Article Preview
TopIntroduction
With the trend of “online to offline” (O2O) in retail, Chinese delivery companies have begun to provide customers with alternative delivery options, such as different last-mile delivery modes and time slots. The delivery modes include attended home delivery (AHD), reception box (RB), and Collection and Delivery Points (CDPs) (Bjørnar & Stein, 2010; Wang et al., 2014).
Given that customers choose different delivery modes and time slots, the routing operation has to deal with the dynamics from customer requests and needs to be optimized with the order acceptance (Ehmke & Campbell, 2014; Xin, 2012). On the one hand, after the customer arrives and chooses the delivery mode and time slot, the delivery company could know the request location and selected delivery options. On the other hand, the delivery company needs to quickly decide whether or not to accept the request based on the RBs layout, the vehicle routing feasibility, and the constraints of the accepted orders (Bruck et al., 2018). The decisions about order acceptance and vehicle routing are dynamically interactive. The acceptance of a new order requires optimizing the vehicle routes, which will constrain the order acceptance in the next stage (Bruck et al., 2018; Florent Hernandez et al., 2017). Particularly, RB delivery could ignore the constraint of the earliest time and only needs to be served before the latest time. To deal with the dynamics inherited from sequential customer arriving and the interaction between order acceptance and vehicle routing, a multicircle order acceptance strategy with strategy-oriented algorithm would be proposed.
Although this topic is widely studied, the following questions are still open. When the customer arrives sequentially and dynamically, how can the order accept decision be made quickly based on the vehicle routing, RB location, and the service constraints of accepted order? If a delivery order is accepted, how can the order acceptance strategy be necessarily adjusted? How could the behavior of customers choosing different delivery modes and time slots be described?
Firstly, one related stream is time slots allocation strategy in Time Slots Management (TSM). It can be divided into static and dynamic strategy. For the static strategy, the time slots allocation stays unchanged (Agatz et al., 2011; F. Hernandez et al., 2017; Mackert, 2019). In contrast, concerning the dynamic one, the time slots allocation will be adjusted (Campbell & Savelsbergh, 2005; Ehmke & Campbell, 2014; Ulmer & Thomas, 2020). Most of the existing literature about TSM focuses on AHD, which is consistent with the situation that most residents of the US and EU live in independent houses. yet the TSM with different delivery modes is found in little literature.
Secondly, Multinomial Logit models were mainly used to describe customer's choice of time slots, such as Asdemir, K. (Asdemir et al., 2009). Recently, the generalized attraction model (GAM) was applied (Vinsensius et al., 2020). It is relevant in the customer choice behavior between delivery modes and time slots. If the customers choose AHD, customers may prefer that the delivery would be finished on time. The research considering the customers’ choice correlation is still relatively rare.