A heuristic for the inventory management of smart vending machine systems

Purpose: The purpose of this paper is to propose a heuristic for the inventory management of smart vending machine systems with product substitution under the replenishment point, order-up-to level policy and to evaluate its performance. Design/methodology/approach: The heuristic is developed on the basis of the decoupled approach. An integer linear mathematical model is built to determine the number of product storage compartments and replenishment threshold for each smart vending machine in the system and the Clarke and Wright’s savings algorithm is applied to route vehicles for inventory replenishments of smart vending machines that share the same delivery days. Computational experiments are conducted on several small-size test problems to compare the proposed heuristic with the integrated optimization mathematical model with respect to system profit. Furthermore, a sensitivity analysis is carried out on a medium-size test problem to evaluate the effect of the customer service level on system profit using a computer simulation. Findings: The results show that the proposed heuristic yielded pretty good solutions with 5.7% error rate on average compared to the optimal solutions. The proposed heuristic took about 3 CPU minutes on average in the test problems being consisted of 10 five-product smart vending machines. It was confirmed that the system profit is significantly affected by the customer service level. Journal of Industrial Engineering and Management http://dx.doi.org/10.3926/jiem.587 355 Originality/value: The inventory management of smart vending machine systems is newly treated. Product substitutions are explicitly considered in the model. The proposed heuristic is effective as well as efficient. It can be easily modified for application to various retail vending settings under a vendor-managed inventory scheme with POS system.


Introduction
Smart vending machines that enable sales and inventory information to be remotely monitored through the Internet using embedded online messaging devices were recently introduced. Webster et al. (1999) developed an experimental Java client-server system with a special purpose hardware interface to control a Pepsi vending machine via the Internet. Jofemar unveiled what it claims is the world's first autonomous hydrogen-powered smart vending machine at the 2009 National Vending Machine Exposition (Fuel Cells Bulletin, 2009). Figure 1 shows an example of a smart beverage vending machine operated at G20 Seoul Summit Conference in 2010.
Smart vending machines are expected to resolve the problems inherent to conventional vending machines such as run-out of products and machine breakdowns left for a long time. Wolson (2000) insisted that smart vending machines are the future of the vending machine industry.
The inventory management of smart vending machine systems with product substitution under the replenishment point, order-up-to level policy creates a decision making problem. The problem includes the determination of replenishment points of products at vending machines, product allocation to vending machine storage compartments, replenishment thresholds at vending machines, and vehicle routes for inventory replenishments.
Little work has addressed the inventory management of smart vending machine systems in the literature. Poon et al. (2010) developed an index to make replenishment decisions for smart vending machines to minimize transportation and stock-out costs. Park and Yoon (2011) proposed a two-phase solution based on a nonlinear integer mathematical model for the operation problem of conventional vending machine supply chains with stock-out based, onestage item substitution. Among reports regarding inventory-routing problems, some are applicable for determining shipping policies in smart vending machine systems (see Archetti et al., 2007;Li et al., 2010;Moin et al., 2011).  In this paper, we propose a heuristic for the inventory management of smart vending machine systems with product substitution under the replenishment point, order-up-to level policy and conduct computational experiments on several test problems in order to evaluate its optimality. Furthermore, a sensitivity analysis is carried out to evaluate the effect of customer service level on system profit using a computer simulation.
The remainder of this paper is organized as follows: A heuristic for the inventory management of smart vending machine systems is proposed in Section 2. Computational experiments are described in Section 3. Finally, conclusions are drawn in Section 4.

A Heuristic for the Inventory Management of Smart Vending Machine Systems
Various constraints are required. The constraints for computing the inventory levels and stockouts of products, restricting the allocation of product storage compartments, limiting the quantities of products for substitution, and restricting the domain of variables are formed by modifying the ones shown in Park and Yoon (2011).
More constraints are needed to determine the replenishment threshold at smart vending machine r, r  . Constraints (2) and (3) Since the objective function is to maximize the profit of a smart vending machine, the time span may be lengthened to the period in which the last remaining product in the vending machine is sold out, generating undesirably a large amount of lost sales. Therefore, Constraint (4) that restricts the amount of lost sales during the time span should be included.
Once the number of product storage compartments and replenishment threshold is determined for all the smart vending machines in the system, the vehicle routes to replenish the smart vending machines that share the same delivery days are constructed. For vehicle routing, the Clarke and Wright's savings algorithm (1964) is employed while minimizing the transportation cost. The saving expected by combining two routes, qr s , is computed as follows: where q denotes the last smart vending machine to visit on one route and r denotes the first one to visit on the other route.

