How do you drive if you can’t see the road? If you’re a driverless car, the answer may be to look under it.
US researchers are experimenting with the use of ground-penetrating radar (GPR) to give vehicles an idea of exactly where they are, even when snow, rain, darkness and other acts of nature are making that difficult.
A team from MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) used localised GPR (LGPR) to send electromagnetic pulses underground to measure an area’s specific combination of soil, rocks and roots.
The mapping process created a unique fingerprint of sorts that the vehicle could later use to localize itself when it returns to that particular plot of land.
In tests, the system’s average margin of error in snow was only about 2.5 centimetres compared to clear weather. It struggled a little more in the rain – likely, the researchers say, because rain wets the soil and increases the disparity between the original mapped LGPR reading and the current condition – but was still only off by an average of 14 centimetres.
Over a period of six months of testing, they never had to unexpectedly step in to take the wheel.
“Our work demonstrates that this approach is actually a practical way to help self-driving cars navigate poor weather without actually having to be able to see in the traditional sense using laser scanners or cameras,” says Daniela Rus, senior author of a paper to be presented to this year’s International Conference on Robotics and Automation in Paris.
A major benefit of mapping out an area with LGPR is that underground maps tend to hold up better over time than maps created using vision or LIDAR (light detection and ranging sensors), because features are less likely to change.
Rus and colleagues say they are the first developers of self-driving systems to trial GPR, which previously has been used in fields such as construction planning, landmine detection and lunar exploration.
And while they only tested the system at low speeds on a closed country road, they are confident it could easily be extended to highways and other high-speed areas.
There is still a lot of work to be done, however. For one, the current system is not capable of performing global localisation without using a GPS prior.
A major focus will be designing mapping techniques that allow LGPR datasets to be stitched together to be able to deal with multi-lane roads and intersections.
In addition, the current hardware is bulky and 1.8 metres wide, so major design advances need to be made before it’s small and light enough to fit into commercial vehicles.
Even if they get it right, it’s never going to be able to do the whole job, because it can’t detect things aboveground. But the MIT team suggests its ability to localise in bad weather means it would couple nicely with LIDAR and vision approaches.