby example of this—it’s an autonomous train that recently hauled 28,000 metric tons of iron ore 280 kilometers across the Australian desert.
Australia is a big place, and it takes a lot of effort to get material out of the middle of Australia (where it’s not useful) to the coast (where it can be taken somewhere that it is). Trains are the most efficient way of doing this, and they travel back and forth through a whole lot of nothing, taking ore from mine to port and bringing the empty cars back again. It’s the kind of repetitive, fixed-route operation that seems like it would ideal for automation, and the Rio Tinto mining group has somehow spent nearly a billion dollars over the last seven years trying to make it happen.
After upgrading the infrastructure along the route of the train, including adding cameras to every single crossing, Rio Tinto started running its locomotives in direct supervised autonomy mode in early 2017. There was still a human present in the train just in case, but the train was operating autonomously, under remote supervision from an operations center in Perth, 1,500 km away. Earlier this month, the train for the very first time made the run using only remote supervised autonomy, meaning that it’s technically more of a robot than it was before, I suppose. Or at least, it’s now crossed some threshold of trust where it’s allowed to embrace its autonomy.
Rio Tinto expects that eventually, “AutoHaul will unlock significant safety and productivity benefits as a result of reduced variability and increased speed across the network, helping to reduce average cycle times.” What that means is that the train has a better understanding of its performance, the load its carrying, and the characteristics of the terrain, enabling it to drive itself faster and more efficiently than a human is able to.
What’s most surprising about this is how long it’s taken—it seems like making a freight train autonomous is a totally obvious thing to do. Think of how well constrained the problem is: The thing is on a track, and it can’t steer even if it wanted to. All you have to worry about is acceleration and braking, and to be blunt, there’s only so much braking you can do if there’s an obstacle on the track at short notice. And if there are these tangible speed and efficiency benefits, why isn’t every single train autonomous? The answer, somewhat depressingly, seems to be that the amount it costs to have one or two humans along for the ride is so low compared to both the cost of installing sensors on a train and the risk of liability if something gets screwed up that it’s just not worth the hassle in most cases. Unlike cars and trucks, the ratio of human to cargo on a train is extreme, so the amount of money that would be saved through automation is negligible. The same goes for cargo ships. Could you automate them? Sure. Do we have the technology? Absolutely. Would it be cost effective? I doubt it.
Honestly, I’m not entirely sure how Rio Tinto is justifying the $940 million it spent to make its AutoHaul system work. Maybe with exceptionally repetitive, high traffic, and very remote routes, it will pay off in the long run. I hope it does, and as much as I’d like to believe that this is the first step toward the automation of rail systems everywhere, that’s probably not something that’s going to happen anytime soon.
[ Rio Tinto ]