New Vehicle System Would Help Autonomous Cars Avoid Collisions

An MIT team has developed a new vehicle system that can help self-driving vehicles see around corners.

Currently, self-driving vehicles navigate using a rotating device called the LiDAR system.

LiDAR, which acts as the car’s eyes, provides a 360-degree view of the surrounding. As a result, autonomous vehicles can drive themselves safely through traffic congestions, potholes, trees, and other obstacles on the road.

While LiDAR is quite impressive, it can only detect a visible object. That means, it’s ineffective at sensing moving objects coming from around the corner, which could result in a collision.

Now, the engineers at MIT have developed a new vehicle system that does just that. It uses tiny changes in shadow to identify approaching objects.

In a statement, co-author of the study and director of the Computer Science and Artificial Intelligence Laboratory (CSAIL), Daniela Rus said:

“Our method can give the robot an early warning that somebody is coming around the corner. So, the vehicle can slow down, adapt its path, and prepare in advance to avoid a collision.”

So, how does the system work?

A Vehicle System That Can See Around Corners

To create the autonomous vehicle system, the researchers had to rely on a previously built system called “ShadowCam.” Using computer-vision techniques, ShadowCam detects and classifies changes in shadows on the ground.

However, applying this technology in autonomous vehicles required a little tweaking.

The researchers developed an ingenious process that combines image registration with a new visual odometry technique. Image registration essentially overlays multiple images to reveal possible variations.

Meanwhile, visual odometry analyses pose and geometry in sequences of images to estimate the motion of a camera in real-time. Next, the researchers used signal amplification to detect weak signals from shadow changes.

This resulted in a system that can help autonomous drive safely around a parking garage.

The first author of the paper, Felix Naser noted:

“By detecting that signal, you can then be careful. It may be a shadow of some person running from behind the corner or a parked car, so the autonomous car can slow down or stop completely.”

In a test, the team was able to prove that the new vehicle system is faster and more efficient than the traditional LiDAR system.

Be that as it may, the MIT team only tested the system in indoor settings where robotic speed and lighting conditions are more consistent. As a result, it was easier for the system to sense and analyze shadows.

Now, the researchers are working on an outdoor system that can work under various light conditions.

Read More: Blind Spot Algorithm Makes Driving Safer (for Autonomous Vehicles, too)