The Study on Multi-scale Prediction of Future Driving Cycle Based on Markov Chain

Abstract

In this paper, a multi-scale single-step future driving cycle prediction method is proposed. The driving cycle is discretized at given time scale, and the mesh accuracy of velocity and acceleration is determined to be 1Km/h and 0.05m/s² respectively, then Markov state transfer matrix can be obtained by categorizing these discretization points and statistical calculation. After that, future driving cycle prediction can be accomplished by proposed method which combines Markov chain and Monte Carlo method. Finally, Root-Mean-Square Deviation is introduced to assess the prediction accuracy. Comparing the prediction accuracy of multi-scale single-step method and traditional fix-scale multi-step method, it can be found that the prediction results of proposed method reach expectant accuracy improving by 7.18% on average.