Distracted Driving Detection in Stop-And-Go Traffic

A recent survey showed that nine in ten US drivers have engaged in distracted driving. Rear-end collisions are the most frequent type of collision, and most of them are due to distracted follower vehicles. When drivers get distracted, they exhibit distinct driving behavior such as fluctuating distances in stop-and-go traffic caused by delayed response. Early detection of such distracted driving accurately is beneficial to mitigate the risk of rear-end collisions. In this paper, we focus on this use case. Ego vehicle uses distance-to-collision measurements and performs time-series analysis to identify fluctuating distance in stop-and-go traffic caused by the delayed response of distracted followers. Whenever the ego vehicle detects a distracted follower, it notifies the driver with a lane change recommendation to avoid collisions. We demonstrate the feasibility of distracted driving detection through outdoor experiments with multiple test vehicles in real-time. Our field trials have shown that we can identify distracted drivers in stop-and-go traffic by about 65% accurate detection with a 35% misclassification rate.