MIT's tech can help self-driving cars 'see' around the corners

Self-driving vehicles rely on LiDAR, aka Light Detection and Ranging, and AI algorithms to 'see' what's on the road and act accordingly.

The tech helps cars dodge incoming obstacles - cars/people - and navigate. But when it comes to corners, say like a bind turn, it struggles to detect the object on the other side owing to the lack of data.

To tackle this problem, the folks at MIT have developed a solution called 'ShadowCam'.

The system employs a series of cameras to analyze a very specific region of the road - the point where two perpendicular paths meet - and discern falling shadows to determine if a vehicle or person is coming from around the corner.

As a self-driving vehicle approaches a T point, the cameras record a series of frames, which are intelligently processed by the system to detect the presence of shadows and changing light conditions.

This data ultimately helps the system determine whether an object is coming from the other side as well as predict its position and relative speed, too.

Once the shadow data is processed and the presence of a vehicle is determined, the system gets enough time to preemptively guide the vehicle to slow down or come to a complete halt.

This could prove extremely beneficial in avoiding accidents between vehicles moving quickly towards an intersection (T-point), even when there's no direct line of sight between them.