Manipulation of autonomous driving made easy

Researchers at UC Irvine managed a simple manipulation of autonomous vehicles that could have dire consequences.

Again and again, researchers or specialized companies succeed in introducing autonomous vehicles X for a U to pretend Also a Artist achieved this with a simple line. Now, researchers at the University of California, Irvine, have discovered that ordinary roadside objects lead to unwanted vehicle behavior. For example, it brakes suddenly.

The objects placed were a box or a bicycle, which led to the behavior of the car. This not only poses a danger to the following traffic, pedestrian traffic could also be affected. Especially if this happens at a motorway entrance or exit, it could lead to accidents or traffic delays.

The researchers identified the problem in the software of the planning module, which controls the system. This component oversees the vehicle’s decision-making processes, which determine, among other things, when to drive, change lanes, or slow down and stop. The software works too carefully, so the problem appears.

To localize the weak points, the team developed software called PlanFuzz. The program explores three different behavioral planning implementations of the industry-standard open-source autonomous driving systems, such as Apollo and carware.

Go to Source