A Data-Driven Framework for Driving Cycle Generation and Analysis

Abstract

This paper presents a methodology for generating realistic driving cycles through a combination of Markov chain modeling, Monte Carlo simulation, and dynamic time warping. The study is focused on the construction of a representative driving cycle for the city of Iğdır in Turkey, taking into account its unique traffic characteristics. The methodology involves two main stages: first, determining reference segments partitioned from original driving datasets based on traffic conditions and road types, using the dynamic time warping technique based on the similarity between each segment time series. The second stage is to stochastically generate a representative driving cycle by employing a combination of Markov chain and Monte Carlo simulation, producing variability and randomness. In this stage, the best driving cycle segment of each segment group from among the generated driving segments utilizing Markov chain modeling and Monte Carlo simulation was selected using the dynamic time warping techniques, considering the reference segments. Finally, a representative driving cycle was constructed by stitching each segment. To assess the generated representative cycle, commonly used kinematic parameters were compared with real-world driving cycle data for Iğdır. The results show that the proposed methodology provides an advanced algorithm for generating a reasonable representative driving cycle, which can contribute to energy consumption analysis, vehicle performance, and emission evaluation. The comprehensive approach provided by the proposed methodology enables an accurate understanding of driving patterns, promoting the development of sustainable mobility solutions.