An Iterative Markov Chain Approach for Generating Vehicle Driving Cycles

For simulation and analysis of vehicles there is a need to have a means of generating drive cycles which have properties similar to real world driving. A method is presented which uses measured vehicle speed from a number of vehicles to generate a Markov chain model. This Markov chain model is capable of generating drive cycles which match the statistics of the original data set. This Markov model is then used in an iterative fashion to generate drive cycles which match constraints imposed by the user. These constraints could include factors such number of stops, total distance, average speed, or maximum speed. In this paper, systematic analysis was done for a PHEV fleet which consists of 9 PHEVs that were instrumented using data loggers for a period of approximately two years. Statistical analysis using principal component analysis and a clustering approach was carried out for the real world velocity profiles. After dividing the real velocity profiles into segments, they were clustered into several different clusters based on statistical data. This data set is also used to generate the Markov chain model technique described above which is the central development of this work. The work of this paper is a part of a larger project in which a mass simulation of a neighborhood of PHEVs will be conducted based on statistical representations of key factors such as vehicle usage patterns, vehicle characteristics, and market penetration of PHEVs. This approach can be used as critical input of the large scale simulation model to generate random velocity profiles of various driving patterns.