Researchers at Michigan State University (MSU) are conducting a study to understand the causes of phantom braking in autonomous vehicles in order to enhance safety. Phantom braking refers to occurrences where the vehicles equipped with driver-assistance technologies suddenly apply the brakes without any visible obstacles present. These incidents can lead to a loss of confidence in autonomous driving technologies, potentially hindering their widespread adoption.
Qiben Yan, an assistant professor in the College of Engineering at MSU, emphasizes the importance of addressing phantom braking. He states, If riders perceive the technology as unpredictable or unreliable, they’ll be less likely to embrace it. To better comprehend why phantom braking happens and how to prevent it, the researchers are focusing on the vision systems of autonomous vehicles. These systems consist of multiple cameras and radar that collect information about the vehicle’s surroundings.
Previous research conducted by Yan and his team uncovered vulnerabilities in these vision systems, which could be exploited by hackers. By projecting lights into the vehicle’s cameras, they were able to deceive the system, resulting in the car hitting the brakes unexpectedly. The researchers also demonstrated the ability to make an object in front of the car disappear from the camera’s view, causing the vehicle to collide with it.
To further investigate the security of autonomous vehicles’ vision systems, MSU has received a $1.2 million multiyear grant from the National Science Foundation. The research team, in collaboration with Virginia Tech, aims to study how cameras perceive phantom attacks and develop measures to enhance the security and resilience of these vision systems against malicious attacks.
Yan and his team are drawing inspiration from neuroscience studies on human perception to understand how artificial intelligence (AI) systems behind vision systems can be deceived. They are particularly interested in programming the AI model to accurately interpret the environment captured by the cameras. By studying different levels of perception, from low-level perception, which involves recognizing the scene, to high-level perception, which involves understanding an object’s speed and its relation to the overall scene, the researchers aim to improve the AI model powering the vision systems.
The research conducted at MSU will contribute to the advancement of autonomous vehicles, making them smarter and safer. Yan highlights that addressing the issue of phantom braking is crucial not only for improving vision technologies but also for ensuring the overall viability, safety, and success of autonomous vehicles in the future.
In conclusion, the study being conducted by researchers at MSU seeks to investigate the causes of phantom braking in autonomous vehicles and develop methods to enhance their safety. By analyzing the vulnerabilities of vision systems and drawing from the field of neuroscience, the researchers aim to program AI models to more accurately interpret the environment captured by the vehicles’ cameras. This research is a significant step toward making autonomous vehicles more reliable and trustworthy in the eyes of the public and ultimately promoting their widespread adoption.
