面向汽车运行工况设计的马氏链非等长交叉进化算法

为解决马尔科夫链和传统遗传算法设计汽车运行工况时效率低、质量差的问题，提出用于设计汽车运行工况的马氏链非等长交叉进化方法.设计子代满足马尔科夫链转移关系的非等长交叉算子，解除等位等长交叉段的限制，使遗传算法更好地适用于汽车运行工况的设计.根据试验数据，应用马氏链非等长交叉进化方法构建非等长初始种群，使用满意准则模型和指数加权平均数设定目标函数，设计三参数汽车运行工况.随机生成3种不同长度的三参数高速公路代表性工况.分析结果表明，期望运行工况与原始数据库特征参数的相对偏差均在设定范围内，速度和加速联合分布相关系数均高于90%，生成工况具有代表性.相比于马尔科夫链和传统遗传算法相结合的设计方法，马氏链非等长交叉进化方法的平均运行工况生成效率提高了66%，运行工况质量更优.

Non-isometric crossover evolution algorithm of Markov chain for designing vehicle driving cycles

A non-isometric crossover evolution algorithm for designing vehicle driving cycles with Markov property was proposed, in order to solve the low efficiency and poor quality caused by the Markov Chain and typical genetic algorithm when designing vehicle driving cycles. Individuals were designed to satisfy the Markov property by exchanging with non-isometric cross segments; a new crossover was designed based on the genetic algorithm, which broke the restriction of allelic crossover segments and was applicable for designing driving cycles. According to a collected highway database, the method was used to design three-parameter highway driving cycles, which included constructing non-isometric initial populations and designing an objective function by using the satisfaction rule model and exponential weighted average. Three kinds of three-parameter representative driving cycles with different lengths were generated. Results show that the relative deviations of indices between the desired cycles and the original database are within a reasonable range and correlation coefficients of velocity and acceleration joint probability distribution are above 90%, which indicates the representativeness of the generated driving cycles. Compared with the results of the method which combines Markov Chain with typical genetic algorithm, the average design efficiency of the new method increases by 66%, and the design quality of driving cycles is better.