An auction-enabled collaborative routing mechanism for omnichannel on-demand logistics through transshipment

Highlights

• An in-depth industrial investigation on dynamic omnichannel on-demand logistics.

• A two-stage auction mechanism designed for dynamic on-demand tasks’ (re)allocation.

• Transhipment-based task generation mechanism identifying the most uneconomic leg.

• Transhipment-based dynamic pickup and delivery routing mechanism.

• Computational study demonstrates the proposed model is effective to bring benefits.

Abstract

Nowadays’ on-demand logistics operations are carried out on separate delivery networks from competitive contractors, with redundant resources and high costs. This article proposes a new paradigm to deal with industrial and societal challenges, developing an auction-enabled collaborative routing mechanism for omnichannel on-demand logistics in a real-time transshipment network. We consider an online service platform for real-time management of on-demand pickup and delivery tasks, where multiple freight shippers can trade with multiple freight carriers. Freight shippers are retailers or individual customers, while freight carriers are a group of logistics service providers. The platform acts as the auctioneer. A two-stage combinational auction mechanism is designed for dynamic on-demand task (re)allocation. Tactical-level auctioning and operational-level routing decisions are optimized together. The transshipment-based task generation method is used to identify uneconomic paths from carrier’s real-time network for outsourcing. A transshipment-based routing algorithm is developed to enable each carrier to make decentralized decisions for network reconstruction and transportation bidding. Our model aims to improve the overall social welfare while bringing benefits to stakeholders involved. The computational results have shown positive society impacts. Specifically, shippers’ payments can be saved while carriers’ profits are increased compared with other operative models which have been investigated in previous research studies or industry. In addition, a substantial reduction in CO₂ emissions and vehicles required can be achieved. The main reason for the improvement in social welfare is due to the optimal network achieved through collaboration. We also numerically analyze the impacts of three key factors: growth in demand density, urgency of tasks and flexible auction interval.

Introduction

Global retail ecommerce sales are expected to increase to US$ 6.5 trillion by the end of 2023 (eMarketer, 2019). The rapid growth of this market leads to on-demand logistics in which delivery orders from both online or offline channels are immediately fulfilled with customer desired time, place and quantities (Hong et al., 2019). Increased service levels expressed in hours to deliver, is also rigidly required. Currently, most on-demand logistics operations are carried out on separate delivery networks from different professional or crowdsourced delivery contractors, with their own dedicated facilities and fleets (Bányai et al., 2018). Many companies face difficulties controlling omnichannel order fulfillment and reducing on-demand delivery costs. Further, although a set of accessible pickup locations (e.g., smart lockers and community stores) exist for every customer, these depots are not employed for cargo transshipment and consolidation.

This article investigates a new paradigm defined as an auction-enabled collaborative routing mechanism with transshipment (ACRT) for omnichannel on-demand logistics through resource sharing. We consider a service platform for the real-time management of on-demand pickup and delivery tasks, where multiple task shippers can trade with multiple task carriers. Task shippers are retailers in the omnichannel ecosystem who could be any grocery, shopping mall, pharmacy or flower shop. Individual customers can also be a shipper since the on-demand delivery of consumer-to-consumer direct shipments are widely operated within cities. Task carriers refer to a group of logistics service providers who own sufficient resources such as vehicles and drivers to complete on-demand deliveries. Effective transportation planning is able to improve social welfare including all stakeholders’ benefits or costs (Rouhani et al., 2016). In this article, the proposed ACRT aims to bring about positive social welfare with novel optimization mechanisms and technologies for shippers, carriers and city residents. Social welfare improvements involve three dimensions (i) shippers: payment for logistics services; (ii) carriers: transportation profit; (iii) city residents: reduction in greenhouse emissions.

