Short-term Speed Forecasting Using Vehicle Wireless Communications

Forecasting the speed trajectories of driving vehicles is essential for vehicle/powertrain predictive optimal control. This paper proposes a simple and effective forecasting method for generating short-term future speed trajectories using vehicle-to-vehicle (V2V) information. Specifically, a series of lead vehicles' speeds and locations are considered to be the potential trajectories that the following car would drive in the near future. Polynomial regression based on weighted least-squares estimation is used to determine a future speed trajectory over a short prediction horizon. The efficacy of the proposed approach is evaluated in single-lane traffic simulations over various driving scenarios. In addition, the performance of the proposed method is also evaluated when V2V is not available. Simulation results show that for a highway drive cycle, the proposed predictor results in root-mean-square errors less than 1 mph with V2V data.