Business and Economic Research

ISSN 2162-4860
2017, Vol. 7, No. 1

The Dynamic Programming Models for Inventory

Control System with Time-varying Demand
Truong Hong Trinh (Corresponding author)
The University of Danang, University of Economics, Vietnam
Tel: 84-236-352-5459 E-mail: [email protected]

Received: February 21, 2017 Accepted: March 16, 2017

doi:10.5296/ber.v7i1.10965 URL: https://doi.org/10.5296/ber.v7i1.10965

Abstract
The concept of dependent and independent demand is important in inventory planning and
replenishment that also requires different inventory control solutions. This paper employs the
dynamic programming technique for inventory control system with time-varying demand to
propose the replenishment policy in terms of the economic order quantity, number of
replenishment, and reorder point where total inventory cost is minimized. The study result
indicates that the dynamic programming models outperform the traditional lot sizing models
in terms of total inventory cost. Moreover, the paper creates opportunities for extending
further researches on dynamic inventory related to capacity constraints and uncertainty
conditions of demand, yield, lead time.
Keywords: Dynamic inventory, Lot sizing, Order point system, Material requirement
planning, Dynamic programming model, Inventory control system
1. Background
In business management, inventory consists of a list of goods and materials held available in
stock. The key questions of what and how much inventory are related. Planning is undertaken
to determine the level of inventory that will be needed for operations, and replenishment is
the process of maintaining this level through some combinations of reorder and other
techniques. To determine the level of inventory needed for operations, it is useful to identify
the source of the demand. The concept of dependent and independent demand is important in
inventory planning and replenishment. An item has independent demand when we can not
control it or tie it directly to another item’s demand. While an item has dependent demand
when the demand for an item is controlled directly, or tied to the production of something
else. Therefore, inventory systems with independent demand and dependent demand also
require very different solutions.

128 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

There is an abundant literature on inventory control policies which extend since the 30’s. The
details on these policies may refer to research of Peterson and Silver (1979), and research of
Zipkin (2000). In inventory planning and replenishment, the traditional lot sizing models are
mostly used for inventory control systems. Each lot sizing method outperforms under some
assumptions and demand conditions in which the demand does not present a monotonous
behavior and varies from period to period. There is also some literature that studies dynamic
inventory control policies based on the investigations of Karlin (1960) and Scarf (1959).
Wagner and Whitin (1958) introduced a dynamic programming model in which demand is a
function of time. Silver and Meal (1973) proposed a heuristic method that finds the optimal
order quantity, minimizing the storage and delivery costs. These deterministic and stochastic
models strongly relied on mathematical background that is not easy to understand and
implement the optimal inventory control policies in reality. This paper attempts to develop
the dynamic programming models for both independent inventory system and dependent
inventory system with time-varying demand. These models are evaluated with traditional lot
sizing models such as Lot for Lot (LFL), Economic Order Quantity (EOQ), Period Order
Quantity (POQ), and Minimum Cost per Period (MCP). The paper provides a basic
framework for extending dynamic inventory researches with capacity constraints and
uncertainty conditions of demand, yield, lead time.
2. Order Point System (OPS)
Order Point System (OPS) is the inventory control system for the independent demand. The
multi-period inventory model with time-varying demand is developed to propose the
replenishment policy in terms of order quantity, number of replenishment, and reorder point.
Figure 1 illustrates the typical independent inventory system that has several end-items with
independent demand.
A multi-period inventory model with time-varying demand for outsourcing materials is
modified on the basis of Wagner’s model (Wagner, 1969) under the following assumptions.
1. Backorder is not allowed.
2. Lead time is known with certainty, and assumed constant during the planning horizon.
3. All relevant costs are assumed constant at each period during the planning horizon.
4. No safety stock is assumed.
5. Ordering and holding costs per period are known.
6. Purchase cost is negligible since prices are assumed constant at each period during the
planning horizon.
7. Inventory level at each period is assumed constant in each period.
8. No quantity discount is allowed.
9. Cost of capital is not considered.

