Systems

Engineering
Procedia

Available online at www.sciencedirect.com

Systems Engineering Procedia 00 (2011) 000000

Systems Engineering Procedia 4 (2012) 298 304

www.elsevier.com/locate/procedia

The 2nd International Conference on Complexity Science & Information Engineering

Agent-based Simulation Model of Single Point Inventory

System
DONG Fu-gui a*, LIU Hui-mei, LU Bing-de
North China Electric Power University,No.2 Beinong Rd., Huilongguan, Beijing, 102206, China

Abstract
Maintaining normal stock amount can reduce ordering cost and improve service level, but excessive stock needs
expensive inventory holding costs and occupies too much floating capital, so it is necessary to seek a balance between
the stock holdings and inventory cost. Using AnyLogic software, the single point inventory system simulation model
is built based on Agent method in this paper. Through comparing the two continuous replenishment strategies, the (R,
S) and (Q, R) strategies, the simulation results show that (R, S) strategy is better than (Q, R) strategy. Then the
optimal inventory policy to minimize inventory cost with a certain service level is analyzed by the optimization
experiment.

Elsevier Ltd.
Ltd.Selection
Selectionand
andpeer-review
peer-reviewunder
underresponsibility
responsibility
Desheng
Dash
2011
2011 Published
Published by
by Elsevier
of of
Desheng
Dash
WuWu.
Keywords: Inventory system; System simulation; (R, S) polic; (Q, R) polic; AnyLogic software

1. Introduction
Inventory is the term used for the future, temporarily idle resources, which can be used to maintain
normal operation, reduce order costs and improve service level while the supply and demand changes
constantly. Too much stock needs expensive inventory holding costs and occupies too much floating
capital, so inventory control in logistics engineering is more important than other logistics activities and it
is necessary to seek a balance between the stock holdings and inventory cost to minimize the total
inventory cost with a certain service level or maximize demand service level with an acceptable cost.
At present, research works about inventory system simulation are very popular. Many papers focus
very much on the impact of information coordination and information-sharing on inventory control in
supply chain, the bullwhip effect, and the optimal inventory strategy, etc. There are four major methods to
model the complex system: Agent-based Modeling and Simulation, System Dynamics model, Petri Net
and Object-Oriented Technology. Paper [1] researched on the information coordination in the activities of

* DONG Fu-gui. Tel.: +00861051963562; fax:+00861080796904.

E-mail address: dfg02@163.com

2211-3819 2011 Published by Elsevier Ltd. Selection and peer-review under responsibility of Desheng Dash Wu.
doi:10.1016/j.sepro.2011.11.079

DONG Fu-gui et al. / Systems Engineering Procedia 4 (2012) 298 304

decision making of production planning and procurement planning in order to fulfill process using the
means of multi-agent based modeling and simulation. Papers[2,3] developed a basic model in a
production-inventory control system using difference equations to discuss the relationship between
stability and bullwhip effect in the supply chain system. Paper[4]studied the integrated stochastic
inventory problem for a two-stage supply chain consisting of a single retailer and a single supplier. By
using batch shipment policy, the paper obtained the result that the expected total cost can be significantly
reduced. Paper[5] investigated the effect of product perishability and retailers stockout policy on system
total cost, net profit, service level, and average inventory level in a two-echelon inventorydistribution
system. In addition, the paper developed an approximate inventory model to the system performance
measures. Paper[6] showed how to model a problem to find optimal number of replenishments in the
fixed-order quantity system as a basic problem of optimal control of the discrete system. Paper[7] studied
the inventory system of an online retailer with compound Poisson demand. The retailer normally
replenishes its inventory according to a continuous review (Q, R) policy with a constant lead time.
Multi-location inventory models are one of the most widely investigated fields in mathematical
inventory theory, but the analytically tractable models suffer from various restrictive assumptions. To
overcome these restrictions simulation can be used[8].
Paper [9] described a simulation model of a multi-echelon inventory cost optimization problems
using Arena and OptQuest to gain the reorder point and order size of the retailers.
The above literature shows that great progress has been made in research on inventory system in
supply chain. This paper describes the single point inventory system simulation model using AnyLogic
software and adds comparison of two continuous inventory policies based on paper [9].
2. Agent-based Simulation Modeling
Agent is a unit of model design that have behavior, memory, timing, contacts, etc. Agents can
represents people, companies, projects, assets, vehicles, cities, animals, ships, products, etc.
The agent-based simulation modeling broke the traditional top-down study mode. It starts at basic
unit of the model to realize independent interaction and to imitate the behavior of the complex systems
through the effective communication between the Agents. It is an effective method of top-down analysis
and bottom-up synthesize.
2.1. Description and Hypothesis
Inventory system simulation model mainly imitates warehouse, but it also contains suppliers and
customers on a certain supply chain. The top-layer model of inventory system is shown in figure 1 below.
In the model, there are three modules: customer, warehouse and supplier. The warehouse accepts
customers demands, handles orders according to the principle of first come first serve, checks storage
and makes a count, then sends replenishment requirement to suppliers. Goods are issued by the supplier
according to the replenishment requirement, accepted by the warehouse and sent to customers.

