11

Automated Storage Systems

E. A. Elsayed

INTRODUCTION

The increasing sophistication and tbe need for increased flexibility of operating systems,
whether in manufacturing or service industries, necessitates the use of sophisticated
systems for storage, retrieval, handling and movement of materials, parts, items, etc.
Not only must such 'logistics' systems receive, and supply items for the operating system
as and when required, but also the control systems used for logistics must complement
those used for the running of the manufacturing or service system. Not surprisingly in
this situation, increasing attention is being given to the use of automated systems for
materials and parts handling, storage, retrieval and distribution. In this chapter we shall
look at tbe type of systems employed, and consider some aspects of the design and per-
formance of such systems. We shall concentrate upon automated storage/retrieval
systems.

TYPES OF SYSTEM

Automated storage and retrieval systems (AS/RS) are becoming common in automated
factories, distribution centres and manufacturing environments. Their application and
size vary from simple storage and retrieval systems, to centralized systems where all otber
production, assembly and manufacturing operations are concentrically located around
them. 'Mini-load' systems are used in assembly operations to make storage on the
factory floor more flexible and responsive to tbe needs of production. They allow work
in progress to be stored adjacent to production, thereby minimizing inventory levels and
allowing manufacturers to institute just-in-time deliveries of materials.
'Micro-load' systems are scaled down mini-load systems. They are smaller in height
and handle lighter loads than tbe traditional mini-load AS/RS. Many of these designs
contain openings in the storage racks where assembly workstations are installed. The
S/R machine delivers materials directly from tbe storage location to the workstation,
thus tbe transport time is reduced to a minimum. In addition, finished products at work-
stations are directly stored for later use by other workstations or for delivery to otber dis-
tribution facilities.
Most of tbe mini-load or the micro-load systems are used for assembling small elec-
tronic components. Anotber use of these mini-load systems is in tbe clean-room manu-
facturing environments where tbe automatic handling of devices such as disk drives and
 Automated Storage Systems 175

semiconductor devices reduces the contamination of the products from manual

handling.
Tbe applications of automated storage and retrieval systems have also been extended
to tbe service area. For example, AS/RS is used in the frozen food processing environ-
ment where the temperature is always kept at -20°F making it extremely hostile to
human operators. In addition, the perishability of the product line, tbe need for quick
response to ever changing retail marketing conditions, and tbe high degree of flexibility
and ready access to various product lines dictate the use of AS/RS to meet all the require-
ments.
Another use of AS/RS in the service area involves the storage and retrieval of cars in
specially designed parking garages. In tbis application, the cost of space in large cities
requires its efficient utilization. Therefore, storage structures are constructed such tbat
each storage location can accommodate one or two cars when stored either in single- or
double-deep storage bins (locations). A computer controlled crane (vehicle) performs
tbe retrieval and storage of cars on command. Thus greater utilization of space, low
manpower requirements and significant savings in beating, ventilation and air condi-
tioning are realized.
Wine producers are employing AS/RS in wine processing, storage and ageing. For
example, the ageing of wine requires storing the wine bottles (or containers) in a dark
place for a period of time. This can be accomplished by constructing underground
AS/RS where computer controlled cranes can perform the storage of wine bottles in
specific locations in the storage structure and tbe retrieval of those bottles tbat have aged
for the specified time. In such an application, space is saved, and the operating cost is
reduced to a minimum.
In this chapter, we describe, in detail, the components of AS/RS, factors that affect
tbe performance of the systems, and different operating policies of sucb systems.

COMPONENTS OF AS/RS

A typical AS/RS, whether it is a large-scale system, a mini-load system, or a macro

system, consists of four basic components. These are:
(a) a storage structure;
(b) storage and retrieval (S/R) machine(s);
(c) conveying devices sucb as conveyors, fork-lift trucks, driverless cars or AGVS (auto-
mated guided vehicle systems) to transport material to areas outside the reach of tbe
AS/RS; and
(d) computers to control tbe storage and retrieval operations within the AS/RS.
Figure 11.1 shows the components of tbe AS/RS, and further details are given below.

