مدلی برای بازیابی اختلالات ترکیبی در تحویل لجستیک : در نظر گرفتن شرکت کنندگان در دنیای واقعی

The existence of uncertainties may result in various unexpected disruption events in logistics delivery, which often makes actual delivery operations deviate from intended plans. The purpose of the paper is to develop a combinational disruption recovery model for vehicle routing problem with time windows (VRPTW), trying to handle a variety and a combination of delivery disruption events. Firstly, a novel approach to measure new-adding customer disruption, which considers the real-world participators (mainly including customers, drivers and logistics providers) in VRPTW, is developed. Then the paper proposes methods of transforming various delivery disruptions into the new-adding customer disruption, and determines the optimal starting times of delivery vehicles from the depot to provide a new rescue strategy (called starting later policy) for disrupted VRPTW. Based on the above, a combinational disruption recovery model for VRPTW is constructed and nested partition method (NPM) is designed to solve the proposed model. Finally, computational results are reported and compared with those of previous works, which verifies the effectiveness of the proposed solution and draws some interesting conclusions.

With the rapid development of e-commerce and mobile commerce, logistics delivery activities have become increasingly important in economic development and daily life. The general process of once delivery activity is: (i) customers' requests are asked; (ii) logistics providers schedule delivery plans; (iii) drivers (delivery staff) travel according to the planed routing and serve customers as they expect. It seems easy to complete the process, especially with the help of advanced technologies nowadays! However, in the real world there are various unexpected disruption events encountered in the delivery process such as vehicles breakdown, cargos damage, changes of customers' requests including service time windows, delivery addresses and demand amount, and so on (Bertsimas and van Ryzin, 1991, Wang et al., 2009a, Li et al., 2009b, Berbeglia et al., 2010, Yeo and Yuan, 2012 and Uskonen and Tenhiälä, 2012). These disruption events often make actual delivery operations deviate from intended plans, which may bring different disturbances on the participators (such as customers, drivers and providers) in logistics delivery. Existing literatures, which we will review later in the following, have put forward effective solutions for the disrupted vehicle routing problem (VRP) with a certain disruption event, but most of the proposed models and algorithms can deal with only a certain type of uncertainty. It is not easy or impossible for each proposed solution to solve actual disrupted VRP with the reality that various disruption events (vehicles breakdown, cargos damage, and changes of customers' service time, delivery addresses, demand amount and so on) often occur successively or even simultaneously. An example is used to illustrate the combinational disruption of vehicle routing problem in Fig. 1. Fig. 1(a) shows the original routing where three vehicles serve seven customers: vehicle 1 serving customers 1 and 2, vehicle 2 serving customers 3 and 4, vehicle 3 serving customers 5, 6 and 7. Fig.1(b) shows several possible disruption events: (i) vehicle 1 breaks down when traveling to customer 2 after serving customer 1; (ii) the delivery address of customer 4 changes when vehicle 2 is serving customer 3; (iii) customer 7 decreases the demand when vehicle 3 is in the way; and (iv) there are three new-adding customers 8, 9 and 10. These disruption events may occur successively or even simultaneously, especially when the number of customers is large. One purpose of this study is to develop a common disruption recovery model for vehicle routing problem with time windows (VRPTW) which may handle a variety and a combination of disruption events. Moreover, most existing researches for disrupted VRP focus on producing new routings with the minimum costs, ignoring the real-world participators in logistics delivery (mainly involving customers, drivers and logistics providers). For some disruption event, there may be several recovery alternatives, but different participators may prefer different alternatives. In Fig. 1(b), the pink and red dotted routings represent some recovery alternatives: (i) for customer 2 who cannot be served on time because of the breakdown of vehicle 1, the pink routing which takes less waiting time is better than the red routing, but the logistics provider may prefer the red routing which can pick up the cargos in vehicle 1 to serve customer 2; (ii) customer 4 who changed the delivery address may like the pink routing better, but the provider would like to send vehicle 2 to the new address of customer 4 after serving customer 3, which need not dispatch more vehicles; (iii) for the new-adding customers 8, 9 and 10, the provider may want to dispatch vehicle 3 to serve them if there are enough cargos in the vehicle, but the driver of vehicle 3 may complain about this because he or she is too tired now; and so on. It is important to consider the satisfaction of customers and drivers (staff) which has important effects on the long-term development of delivery firms, as well as delivery costs (Sessomboon et al., 1998, Jozefowiez et al., 2008, Wang et al., 2009b, Ding et al., 2010 and de Haan et al., 2012). We try applying the thought of Disruption Management to produce solutions which are satisfactory to the above three participators, which is the second purpose of the study. To sum up, the major contributions of this paper include: (i) proposing a novel approach to measure new-adding customer disruption quantificationally, considering the real-world participators in logistics delivery; (ii) designing methods of transforming different disruption events into the new-adding customer disruption; (iii) developing a combinational disruption recovery model for VRPTW and its nested partitions method. The paper is organized as follows. Section 2 reviews some related literatures. A new approach to measure the new-adding customer disruption, which considers the real-world participators in VRPTW, is developed in Section 3. Section 4 transforms different disruption events into the new-adding customer disruption, determines vehicles' optimal starting times from the depot, and constructs a combinational disruption recovery model for disrupted VRP. Section 5 designs the nested partitions method for the proposed recovery model. In Section 6, computational experiments are demonstrated to verify the effectiveness of the model and the algorithm. Lastly, conclusions are drawn, with recommendations in future works.

There are various unexpected disruption events encountered in the delivery process. These disruptions often make actual delivery operations deviate from intended plans, which may bring different disturbances on the real-world participators (customers, drivers and providers) in logistics delivery. Although the rescheduling method can produce relatively economic rescue plans for disruption events of vehicle routing problems, it may bring great disturbance to the whole delivery system. Disruption Management provides a good idea to minimize the negative effects on the participators in logistics delivery. Most existing researches for disrupted VRP mainly focus on producing new routings with the minimum costs, ignoring the real-world participators. For delivery disruption events, there may be several recovery alternatives, but different participators may prefer different alternatives. It is important to consider the satisfaction of the participators which has important effects on the long-term development of delivery firms when working out recovery plans for disrupted VRP. We measure the new-adding customer disruption from aspects of customers, drivers and logistics providers, and applies the thought of Disruption Management to produce solutions which are satisfactory to the three participators. For the reality that a variety of delivery disruptions often occur successively or simultaneously, we propose methods of transforming various disruption events into new-adding customer disruption, which facilitates to develop a common and combinational VRPTW disruption recovery model. Considering vehicles' optimal starting times from the central depot can not only reduce the waiting costs of in-transit vehicles but also provide a new rescue strategy for the disrupted VRP. Nested Partitions Method (NPM) is used to solve the proposed recovery model. We focus on various customer disruption events in computational experiments, giving no consideration to vehicle disruption events and cargo disruption events which need further efforts. One of next works is to develop more effective and efficient multi-objective optimization algorithms for the proposed combinational disruption recovery model, which will be helpful to developing a decision support system for various disruption events of delivery practices.