Fig. 1 illustrates the differences between the conventional mode (a) and the transshipment-based collaborative mode (b) for omnichannel on-demand logistics. In the conventional delivery mode, shippers select carriers from the transportation market randomly or based on experience. Service pricing is usually fixed within a certain delivery area. Carriers process tasks separately, they are unable to cope with lumpy demands during peak hours but become under-utilized during off-peak periods. Carriers are self-interested and independent to each other while making their own decisions. In the proposed ACRT, shippers can purchase transport services more economically from an auction-enabled on-demand service platform. Carriers can outsource tasks through an arbitrary open facility to optimize delivery plans with reduced costs. Parcel lockers or community stores near the customers’ house can be utilized for transshipment.

It has been demonstrated that auction mechanisms provide a very effective way for collaboration in logistics (Karaenke et al., 2019). Several studies have addressed auction approaches in decentralized transportation optimization with carriers’ collaboration. Existing models are either focused on task allocation from shippers to carriers (Andres Figliozzi et al., 2003, Figliozzi et al., 2004, Figliozzi et al., 2005, Mes et al., 2007, Mes et al., 2013) or on shared task reallocation among carriers (Berger and Bierwirth, 2010, Dai and Chen, 2011, Dai et al., 2014, Gansterer and Hartl, 2016, Lyu et al., 2019). The problems for freight task sharing were examined in a static context where all tasks were revealed before planning. Social welfare for shippers and residents is scarcely considered. Here, we incorporate the allocation of new tasks and shared tasks reallocation through dynamic on-line auctions. To the best of our knowledge, there are limited studies considering combined decision-making for multiple stakeholders (including both shippers and carriers) with multiple objectives for on-demand logistics.

To enable more flexible trading and sharing in the carrier coalitions context, real-time transshipment is integrated into the ACRT paradigm. Dynamic tasks are allowed to be executed by multiple carriers. The advantage of transshipment-based collaborative routing is illustrated in Fig. 2. In a non-transshipment network, the paired pickup and delivery node of task 2 of one seller carrier have to be outsourced together to another actual carrier (Fig. 2a). In contrast, only the delivery node of task 2 is outsourced in the transshipment-based network (transshipped via the pickup node of task 3), instead of outsourcing the entire task 2 (Fig. 2b).

The routing component of the ACRT is a modified version of the dynamic pickup and delivery problem with transshipment and time windows (DPDPTTW). There has been a limited amount of research that has focused on the DPDPTTW in the transportation science and operations research literature. Thangiah et al. (2007) adapted the P&D heuristic of Shang and Cuff (1996) in examining DPDPTTW applications, however, the P&D heuristic in Shang and Cuff (1996) was designed to solve the static PDPTTW. In addition, the detailed algorithm for the design of transshipment insertion is missing in both studies. The above gaps constitute one of the objectives of this article.

In particular, this article aims to answer the following questions:

• What is an effective auction mechanism for the optimal procurement of on-demand delivery services as well as carrier collaborations in the ACRT paradigm?

• What is a dynamic task generation strategy for determining the most uneconomic path from a carrier’s network?

• How can the dynamic transshipment-based routing problem with ACRT requirements be solved?

• What are the impacts of key factors (e.g., demand density growth, urgency of tasks and auction interval) on individual, and total logistics network performance?

To cope with the complex ACRT problem, three mechanisms are designed including, (i) a two-stage auction mechanism (TAM), (ii) a transshipment-based task generation mechanism (TTGM), and (iii) a transshipment-based routing mechanism (TRM) (see Section 4). Compared with the conventional mode, the ACRT is evaluated via simulated omnichannel scenarios like business-to-customer (B2C) fulfillment, customer-to-customer (C2C) fulfillment and hybrid fulfillment of B2C and C2C. We further analyze the impact of key factors on the performance of the ACRT to investigate the practical significance of the proposed solution (see Section 5). A list of abbreviations used in this article can be found in Appendix A.

This article is organized as follows. In Section 2, we review the relevant literature. We describe the new paradigm of the ACRT for omnichannel on-demand logistics in Section 3. In Section 4, integrated methods of the ACRT involving TAM, TTGM and TRM are proposed. We present a computational study in Section 5 to quantitatively assess the solution feasibility with associated key impact analyses. Our conclusions and future research directions are given in Section 6.