Storage structure

Tbe most widely applied AS/RS type is the unit load system (a unit load is defined as a
pallet or standard container in contrast with packaged or bulk material). Tbe unit loads
are usually stored in a single-deep system where tbe storage location is one unit deep.
Other types of storage structures are tbe double-deep or multi-deep storage systems
176 Technologies

S/R machine

Workstations

Order picking
stations

Figure 11.1 Components of an AS/RS.

(deep-lane systems). These systems are used with relatively few SKUs (stock keeping
units) where large quantities are involved such as in the food processing industry.
Storing or retrieving items (loads) in deep-lane systems may be performed by using
different methods. The first method is the flow-through rack. In this method, the flow-
tbrougb racks have an input and an output side, and loads to be stored are transported by
a specially designed S/R machine to the input side of tbe racks. A similar S/R machine
on the output side of the flow-through racks is used for retrieving loads. The flow-
through racks may utilize gravity feed or powered conveyors. The choice of the conveyor
system for tbe racks depends on tbe size and weight of loads and the number of loads in
tbe deep lane (double-deep, three loads deep, . . . ten loads deep).
Tbe second and most common method for load storage and retrieval in deep lane
systems uses S/R machines each equipped witb a rack entry module (REM). Tbe loads
are stored in deep lanes but are moved by the REM rather than some type of in-rack
conveyorization as discussed in tbe first method. The REM transports the load until it is
adjacent to a stored load, deposits it, and returns to the S/R machine. This device typi-
cally communicates with the S/R machine via a radio frequency (RF) link. A number of
these systems are being used in the frozen food industry where throughput may not be
great and bigb density storage permits efficient use of energy by the refrigeration system
(Budill, 1983).
Having discussed the types of storage structures of AS/RS, it may be appropriate to
present an overview of the sizes of these structures. Most of AS/RS structures are free-
standing steel structures that range from 400 ft to 800 ft in length, and from 50 ft to 100 ft
in height for large systems. Tbe storage bin size ranges from 18 in x 36 in for mini-load
systems to 10 ft x 12 ft for large systems, while tbe depth of tbe storage bins varies
between 36 in for single-deep systems to 360 in for deep-lane systems. Mini-ioad systems
are normally installed in existing buildings and generally are under 30 ft in height.
The storage structure and its dimensions have a major effect on system performance.
This will be explained later in this chapter.
 Automated Storage Systems 177

Storage and retrieval machines

The S/R machine is used for transporting, storing and retrieving tbe loads within tbe
AS/RS system. It operates at a horizontal speed of 400-500 ft/min and an acceleration
of 1.0 to 1.8 ft/sec^, while its vertical speed ranges between 70 and 100 ft/min. The S/R
machine travels over a floor mounted track and is guided at tbe top by bearings which
slide in a slot equivalent to the aisle length.
The S/R machine is capable of travelling horizontally and vertically at the same time.
Hence, tbe time to travel from point / (coordinates^f,,,^,) to pointy (coordinates A'^,.Vy) is
the maximum of the horizontal and vertical travel times. Tbe S/R machine bas another
direction of motion: tbe shuttle subcycle which includes extension, pick up and retrac-
tion of tbe picking bead. Storing or retrieving items can be accomplished by many
methods sucb as having one or more operators on board, having an extractor which
extends under the load to be retrieved, or having a robot guidance system where a robot is
guided by a vision system to identify items to be picked. Tbe technology of using robots
with guidance systems is in an infant stage and further research is needed to apply effi-
ciently such technology. A typical S/R machine is shown in Figure 11.2.

S/R machine

Extractor

Figure 11.2 A typical S/R machine. (Courtesy of Hartman Materials Handling Systems, Inc.)
178 Technologies