129 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

Order quantity

Stock Stock

System

Lead time

Figure 1. The independent inventory system

The dynamic programming model for the independent inventory system (OPS) can be
expressed as follows.
Minimize:

m T m T

 O  N
i 1 t 1
i i ,t  ∑∑H i  X i ,t (1)
i 1 t 1

Subject to:

X i ,t  X i ,t 1  Di ,t  Qi ,t ;i  1..m,t  1..T (2)

Ri ,t  Qi ,t  LTi  0 ;i  1..m,t  1..T  LTi (3)

Qi ,t  N i ,t  LSi ;i  1..m,t  1..T (4)

Qi ,t  N i ,t  M ;i  1..m,t  1..T (5)

X i ,t  0, Qi ,t  0, and N i,t  0,1 ;i  1..m,t  1..T (6)

Where,
Oi = Ordering cost per replenishment for item i at period tth
Hi = Holding cost per unit for item i at period tth
Ni,t = Number of orders taking place for item i at period tth
Di,t = Demand for item i at period tth
Qi,t = Order quantity each time an order takes place for item i at period tth
Xi,t = Inventory level for item i at the end of period tth
LTi = Lead time for each replenishment for item i at period tth

130 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

Ri,t = Reorder point for item i at period tth

t = Period of time during the planning horizon
m = Total number of items
T = Total number of period of time during the planning horizon
The multi-period inventory system under study has three products (end-items) with
independent demand. These items have lumpy demand due to seasonality, trend, and
economic conditions. Information about demand and properties of the system are given as
inputs of the inventory control models. Table 1 gives the demand of three end-items in next
eight periods.
Table 1. The demand of items in the planning horizon
Item 1 2 3 4 5 6 7 8
A 200 200 300 300 350 350 400 400
B 300 300 300 300 300 300 300 300
C 200 250 300 350 350 300 250 200
The properties of the inventory system provide information related to inventory costs, initial
inventory, lead time and lot size. This information is given in Table 2.
Table 2. The properties of the independent inventory system
Ordering cost Holding cost Initial Inventory Lead time Lot size
Item (i)
(Oi) (Hi) (X0i) (LTi) (LSi)
A 1000 2 200 1 100
B 1500 3 600 2 100
C 2000 5 400 1 100
There are many different methods for determining replenishment policy such as Lot for Lot
(LFL), Economic Order Quantity (EOQ), Period Order Quantity (POQ), and Minimum Cost
per Period (MCP). These lot sizing methods are used to compare with the dynamic
programming model that is called the OPS model. Table 3 shows the total cost of models
under the study. The result indicates that the OPS model is better than other lot sizing models
in terms of total inventory cost.
Table 3. Total inventory cost of the models under the study
Method LFL EOQ POQ MCP OPS
Item A 7000 9580 6100 6100 6100
Item B 9900 12660 8100 8100 8100
Item C 15000 16800 17500 13500 13000
Total cost 31900 39040 31700 27700 27200

3. Material Requirement Planning (MRP)

Material Requirement Planning (MRP) is used for the dependent inventory system. The MRP

131 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

model uses a lot of data about items and components. The term of “item” is used to refer to
final product, components and components of components. For each item, it needs to know:
• The lead time, the time leg between the release of an order to the shop floor or to a
supplier and the receipt of the items.
• The lot size, a minimum production quantity (referred to as a minimum lot size for items
that are manufactured in-house) or a minimum order quantity for purchased items.
• The inventory status (stock on hand that calculates based on initial inventory, scheduled
receipt and demand requirement in each period).
• Components needed, which is often referred to as a bill of materials (BOM).

