MIT is teaching self-driving cars to deal with the unknown
There are degrees of challenge for self-driving cars. Grid-based cities are the lowest of low-hanging fruits, and that’s one reason why early tests took place in California, where not only are the roads straight, but there’s little rainfall and conditions are generally favourable. The problem, of course, is that the majority of the world isn’t made up of these dream circumstances – how does a driverless car cope when there’s no 3D-map to guide it?
Country roads are the best example of this: they have very few markings, and are often completely unmapped because so few people use any given one that there’s no real incentive to map them. But if driverless cars can’t cope, then does that mean future generations will have to learn to drive just for these scenarios?
Not if MIT can help it. Researchers are currently working on a new framework called Maplite which they hope will allow driverless cars to figure out roads they’ve never driven on before, without the need for 3D maps. Combining Google Maps GPS data with the car’s own LIDAR and IMU sensors, the idea is that the car can figure out the way the road is twisting and turning without needing to know exactly what lies ahead.
This is how that looks in practice:
If you’re thinking that that sounds a lot like how humans navigate, then you’re right. But that’s wholly different to how self-driving cars tend to operate. “The reason this kind of ‘map-less’ approach hasn’t really been done before is because it is generally much harder to reach the same accuracy and reliability as with detailed maps,” lead author on a related paper, Teddy Ort, told MIT News. “A system like this that can navigate just with onboard sensors shows the potential of self-driving cars being able to actually handle roads beyond the small number that tech companies have mapped.”
Ort doesn’t believe that this system will see the end of 3D-map usage, or spell a change in how self-driving cars operate more generally, chiefly because the more information autonomous vehicles have, the better. “I imagine that the self-driving cars of the future will always make some use of 3D maps in urban areas. But when called upon to take a trip off the beaten path, these vehicles will need to be as good as humans at driving on unfamiliar roads they have never seen before.”
It’s not quite road-ready yet, though. For one thing, Maplite can’t deal with mountain roads thanks to the elevation. For obvious reasons, you want your car to be extremely confident at second-guessing roads in these conditions, but this research is a very good sign that we’ll be able to tear up our driving licences in the next couple of decades without regretting it when a car downs tools at the city limits.