Allocation of S/R machines to the aisles of the AS/RS is based on tbe demand require-
ments for order storage and retrieval. For example, in systems witb high throughput
requirements a dedicated S/R machine to each aisle may be recommended. In tbis case,
each aisle is served by a dedicated S/R machine and no transfers of tbe S/R machines
between aisles are allowed. On the other band, if an exceedingly large number of unit
loads must be accommodated in a system which has relatively low throughput require-
ments it may not be economically feasible to provide a dedicated S/R machine to each
aisle. Therefore, an aisle transfer car may be considered.
A transfer car is, as the name implies, a device tbat can be used to transfer tbe S/R
machine from one aisle to another. Detailed performance analysis must be made to
ensure that the system throughput requirements are met when fewer S/R machines than
the number of aisles are employed. This is because a 70 ft high S/R machine may weigh
over 30,000 lb, and tbe cost of the transfer car needed to transfer such a machine
becomes a major percentage of the S/R machine cost, and it requires substantial space at
botb ends of tbe aisle.
The simplest type of transfer system occurs when a transfer car is dedicated to only one
S/R machine. In tbis type of application a ratio of one S/R machine to four to five aisles
would be a good selection by the planner (Budill, 1983).
It should be noted, however, that an optimum aisle transfer strategy must be
developed since it is not efficient to move S/R machines randomly between aisles. In
addition, aisle transfer systems require accurate installations because each aisle must be
identical.

Conveying devices

The conveying system transports material to areas beyond the reach of the S/R
machines. Horizontal handling systems are needed to transport material, components,
parts or products to manufacturing operations or delivery points which cannot be ser-
viced directly by the AS/RS.
Such automated storage and retrieval systems and the horizontal handling systems
which support manufacturing operations are of critical concern. Manufacturing opera-
tions may be based on high volume production at the originating operations such as
automatic screw machines and press operations. Tbe AS/RS must perform the storage
and retrieval functions at a rate that can meet the requirements of such high volume con-
tinuous processes otherwise it may block the transportation system which may ultimately
shut down the entire production system. A steel or aluminium mill or an automated
stamping operation are typical examples of this kind of system.
Conveying loads away from tbe AS/RS may be accomplished by employing otber
transportation systems sucb as conveyors, fork-lifts, driverless cars, cranes, or AGVS.
There are many types of conveyor system that can be used to perform sucb tasks. The
conveyor is tbe least expensive transportation (or temporary storage) mode per foot but
it is the least flexible system. Its reliability and operating policies are important factors in
ensuring delivery of loads at their destinations to meet production or distribution sche-
dules. Failure of one section of the conveyor system may result in a total blockage of tbe
transportation system and tbe production operations.
Automatic guided vehicle systems (AGVS) are one of the fastest growing segments of
the materials handling industry. Unlike conveyor systems, AGVS are more flexible with
respect to routeing, scheduling and adaptation for different types of loads. Tbe AGVS
 Automated Storage Systems 179

include vehicles, guidepath, controls and interfaces for automatic routeing and position-
ing. Tbe vehicles are guided in tbeir patbs either by optical or magnetic guidance tech-
niques. For optical guidance, vehicles track a guidepath placed on tbe floor. A high
contrast between tbe floor and tbe guidepath is required for reliable tracking. A reflec-
tive tape or painted stripe on the floor is a common method. The vehicle focuses a light
beam on the guidepatb, and by measuring the amplitude of tbe reflected light is able to
track tbe path accurately. Other applications use ultra-violet systems for path tracking.
For magnetic guidance, a slot is cut into the floor surface and one or more wires are
placed in the floor. Once tbe wires are in place, the slot is filled witb a grout for a smooth
and level finish. The wires are energized with a low-current, low-frequency alternating
current signal which generates a magnetic field around the wire. Vehicle sensors detect
this magnetic field for accurate tracking (Adams, 1983).
In summary, an efficient and appropriate conveying system is needed to transport
loads beyond the reach of tbe AS/RS. Selection of the appropriate conveying system
depends on the application and the environment in which it will be operating.

Computer controls

The computer control of the AS/RS consists mainly of two types:

(a) individual machine controls, where tbe S/R machine has its own on-board computer
which controls the movements and storage and retrieval functions of the picking
head of the S/R machine, and
(b) the main warehouse or factory computer control where scheduling and routeing of
the AS/RS takes place as well as the interfacing between tbe AS/RS and other opera-
tions in tbe warehouse or factory.
The control system consists of the automation or logic control, tbe motor control
system, the primary positioning system and the safety system. The motor control system
regulates the acceleration and speed of tbe crane, picking head and REM. The logic con-
trol system receives commands from the hardware control system to determine the neces-
sary actions of other control systems such as tbe motor control system.
Tbe main warehouse or factory computer integrates tbe functions of tbe AS/RS with
other operations in tbe warehouse or on tbe factory floor. It controls tbe storage and
retrieval processes in tbe AS/RS, places orders to replenish the depleted inventories, and
performs order grouping, scheduling of orders and routeing of the S/R machine. Effec-
tive operation of the AS/RS requires efficient control algorithms that can co-ordinate
the largenumber of activities and operations in the warehouse or on the factory floor in
real time.

