With more and more self driving cars hitting the road, risks that rarely occur in lab conditions begin to surface.
One such problem which has become somewhat common is the innability for self driving cars to compute exactly where they are when driving in bad weather. Since most self driving cars use LIDAR systems (light detection systems) combined with cameras placed around the vehicle, the cars ‘sight’ is limited in a similar way to human vision – if it can’t make out objects or markers through luminosity, it can’t see it.

While developers are aware of this problem, coming up with a solution – such cameras with more capable sensors – proves to be a challenge.

But according to a team at MIT, using convential cameras may not be the way forward. Rather, the team has demonstrated a new system which maps out the ground underneath the car using a new radar system – Localising Ground Penetrating Radar (LGPR).

The team at MIT claim that the radar system can identify it’s exact position by mapping out the materials and elements underneath it – such as tarmac or steel – which it records and uses as a reference for a ‘map’, which it uses to determine it’s exact location and where it is currently heading.

The team also claim that the radar updates in real time, allowing the car time to adjust for changes in detected terrain – such as the car slowly drifting off of the road.

But perhaps the greatest advantage this new system holds is it’s ability to ‘see’ at a depth of up to ten feet. This means that the technology will likely work in many conditions, and on many types of surfaces, since the ‘map’ it builds can see various elements and patterns at considerable depth.

As the team continues to work on this new system, self driving car manufacturers are ceratin to take notes and begin implementing similar technology in the future.

Source: https://www.csail.mit.edu/news/self-drive-snow-look-under-road

Image credits: MIT CSAIL