Intent Inference of Driver's Deceleration Maneuver and Its Safety Impact on a Platoon of Vehicles

Rear-end collisions in longitudinal car-following situations occasionally cause severe multiple collision accidents, especially in high-density and high-speed traffic situations. However, if a preceding car driver's deceleration intention could be inferred in advance, even a second earlier, the risk of a collision could be decreased not only between vehicle pairs but also for car platoons. This paper reports on the application of our previously reported intent inference system to a platoon of vehicles and evaluates its impact on safety. The system utilizes an unscented Kalman filter (UKF) and attempts to infer the intent of a vehicle ahead 1.5 s in advance of its driver's most likely action. The inferred intention is then provided to the driver of the assisted vehicle through an interface based on a windshield display, and earlier experiments have proved that the system is quite effective for collision risk mitigation. In this paper, we report on an investigation into the safety impacts of our system on a platoon of vehicles following an assisted vehicle. The results of a traffic simulator experiment showed that the system could provide significant safety benefits to multiple cars traveling behind an assisted vehicle in a vehicle platoon.