Preventing driver mistakes before they happen
A new car safety system that anticipates driver errors a few seconds before they occur could mean a breakthrough in safety technology. The system was developed by a Cornell University professor and will be introduced at a conference in Rome in June.
The system is based on a computer algorithm that observes the driverês body language combined with cameras or radar that observe the environment outside the car. The algorithm can then calculate the probability that the driver will take an action such as a potentially dangerous turn or lane change. If that is likely, the system will alert the driver with a light, sound or vibration.
The algorithm predicted the driverês actions correctly 77 percent of the time, an average of 3.5 seconds in advance. That could be enough time to save lives, said Ashutosh Saxena, an assistant professor at Cornell who developed the system with a graduate student.Download Bulletin PDF