Synthesis of real-world driving cycles using stochastic process and statistical methodology

This paper proposes a procedure for synthesising real-world driving cycles to reproduce naturalistic driving patterns for arbitrary driving distances. The procedure combines stochastic processes and statistical methodologies. Vehicle dynamics equations are investigated and two states, velocity and acceleration, are selected to represent real-world driving behaviour with the Markov chain. Then, the information is extracted from the naturalistic driving data measured in Southeast Michigan in a form of transition probability matrices (TPMs). Statistical methods are utilised to guarantee the representativeness of synthesised cycles. Results demonstrate the ability to capture features of a whole category of naturalistic data with a single synthetic cycle.