Order Quantity

M
A B
T Stock Final-Assembly Stock
E
R D
I
A C A
L E
Sub-Assemblies

Lead time

Figure 2. The dependent inventory system

An optimization inventory model is not needed to use MRP calculation, the purpose of the
study is to create an optimization problem that matches MRP not for its own sake but to get
started with models that match classic planning systems. Using this model as a starting point,
it is easy to go on to more sophisticated models (Voß and Woodruff, 2006).
The dynamic programming model for the dependent inventory system (MRP) can be
expressed as follows.
Minimize:

m T m T

 Oi  N i,t  ∑∑H i  X i,t

i 1 t 1
(7)
i 1 t 1

Subject to:

 m

X i ,t  X i ,t 1   Di ,t   Pi , j  R j ,t   Qi ,t ;i  1..m, t  1..T (8)
 j 1 

132 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

Ri ,t  Qi ,t  LTi  0 ;i  1..m,t  1..T  LTi (9)

Qi ,t  N i ,t  LSi ;i  1..m,t  1..T (10)

Qi ,t  N i ,t  M ;i  1..m,t  1..T (11)

X i ,t  0, Qi ,t  0, and N i,t  0,1 ;i  1..m,t  1..T (12)

Where,
Oi = Ordering cost per replenishment for item i at period tth
Hi = Holding cost per unit for item i at period tth
Ni,t = Number of orders taking place for item i at period tth
Qi,t = Order quantity each time an order takes place for item i at period tth
Xi,t = Inventory level for item i at the end of period tth
Di,t = External demand for item i at period tth
Pi,j = Number of item i need to make one j
Ri,t = Reorder point for item i at period tth
LTi = Lead time for each replenishment for item i
LSi = Minimum lot size for item i
M = A large number
t = Period of time during the planning horizon
m = Total number of items
T = Total number of period of time during the planning horizon
The multi-stage inventory system is considered to illustrate the inventory control policy with
time-varying demand. Suppose that there is a single end-item A that has a bill of materials as
shown in Figure 3.

133 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

A (1)

B (1) C (1)

D (2) E (3)

Figure 3. BOM structure for the system

In this system, there are two items with independent demand, A and B. Item B is also a
component to make end-item A, so item B has both dependent and independent demand. The
assembly of A requires 1 item B and 1 item C. In order to assembly 1 item C, it requires 2
items D and 3 items E. Items B, D, E are raw materials that are ordered from outside
suppliers. Table 4 describes the properties of the inventory system.
Table 4. The properties of the dependent inventory system
Ordering cost Holding cost Initial Inventory Lead time Lot size
Item (i)
(Oi) (Hi) (X0i) (LTi) (LSi)
A 1000 8 300 1 100
B 1500 6 600 2 100
C 1800 5 400 1 100
D 1300 4 500 1 100
E 2000 7 600 2 100
The demand for item A and item B in the next eight periods is given in Table 5.
Table 5. The demand for item A and item B in the planning horizon
Item 1 2 3 4 5 6 7 8
A 200 0 300 300 350 350 0 400
B 0 300 0 300 300 0 300 0
Based on information about bill of materials (BOM) and data from Table 4 and Table 5, the
MRP model provides optimal solution for scheduled receipts, stock on hand and planned
order release in each period. The optimal inventory model is developed for such system based
on previous dynamic programming model. The objective of the model is to match demand
requirements and minimize total inventory cost.
The effectiveness of MRP model is depended on lot sizing methods. Some traditional lot
sizing methods are used to compare lot sizing generating in the MRP model under total
inventory cost. Lot sizing methods are employed in the study including Lot for Lot (LFL),
Period Order Quantity (POQ), and Minimum Cost per Period (MCP). Table 6 shows total
inventory cost of the models under the study.

