Tesla is known for its various technological advancements in the automobile industry. However, a new research targeted at its tech (Autopilot) suggests that it is not as advanced as we thought.
According to a research report by Tencent’s Keen Security Lab, Tesla’s Autopilot module comes with a fault that can trick its lane recognition feature into driving the car in the opposite direction.
The discovery has been supported with an “adversarial” example which suggests that when three red patches were added to the picture taken by the car’s camera, the spots were recognized as a lane.
Additionally, physical red patches were added to the ground in an intersection. When the car was in Autosteer mode, its lane recognition considered the red patches as the continuation of the lane it was on and neglected the real left lane.
Upon reaching the middle of the intersection, it took the real left lane (it now considers the actual left lane as the real right lane), leading to a wrong path altogether.
As for the basics, Tesla’s lane recognition feature uses computer vision to carry forward its tasks and can perform in an environment such as no strong light, rain, snow, sand and dust interference.
However, the research implies that the technology is not as good as it seems, as it can be prone to malicious attacks and can easily be manipulated.
We hope Tesla has a way to fix this issue ASAP. What are your thoughts on the same? Comment below and keep reading Fossbytes.
