Computer Vision – Helping AVs See Around Corners

Engineers have created a computer vision system that can detect minute changes in ground shadows to determine if a moving object is around the bend to increase the safety of autonomous systems.

One day, autonomous vehicles may use the system to swiftly avoid collisions with other vehicles or pedestrians approaching from behind a building’s corner or within a space between parked vehicles. The system may be used by robots that travel through hospital corridors to transport supplies or medications to avoid running into people.

The researchers outline fruitful trials with a self-driving vehicle driving through a parking garage and a self-driving wheelchair traversing a set of corridors. The car-based computer vision system outperforms conventional LiDAR, which can only identify visible objects by more than half a second when sensing and stopping an approaching vehicle.

Although it might not seem like much, the researchers assert that fractions of a second count when it comes to swiftly moving autonomous cars.

The technique can provide an early warning that someone is approaching a corner for applications where robots are traveling through environments with other moving objects or people, allowing the vehicle to slow down, adjust its course, and get ready ahead of time to prevent a collision. The big goal is to give fast-moving street cars something like “X-ray vision.”

Read more

Related Content: Compact Lidar Steers Small Autonomous Vehicles