Researchers have developed a computer-assisted driving system that uses sensors to preempt and warn drivers of dangerous maneuvers.
Despite huge developments, consumers still have a significant wait before driverless cars become a viable option. In the meantime, researchers from Robo Brain and Brain4Cars, have developed a promising system for computer-assisted driving, which uses sensors and deep learning to preempt bad driving and alert the driver before they actually do it.
The majority of road deaths are caused by drivers attempting unsafe maneuvers, which is why Robo Brain monitors the drivers themselves, as well as external factors. The system is comprised of multiple sensors including cameras, wearable devices, and tactile sensors. The devices monitor the driver in real-time, and an algorithm learns the individual’s driving behavior. Over time it is able to anticipate a dangerous driving maneuver up to 3.5 seconds in advance and provide a warning that will discourage the driver from going ahead with it. For example, it might learn that glancing to the right frequently means the driver is going to overtake and alert them if it is likely to result in a crash.
The system was created by the Departments of Computer Science at Cornell and Stanford Universities. Researchers collected thousands of miles of natural driving data from a variety of drivers. It combines this data with the individual’s habits, which it can learn very quickly: in recent tests, after just over 100 miles, the system was able to predict the actions of 10 drivers with 90 percent accuracy.
Computer-assisted driving could help to pave the way for acceptance of driverless cars, winning over skeptics by showing the improvements to safety that computers can provide. Are there other technologies that provide a middle ground for users who are apprehensive of big changes?