Traffic Congestion Mitigation by Deceleration Control with Short-term Velocity Forecasting Using V2X

Traffic congestion on highways on sags is caused by temporal road capacity reduction brought by slow-down of some vehicles at a gentle uphill section. As a countermeasure to this type of congestion, intentional deceleration of vehicles running behind the congested car group is known to be effective. To enhance the performance of this control, early detection of congestion is a key issue. Although existing studies of deceleration control mainly assume that congestion is detected using current velocity values, it would be effective if the likeliness of congestion occurrence before it actually occurs using time sequence of velocity data. This paper proposes a method to forecast traffic congestion by time-series forecasting using velocity information of autonomous vehicles periodically reported to a server via V2X, and starts velocity control immediately after the forecasted velocity value becomes lower than a threshold. Performance evaluation by traffic flow simulation is conducted and its results show that the proposed method largely reduces the percentage of vehicles whose average velocity is below 50km/h during congestion compared to the method that determine the starting time of deceleration control without velocity forecasting.