Self-driving vehicles use a combination of sensing technologies to be effective, usually optical cameras, and lidar. Unfortunately, both of these rely on reflected light and are made problematic by snowy conditions. However, there may be a way around it.
MIT’s Lincoln Laboratory long ago developed a Ground Penetrating Radar (GPR) system to help the military detect landmines. However a startup named WaveSense has been attempting to apply that technology to self-driving cars, and they may have just succeeded.
GPR scans the ground underneath a vehicle upto a depth of around 10 feet, building a picture of what’s going on beneath the surface you’re driving on. In self-driving cars, WaveSense compares this data to GPR data that’s previously been collected for road networks, and is then capable of placing the car to within even a few centimetres.
Sure, you can say that this technology relies on the roads being mapped beforehand, but then again so does lidar. In fact, WaveSense says remapping with ground-penetrating will be required less frequently because the rock underneath a road isn’t as likely to change as the actual ground.
There’s also the plus point that adding a GPR sensor isn’t expensive at all, only costing about $100 per vehicle according to the startup’s CEO. After all, it’s been through heavy development and testing by the military and is about as rock solid as it can get. In addition, it’s the first sensing method for self-driving cars that’s completely unfazed by snow and its reflectiveness.