A stochastic microscopic based freeway traffic state and spatial-temporal pattern prediction in a connected vehicle environment

Traffic state prediction forms the basis for effective and efficient traffic control and management strategies. A model-based traffic state prediction approach based on a stochastic microscopic three-phase model is developed to predict traffic flow, speed, and travel time in short prediction horizons consisting of multiple time steps ahead. The proposed model utilizes connected vehicles’ trajectory data including location and speed information and fuses this information with detector measurements using an Adaptive Kalman filter. Stochastic driver behaviors in merging, lane-changing, and over-acceleration are considered in the three-phase microscopic model, which allows for a precise prediction of macroscopic parameters for a relatively long stretch of freeway. Traffic flow and speed predictions are conducted for each lane individually and, for a whole segment. Per-lane predictions provide valuable information regarding different speed fluctuations in each lane for identifying congestion and applying proactive freeway controls. Predicted traffic parameters are used for tracking and predicting the spatial-temporal traffic patterns in real-time. The accuracy of the proposed model is examined and validated for various penetration rates of connected vehicles and prediction horizons and outperforms the baseline prediction methods.