Synthesis of naturalistic vehicle driving cycles using the Markov Chain Monte Carlo method

Contemporary methods of designing and testing mechanical vehicles are based on simulation techniques that require the use of precise models of vertical and horizontal dynamics and sequences of random events occurring in road traffic conditions. This problem is relevant to the optimization of vehicles with internal combustion engines, electric vehicles (EV) and hybrid electric vehicles (HEV). To select the most appropriate drive system architecture for a particular vehicle class and driving cycle, it is necessary to optimize the size of components according to their cost functions, such as the lowest 2 CO emissions, the lowest weight, fuel savings or any combination of these attributes in the architecture [1, 7, 9 and 18]. Regardless of the simulation technique used: quasi-static using a “Backward-facing” vehicle model or a dynamic simulation with a “Forward-facing” model, understanding of the representative driving cycle is essential. In the first case, for an open-loop system, the time series of speed is imposed on the input of the vehicle model in order to calculate rpm and torque on the wheels. In a closed-loop vehicle model, on the other hand, the driving cycle is a setpoint for the driver block, which generates a suitable engine torque. The time and cost constraints associated with the design and testing of various possible vehicle architectures require methods of driving cycle synthesis that can meet the modelling and simulation requirements of automotive engineers throughout the R&D process. It is not possible to optimize the parameters and gradually increase the autonomy of the vehicles based on standard driving cycles, and such optimization cannot prevent “cycle beating”. To ensure that the synthesized time series based on the collected databases are representative, it is necessary to use algorithms adopting techniques based on stochastic and statistical models [6, 19]. To define the equivalence criteria, the synthesis process is concluded with a verification of the results, i.e. each generated cycle, through statistical analysis in the time or frequency domain. A combination of multiple criteria is frequently used [2, 4]. The methods of driving cycles construction require quantization of traffic parameters. Depending on their function (emissions estimation, fuel consumption estimation or traffic engineering, etc.), the defined states can be synthesized for micro-trips, segments, heterogeneous classes or modal cycles [17]. Micro-trips are driving models between stops including periods of inactivity. Traffic signals and overloads contribute to “stop-go” driving patterns, and result in increased fuel PuchAlski A, komorskA i, ŚlęzAk m, NiewczAs A. synthesis of naturalistic vehicle driving cycles using the markov chain monte carlo method. eksploatacja i Niezawodnosc – maintenance and reliability 2020; 22 (2): 316–322, http://dx.doi.org/10.17531/ein.2020.2.14.