299

300

DONG Fu-gui et al. / Systems Engineering Procedia 4 (2012) 298 304

customers

goods
demand

warehouse

replenishment
requierment

supplier

Figure 1. Top-layer model of inventory system

There are two replenishment strategies, cyclical and continuous strategies. The former checks the
inventory status on a fixed cycle and is divided into (t, S) and (t, R, S) policy, the latter concerns about the
continuing changes in the inventory level and contains (Q, R) and (R, S) policy. (Q, R) policy is better to
work in lower volatility conditions but (R, S) policy is better to work in higher volatility conditions.
There are different ordering policies and allocating strategies in inventory system. This paper selects
continuous replenishment strategies and independent ordering strategies with no price discount to
simulate an inventory system of single point. Customers' demands is random and have time constraint. If
the order is finished within the specified timeframe, then its not necessary to consider the loss due to
shortage, or else the partial loss is included. This model assumes that customers demand is normally
distributed with mean 500, variance 12. The demand lead time is normally distributed with mean
2.5.variance 0.8. The maximum inventory is set to 3000 and the safety stock quantity is 1500. And the
model supposes that suppliers production ability is infinite.
2.2. Simulation model
This model contains warehouse Agent and supplier Agent. The former mainly achieves three
functions: customer order management, inventory control and cost statistics, the latter mainly imitates
replenishment behavior. The working mechanism of the whole model is shown in figure 2.
a new day
begins

Check
inventory

Request
Accept
order

Send
replenishment
request

Accept
order

Sort order

Receive
replenishment

Sort order

Replenish

Handle
orders

Warehouse
Agent

Supplier
Agent

Figure 2. Working mechanism of the whole model

Each day begins with inventory checking, orders handling, and random demand producing. The
customers demand will trigger a series of events and make agents properties or behaviors change.

DONG Fu-gui et al. / Systems Engineering Procedia 4 (2012) 298 304

Firstly, the customer produces random demand, and then warehouse agent takes orders, records customer
demand information and handles order according to the principle of first come first serve. If the current
inventory meets this order warehouse will provide goods at inventory, or else check pipeline stock. If the
transportation inventory can satisfy the needs for the rest order demand, they consider making up the
order with the pipeline stock. The goods will arrive during the demand lead time, or else renew this
shortage loss. The flow chart of handling orders is shown in figure 3.
Begin
Remove order
No order

End

Y
Take an order

order quantity<=
current inventory

Meet demand with

current inventory

Y
pipeline stock can
meet rest demand

Renew shortage loss

N

Y
Meet rest demand with
pipeline stock

Pipeline stock can arrive

during this demand lead time
Y
Meet demand with
pipeline stock

Figure 3. Flow chart of handling orders

Begin

No order
N

Take an order

Remove order

Replenish according
to order

Renew the number of

pipeline stock

End

Figure 4. Supplier agents work flow diagram

After the warehouse agent successfully deals with an order inventory the system state will change.
The warehouse will begin to check inventory as soon as it successfully deals with an order. Using (Q, R)
policy warehouse will order Q goods each time while the current inventory quantity is smaller than R.

