关键词: 用户满意度, 非对称软时间窗, 路径规划, 柯西变异, 遗传算法

Abstract: In the context of the booming development of Internet technology, people's lifestyle is constantly changing, and the functional demand of the Internet is also increasingly emphasized. As an important livelihood issue closely related to people's life and health, the combination of medical care and information technology is very necessary. During the rapid development of online retail industry, it coincides with the government departments, continuous introduction of policies to encourage medical institutions to use the Internet to provide more convenient and efficient diagnosis and treatment services for the public, and the online medical form emerges at the historic moment. It not only has basic medical services such as online consultation and online prescription, but also provides users with drug distribution services. Compared with other commodities, drugs have the particularity of strong rigid demand and complex categories, which lead to higher requirements for timeliness and safety in the process of drug distribution. Therefore, in the process of drug distribution, how to comprehensively consider the characteristics of drug distribution, rationally plan the distribution path, so as to improve the distribution efficiency and achieve the optimal cost has become an urgent problem to be solved in drug distribution.
Firstly, based on the fact that the user's satisfaction with the delivery service will affect the delivery cost in the process of drug delivery, and the drug distribution adopts the principle of non-contact delivery, the rider will not cause losses to the user when the order is delivered earlier than the expected time window of the delivery center, that is, the penalty coefficient before the earliest expected delivery time of the system is 0. Combined with the waiting cost caused by the lack of inventory in chain pharmacies and the delay in taking medicine caused by other merchants, a PDVRP model with asymmetric soft time window considering customer satisfaction is established. The total cost of delivery in the model is composed of three parts: transportation cost, waiting cost and penalty cost, and the path optimization is carried out with the goal of minimizing the total cost.
Secondly, considering the NP-hard characteristics of the delivery path problem with time window, in order to improve the problem of insufficient local search ability and slow convergence speed of the genetic algorithm, an improved genetic algorithm based on Cauchy mutation strategy is designed. The chromosome representing the delivery route of the rider is constructed by integer coding, the reciprocal of the objective function is taken as the fitness function, and the ranking-based roulette selection is selected to eliminate the huge difference of the fitness value and make the advantages of the better individuals more obvious. A new adaptive forward continuous crossover operator is designed to retain its own excellent genes and avoid code loss, and the Cauchy mutation strategy is integrated to jump out of the local optimum and enhance the global search ability of the algorithm.
Finally, in order to verify the effectiveness of the model, two different test sets are generated based on the order-related data of a platform given in the reference. The model is solved by simulation under the two different test sets, and the rider order allocation, path arrangement and cost distribution under the optimal delivery path are obtained. Then the model is solved by the traditional genetic algorithm and the improved genetic algorithm without the fusion of Cauchy mutation, and the evolution curves of the three algorithms and the cost structure under the optimal solution of each algorithm are compared. The results show that the convergence speed and the optimal solution of the improved genetic algorithm with the fusion of Cauchy mutation are better than the other two algorithms. Therefore, the effectiveness and superiority of the model and algorithm in this paper are verified.
In summary, the research content of this paper improves the problems of previous studies, such that few studies comprehensively have considered the particularity of drug distribution and customer satisfaction, but most have ignored the benefits of drug distribution efficiency, resulting in excessive penalty cost and too ideal inventory estimate for chain pharmacies. An improved genetic algorithm combined with Cauchy mutation, which can improve the convergence speed of the algorithm and jump out of the local optimum, is proposed for the optimization scheme of drug distribution path planning for chain pharmacies, which has certain reference significance for future drug distribution research. In further research, the dynamic changes of order quantity and actual road conditions can be introduced into the cost model, and various complex problems such as the remaining life of distribution vehicles can be considered to expand the application scenarios of the algorithm.

Key words: user satisfaction, asymmetric soft time window, path planning, Cauchy variation, genetic algorithm