Computational Experiments
The five-product smart vending machine locations were generated using a uniform distribution on [0, 100]. The location of a depot was set to (50, 50). The daily demand at each smart vending machine was generated from a Poisson distribution, the average of which was generated from a uniform distribution on [150,210]. The values of five products were set to 0.525, 0.875, 1.225, 0.7, and 1.05, respectively. The rates of inventory carrying cost, smart vending machine operating cost, and substitution cost were set to 0.01, 0.1, and 0.05, respectively. The vehicle capacity was set to 3200. The fixed and variable transportation costs were set to 100 and 0.4, respectively. The allowance ratio of lost sales was set to 0.3.
It was assumed that the warehouse at the business office always stores the sufficient amount of products to replenish the smart vending machines in the system. Computational experiments were carried out on an IBM compatible Pentium IV PC (2.21GHz CPU, 1GB RAM).

Comparison of the Integrated Optimization Model and the Proposed Heuristic
We evaluated the performance of the proposed heuristic by comparing it with the integrated optimization mathematical model. The optimization mathematical model determines the number of product storage compartments and replenishment thresholds at smart vending machines and the vehicles routes for daily replenishments in an integrated way while maximizing the system profit during a planning horizon. The integrated optimization model was constructed by basically combining the mathematical model presented in Section 2 and the constraints for vehicle routing proposed by Abdelmaguid and Dessouky (2006).

Sensitivity Analysis
We carried out a sensitivity analysis to evaluate the effect of the customer service level on system profit using the proposed heuristic. We formed a base problem consisting of eighty smart vending machines with 25  M and 30  Q . The operation of the smart vending machine system was modeled with Arena 10.0 (Kelton et al., 2007). Salmeron (2009) stated that simulation is one of the most commonly used methodologies employed in operations management research and listed some emerging areas to apply including system design.
During the simulation, when a vehicle faces deficiencies of some products when filling a smart vending machine, it stocks the storage compartments as much as possible and travels to the next vending machine with the remaining supply. When a vehicle still has supply after serving the last vending machine on its tour, it returns to the depot with the remainder. The daily operation of vehicles is limited to 8 hours. Table 2 shows the average daily profit per smart vending machine for four different service levels in three replications, which were obtained by simulating each level for 365 days. The highest average daily profit per smart vending machine was obtained at a 70% service level.
For a statistical confirmation of our finding, we performed a one-way analysis of variance (ANOVA) using an Excel spreadsheet, as shown in  Table 3. Test statistics of ANOVA for the service level Table 4 shows the partial solutions of the five smart vending machines randomly selected, which were obtained in the first replication with the customer service level of 0.7. Table 5 shows the monetary outputs obtained in the first replication with the customer service level of 0.7 by simulating the system for 365 days.   Table 5. System outputs obtained in the first replication with the service level of 0.7

Conclusions
We proposed a heuristic for the inventory management of smart vending machine systems with product substitution under the replenishment point, order-up-to level policy. The heuristic solves the problem in a decoupled way. First, it determines the replenishment points of products at smart vending machines. Second, it determines the number of product storage compartments and the replenishment thresholds at smart vending machines. Third, it determines the vehicle routes for daily replenishments. The proposed heuristic yielded pretty good solutions with 5.7% error rate on average compared to the optimal solutions obtained by solving the integrated mathematical model. The heuristic took about 3 CPU minutes. It was confirmed through the sensitivity analysis using a computer simulation that the system profit is significantly affected by the customer service level.
Future research is necessary to develop more efficient and effective solutions for the inventory management of smart vending machine systems. Furthermore, it is required to develop an algorithm to find the optimal allowance ratio of lost sales during the time span which maximizes the system profit.