301

302

DONG Fu-gui et al. / Systems Engineering Procedia 4 (2012) 298 304

The process of (R, S) policy is quite similar to the process of (Q, R), the difference lies in the order
quantity. Warehouse will send requests to replenish inventory to the highest level.
The supplier agent mainly realizes replenish according to warehouses request, and the process of
dealing with orders is similar to the warehouse. Figure 4 is supplier agents work flow diagram.
3. Simulation results
3.1 Simulation experiment results
The simulation indices in this model are order fill rate (customer service level) and total cost. The
table1 respectively records the results of taking different replenishment strategies.
Table 1. Comparison of two replenishment policy

R,Spolicy

Q,Rpolicy

count

mean value

shortage costs

500091

69172.798

Inventory holding costs

500091

3962.342

count

mean value

shortage costs

500175

249487.521

Inventory holding costs

500175

3287.328

Service level is shown in red. As showed by the diagrams that the service level of (R, S) policy is
slightly higher than (Q, R) policy. The last row records the shortage costs and inventory holding costs of
one simulation adopting (R, S) and (Q, R) policy respectively. The table shows that (R, S) policy service
is better, and the shortage costs is lower, but inventory holding costs is slightly higher than those under
(Q, R) policy. But as a whole, the (R, S) policy is better than (Q, R) policy.

DONG Fu-gui et al. / Systems Engineering Procedia 4 (2012) 298 304

3.2 Optimization experiment results

Optimization is the process of finding the optimal combination of conditions so that can find the best
possible solution. AnyLogic optimization is built based on the OptQuest Optimization Engine which can
automatically finds the best parameters of a model with respect to certain constraints.
In this model, (R, S) policy is adopted, the order point is considered as an optimization parameter,
and total cost is to be minimized. A requirement that the service level should be up to 0.95 is an
additional restriction imposed on the solution found by the optimization engine. Requirements will be
checked at the end of each simulation, and if they cant meet the standard, the parameters will be rejected.
Otherwise the parameters will be accepted.
The results of three optimization experiments are shown in table 2. It indicate that if safety stocks
can meet two or three days demand the warehouse can reach the lowest cost with a certain service level.
Table 2. Results of three optimization experiments

optimum order point1044

optimum order point1377

optimum order point1270

4. Conclusion
Based on agent method, this paper describes the single point inventory system simulation model
using AnyLogic software. In this model the customers demand and lead time cant be determined ahead.
A comparison is made between two continuous replenishment strategies. Simulation experiment verifies
the exactness of this model and concludes that (R, S) policy is better than (Q, R) policy. At last,
optimization experiments are made to testify the optimize storage policy under certain conditions.

303

304

DONG Fu-gui et al. / Systems Engineering Procedia 4 (2012) 298 304

References
[1]Zhang qing,Liu zhixue. Multi-agent-based modeling and simulation for information coordination in supply chain.
Application Research of Computers 2009;26(10):3709-3711.
[2]Zhang chong,Wang haiyan. Analysis of stability and bullwhip effect in production-inventory systems. Journal of Southeast
University(English Edition) 2011;27(01):101-106.
[3]Li wenfeng,Yuan bing,Zhang yu. Logistics System Modeling and Simulation. Beijing: Science Press.2010
[4] Yu-Cheng Hsiao. Integrated logistic and inventory model for a two-stage supply chain controlled by the reorder and shipping
points with sharing information. International journal of production economics 2008;115(1):229-235.
[5] Kanchana Kanchanasuntorn, Anulark Techanitisawad. An approximate periodic model for fixed-life perishable products in a
two-echelon inventorydistribution system. International Journal of Production Economics 2006;100(1):101-115.
[6] Konstantin Kosti. Inventory control as a discrete system control for the fixed-order quantity system. Applied mathatical
modeling 2009;33(11):4201-4214.
[7] Shuo Huanga, Sven Axsterb, Yifan Doua, Jian Chen. A real-time decision rule for an inventory system with committed
service time and emergency orders. European journal of operational research 2011;215(1):70-79.
[8] P. Kchel,U. Nielnder. Simulation-based optimisation of multi-echelon inventory systems. International journal of
production economics 2005;93-94:505-513.
[9]Wan jie,Zhao cong,Han jianfeng. Research on the Cost Simulation of Supply Chain Sys-tem Based on (R,Q)Policy. Journal
of Hebei University of Technology 2010;39(1):79-83.