Self-Adaptive Hyper-Heuristic Markov Chain Evolution for Generating Vehicle Multi-Parameter Driving Cycles
Man Zhang, Shuming Shi, Wendong Cheng, Yunbo Shen
Abstract
Representative driving cycles are very important for testing energy consumption and pollutant emission, optimizing control strategy, and designing power resource components of vehicles. Multi-parameter representative driving cycles are needed to improve vehicles' adaptability to the driving environment. Thus, high-dimensional driving cycles need to be efficiently generated. Additionally, the generation method should be flexible enough to facilitate use by automobile engineers. This study introduces a hyper-heuristic framework into Markov chain evolution (MCE). A boundary variable is introduced to refine strategies, and multiple evolution strategies are proposed for self-adaptivity. Then an evaluation function based on the desired driving cycles is designed using allocation and update mechanisms. Finally, an efficient framework is established for generating multi-parameter driving cycles. As an example, the generation efficiency of this framework is increased by 63.25% over the standard MCE method through collecting real-world driving data and considering representative driving cycles with three parameters (velocity, acceleration, and road slope). Analyzing the proportions of hyper-heuristic evolution strategies indicates the self-adaptivity of this method. Compared with an adaptive MCE method with two strategies, the proposed method has greater running efficiency and application flexibility.