134 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

Table 6. Total inventory cost of the models under the study

Method LFL EOQ POQ MCP MRP
Item A 6600 6952 6600 6600 9800
Item B 10200 11244 10200 10200 7800
Item C 10200 19622 12350 10150 9650
Item D 7100 6624 4600 5900 3400
Item E 12300 10200 8200 10200 4000
Total cost 46400 54642 41950 43050 34650
The above result indicates that the MRP model has the least total inventory cost. It reveals
that the MRP model is better than other models with proposed lot sizing methods in terms of
total inventory cost. Moreover, it is interested in extending the MRP model to more
sophisticated models with capacity constraints.
4. Conclusions
The paper has developed the dynamic programming models for both the independent and
independent inventory systems. These dynamic programming models are very basic for
extending to sophisticated inventory control models. Some findings are summarized as
follows.
For the independent demand, the dynamic programming model for the independent inventory
system (OPS) is developed for the multi-period inventory model with time-varying demand.
The result indicates that the OPS model provides the optimal inventory solution in terms of
total inventory cost. Moreover, the model may extend to inventory control policy with
uncertainties in demand, yield and lead time. (Babai and Dallery, 2006). According to the
assumptions of the model, its application is limited to some cases in
practice, especially the accuracy of the forecasted demand in each period. However, it is
found that the model can provide an alternative replenishment policy with significant saving
to the decision maker in managing their system efficiently.
For the dependent demand, there is no perfect model for Material Requirement Planning. In
fact, the MRP model has a number of well-known and very severe problems. Perhaps the two
most serious problems are that lot sizing can cause nervousness and there are no capacity
constraints. Even having serious problems, optimal MRP model can still be very useful for
solving the lot sizing problems. For one thing, it is usually much better than non-planning
model at all. This is particularly true in industries with changing demand patterns where
standard orders cannot be used. The MRP model provides a good starting point for planning
and for the ordering of raw materials. The study result indicates that the MRP model is better
than other traditional lot sizing models as a whole. In addition, using this model as a starting
point, it is easy to go on to more sophisticated models, especially capacity constraints.
References
Babai, M. Z., & Dallery, Y. (2006). A dynamic inventory control policy under demand, yield
and lead time uncertainties. In Service Systems and Service Management, 2006 International

135 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

Conference on (Vol. 2, pp. 1026-1031). IEEE. https://doi.org/10.1109/icsssm.2006.320649

Karlin, S. (1960). Dynamic inventory policy with varying stochastic demands. Management
Science, 6(3), 231-258. https://doi.org/10.1287/mnsc.6.3.231
Peterson, R., & Silver, E. A. (1979). Decision systems for inventory management and
production planning. New York: Wiley.
Scarf, H. (1959). The optimality of (S, s) policies in the Dynamic Inventory Problem.
Mathematical Methods in the Social Sciences, Stanford University Press, Stanford,
California.
Silver, E. A., & Meal, H. C. (1973). A heuristic for selecting lot size quantities for the case of
a deterministic time-varying demand rate and discrete opportunities for replenishment.
Production and Inventory Management, 14(2), 64-74.
Voß, S., & Woodruff, D. L. (2006). Introduction to computational optimization models for
production planning in a supply chain (Vol. 240). Springer Science & Business Media.
Wagner, H. M. (1969). Principles of operations research: with applications to managerial
decisions. In Principles of operations research: with applications to managerial decisions.
Prentice-Hall.
Wagner, H. M., & Whitin, T. M. (1958). Dynamic version of the economic lot size model.
Management science, 5(1), 89-96. https://doi.org/10.1287/mnsc.5.1.89
Zipkin, P. H. (2000). Foundations of inventory management (Vol. 2). New York:
McGraw-Hill.
Appendix
Appendix 1. The OPS Model in Lingo
MODEL:
! The dynamic programming model of independent inventory system;
! Keywords: Order Point System (OPS);
SETS:
! Index of items;
ITEM/1..3/:O,H,X0,LT,LS;
! The planning horizon;
TIME/1..8/;
! Set of item & time, input & output;
LINK(ITEM,TIME):D,N,Q,R,X;
ENDSETS
DATA:
! The properties of the system;
O,H,X0,LT,LS =
1000 2 200 1 100
1500 3 600 2 100

