Abstract: Embodiments described herein improve fuel economy by controlling a vehicle powertrain based on a predicted vehicle velocity. The vehicle velocity is predicted based on vehicle-to-vehicle data when a prediction horizon is a longer prediction horizon and the vehicle velocity is predicted based on historical drive cycle data when the prediction horizon is a shorter prediction horizon. A time duration of the shorter prediction horizon is shorter than the time duration of the longer prediction horizon. A plurality of drive cycles are established for both the longer and the shorter prediction horizons using a neural network. A shorter prediction horizon drive cycle uses nonlinear autoregressive exogenous model neural networks and the longer prediction horizon drive cycle uses two layer feedforward neural networks. The predicted vehicle velocity is determined from a similar drive cycle of the plurality of drive cycles of either the shorter and/or the longer prediction horizon drive cycles.

Abstract: Embodiments described herein improve fuel economy by controlling a vehicle powertrain based on a predicted vehicle velocity. The vehicle velocity is predicted based on vehicle-to-vehicle data when a prediction horizon is a longer prediction horizon and the vehicle velocity is predicted based on historical drive cycle data when the prediction horizon is a shorter prediction horizon. A time duration of the shorter prediction horizon is shorter than the time duration of the longer prediction horizon. A plurality of drive cycles are established for both the longer and the shorter prediction horizons using a neural network. A shorter prediction horizon drive cycle uses nonlinear autoregressive exogenous model neural networks and the longer prediction horizon drive cycle uses two layer feedforward neural networks. The predicted vehicle velocity is determined from a similar drive cycle of the plurality of drive cycles of either the shorter and/or the longer prediction horizon drive cycles.