Vehicle Velocity Prediction and Energy Management Strategy Part 1: Deterministic and Stochastic Vehicle Velocity Prediction Using Machine Learning 2019-01-1051

There is a pressing need to develop accurate and robust approaches for predicting vehicle speed to enhance fuel economy/energy efficiency, drivability and safety of automotive vehicles. This paper details outcomes of research into various methods for the prediction of vehicle velocity. The focus is on short-term predictions over 1 to 10 second prediction horizon. Such short-term predictions can be integrated into a hybrid electric vehicle energy management strategy and have the potential to improve HEV energy efficiency. Several deterministic and stochastic models are considered in this paper for prediction of future vehicle velocity. Deterministic models include an Auto-Regressive Moving Average (ARMA) model, a Nonlinear Auto-Regressive with eXternal input (NARX) shallow neural network and a Long Short-Term Memory (LSTM) deep neural network. Stochastic models include a Markov Chain (MC) model and a Conditional Linear Gaussian (CLG) model. To derive the prediction models, numerous data inputs are used, including internal vehicle data (CAN bus information) and external vehicle data (radar and V2I information). Two data sets representative of real world driving in Ann Arbor, Michigan are used for model development and validation. One of these data sets reflects highway-focused single-car driving, and the other one is representative of a mixed highway/urban three-car connected driving. Time shift, a novel index which reflects the time lag between predicted and actual vehicle speed values, is introduced to assess the prediction accuracy of vehicle velocity. Also, a more standard Mean Absolute Error (MAE) metric is used to evaluate the prediction results. In order to improve the vehicle speed prediction accuracy, data augmentation with additional labels is used to cue machine learners on different features present in the driving trajectories. The results show that these artificially augmented labels can significantly improve the prediction accuracy both in terms of MAE and time shift metrics. The results also indicate that deterministic models can provide more accurate performance on average while stochastic models may be less accurate in terms of the average velocity prediction but provide information on the prediction error distribution which can be exploited in stochastic and scenario based model predictive control. Overall, LSTM deep neural networks have been able to achieve the best accuracy in predicting vehicle velocity. For 10 sec ahead vehicle velocity prediction, the LSTM model demonstrates prediction accuracy with MAE of about 1 m/s and time shift of 0 to 4 seconds. The results also suggest substantial benefits in using external vehicle data, indicative of the current traffic situation, for vehicle speed prediction.