工业物联网的工业边缘云部署算法  

Industrial edge cloud deployment algorithm for industrial internet of things

作　　者：颜晓莲 章刚 邱晓红[1] 陈庆奎[3] YAN Xiaolian;ZHANG Gang;QIU Xiaohong;CHEN Qingkui(School of Software Engineering,Nanchang Campus of Jiangxi University of Science and Technology, Nanchang 330013, China;Peking University Science Park of Jiangxi Province, Nanchang 330033, China;School of Optical-Electrical and Computer Engineering, University of Shanghai for Sciences and Technology, Shanghai 200093, China)

机构地区：[1]江西理工大学南昌校区软件工程学院,江西南昌330013 [2]江西北大科技园,江西南昌300033 [3]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机集成制造系统》2022年第2期574-583,共10页Computer Integrated Manufacturing Systems

基　　金：江西省教育厅科技资助项目(GJJ170571);国家自然科学基金资助项目(61572325)。

摘　　要：针对动态调整不同区域的生产线时,因工业边缘云资源有限,覆盖生产线的工业边缘云部署不合理而造成实时性运维服务质量下降和企业成本增加等问题,采用带约束的多目标优化和带约束的最小子集划分思想讨论工业边缘云部署问题,提出一种启发式遗传算法。基于问题的特点,该算法采用二进制编码,降低了算法实现的难度;采用多轮随机不重复解策略筛选多样化的可行解作为初始种群,提高了搜索速度和搜优概率;根据混合选择法有目的地选择较优个体和较差个体,从而保持种群多样性;采用多轮多维度多点交叉法实现个体间较优与较优、较优与较差、较差与较差的深度交叉,维持了种群多样性,并探索了新区域;利用较优个体优先单点变异策略,对交叉操作产生的较优新个体所在区域优先进行局部深挖,在深挖过程中不断调整深挖方向,从而拓展种群多样性,提升全局搜索能力。实验从期望负载偏差率、期望服务延时偏差率、算法收敛率及解误差率4方面验证了算法的有效性、收敛性和全局搜索能力。If the industrial edge cloud deployment covering the production line is unreasonable,it is easy to cause problems such as the decline of real-time operation and maintenance service quality and the increase of enterprise costs under the dynamic adjustment of production lines in different regions and the limited cloud resources.By using the idea of constrained multi-objective programming and constrained minimum subset partition,the industrial edge cloud deployment problem was discussed,and a heuristic genetic algorithm was proposed.Based on the characteristics of the problem,the binary coding was adopted to reduce the difficulty of algorithm implementation.The multi-round random non-repetitive solution strategy was used to select the feasible solutions as the initial population,so as to improve the search speed and search probability.According to the mixed selection method,the better individuals and the worse individuals were selected purposefully to maintain the diversity of the population.The method of multi-round,multi-dimension and multi-point crossover was adopted to realize the deep crossover of better and better individuals,better and worse individuals,worse and worse individuals,to maintain population diversity and to explore new areas.The preferred individual preference single-point mutation strategy was adopted to carry out local deep excavation in the region where the preferred new individuals generated by cross-operation were located,and the direction of deep excavation was adjusted continuously in the process of deep excavation,so as to maintain the diversity of population and enhance the global search ability.The validity,convergence and global search ability of the algorithm were verified from expectation load deviation rate,expectation service delay deviation rate,algorithm convergence rate and solutions error rate.

关 键 词：工业物联网 边缘云部署 NP难问题 遗传算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]