136 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

2000 5 400 1 100;

! The demand requirements;
D=
200 200 300 300 350 350 400 400
300 300 300 300 300 300 300 300
200 250 300 350 350 300 250 200;
ENDDATA
! A large number;
M=10000;
! The objective function;
MIN=@SUM(LINK(I,T):O(I)*N(I,T))+@SUM(LINK(I,T):H(I)*X(I,T));
! The constraints for inventory status;
@FOR(ITEM(I):X(I,1)=X0(I)-D(I,1)+Q(I,1));
@FOR(LINK(I,T)|T #LT# @SIZE(TIME):X(I,T+1)=X(I,T)-D(I,T+1)+Q(I,T+1));
! The constraints for planned order release;
@FOR(ITEM(I):R(I,1)=@SUM(TIME(T)|T #LE# LT(I)+1:Q(I,T)));
@FOR(LINK(I,T)|T #GT# LT(I)+1:R(I,T-LT(I))=Q(I,T));
! The constraints for scheduled receipts;
@FOR(LINK(I,T):Q(I,T)>=N(I,T)*LS(I);Q(I,T)<=N(I,T)*M);
! The decision variables;
@FOR(LINK(I,T):@GIN(X);@GIN(Q));
@FOR(LINK(I,T):@BIN(N));
END
Source: Author’s work

Appendix 2. The MRP Model in Lingo

MODEL:
! The dynamic programming model of dependent inventory system
! Keywords: Material Requirement Planning (MRP);
SETS:
! Index of items;
ITEM/1..5/:O,H,X0,LT,LS;
! The planning horizon;
TIME/1..8/;
! Set of item & time, input & output;
LINK(ITEM,TIME):D,N,Q,R,X;
! Bill of material structure;
PART(ITEM,ITEM):P;
ENDSETS
DATA:
! The properties of the system;
O,H,X0,LT,LS =
1000 8 300 1 100

137 http://ber.macrothink.org
 Business and Economic Research
ISSN 2162-4860
2017, Vol. 7, No. 1

1500 6 600 2 100

1800 5 400 1 100
1300 4 500 1 100
2000 7 600 2 100;
! The demand requirements;
D=
200 0 300 300 350 350 0 400
0 300 0 300 300 0 300 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0;
! Bill of material structure;
P=
0 0 0 0 0
1 0 0 0 0
1 0 0 0 0
0 0 2 0 0
0 0 3 0 0;
ENDDATA
! A large number;
M=10000;
! The objective function;
MIN=@SUM(LINK(I,T):O(I)*N(I,T))+@SUM(LINK(I,T):H(I)*X(I,T));
! The constraints for inventory status;
@FOR(ITEM(I):X(I,1)=X0(I)-D(I,1)-@SUM(ITEM(J):P(I,J)*R(J,1))+ Q(I,1));
@FOR(LINK(I,T)|T #GT# 1:X(I,T)=X(I,T-1)-D(I,T)-
@SUM(ITEM(J):P(I,J)*R(J,T))+Q(I,T));
! The constraints for planned order release;
@FOR(ITEM(I):R(I,1)=@SUM(TIME(T)|T #LE# LT(I)+1:Q(I,T)));
@FOR(LINK(I,T)|T #GT# LT(I)+1:R(I,T-LT(I))=Q(I,T));
! The constraints for scheduled receipts;
@FOR(LINK(I,T):Q(I,T)>=N(I,T)*LS(I);Q(I,T)<=N(I,T)*M);
! The decision variables;
@FOR(LINK(I,T):@GIN(X);@GIN(Q));
@FOR(LINK(I,T):@BIN(N));
END
Source: Author’s work

Copyright Disclaimer
Copyright for this article is retained by the author(s), with first publication rights granted to
the journal.
This is an open-access article distributed under the terms and conditions of the Creative
Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

138 http://ber.macrothink.org