Having presented the components of automated storage and retrieval systems, we now
discuss in detail the parameters which affect the performance of such systems. Also dis-
cussed are design guidelines for these parameters.

PERFORMANCE OF AUTOMATED STORAGE AND RETRIEVAL SYSTEMS

Performance of automated storage and retrieving systems may be measured by the delay
time in storing or retrieval items, throughput rate, the average cost per storage or
retrieval operation, and the total time required to store/retrieve a batcb of orders. Tbe
180 Technologies

total time to store or retrieve a batch of orders is, perhaps, the most widely used measure
of performance of AS/RS. There are many parameters that affect the performance of
automated storage and retrieval systems. Although some of the parameters are inter-
related, we shall divide them into three main groups;
(a) operating policies;
(b) physical design; and
(c) demand requirements.

Operating policies

One of the major problems in AS/RS is how to operate the system efficiently, since the
operating procedures have dominant effects on the system performance and cost. Some
of the operating issues which arise in the decision-making stage are: dedicated or shared
storage policies for items, relative locations of items in the storage racks, sequential pick-
ing V. batch picking, assignments of orders to batches, routeing of the picking crane to
perform the required picks in the least amount of time, and order interleaving (perform
storage and retrieval operations in the same cycle) v. performing retrieval and storage
operations in different cycles.
The operating policies can be subdivided into three areas: storage policies, order
batching, and S/R machine routeing.

Storage Policies

CPO (cube-per-order) method The problem of locating (assigning) items in the storage
structure is of extreme importance since it directly affects the variable cost of order
storage and retrieval. There are, in general, four factors which affect this variable cost:
the compatibihty of products, whether they are complementary, their popularity, and
how much space they require. Compatibility of two types of products implies that these
products can be stored next (or adjacent) to each other without fear of spoilage or con-
tamination of either type of product. Complementary products are those products which
are frequently ordered simultaneously such as bolts and nuts or pistons and piston rings.
These complementary products should be stored next to each other. The popularity of a
product is measured by the average number of trips per day to store or retrieve items of
the product. The product popularity is the one most commonly recommended for stock
location. Thus products having high popularity values should be stored closest to the
deposit and picking (DP) station of the AS/RS. The space required for items is the fourth
factor which affects the variable storage and retrieval cost. It is recommended that items
requiring the least amount of warehouse space be placed closest to the DP station.
Obviously, the above four factors cannot be met simultaneously and some com-
promise is needed. Assuming that all products in the warehouse are compatible, Heskett
(1963) developed a new criterion which combines the space and popularity of the item
and is referred to by the 'cube-per-order' index or CPO. The CPO establishes a quantita-
tive trade-off between the dual objectives of placing closest to the DP station those items
taking up the least space, and also those items which are the most popular. To calculate
the CPO for each product, /, one needs to determine the average number of shipping
 Automated Storage Systems 181

Figure 11.3 Allocation of items to storage zones.

units per order, n^, the average number of cubic feet of storage required per order, v^, and
the average number of orders to be received per shipping day, d^, during the time
horizon. The last two numbers are then multiplied together with the number of days
specified as the maximum days' demand target to yield the amount of cubic footage of
staging area space to be allocated for each item. The CPO is calculated as the ratio of this
latter number to the average number of orders per shipping day, or:

CPO, = - ^ m

After calculating the CPO for each product, the products are then ranked based on
their cube-per-order index; the product with the lowest index is ranked first. From this
ranking, the staging area layout immediately follows. The product with the lowest index
goes in Zone 1, closest to the DP station, using up as much space as needed to accom-
modate the demand for the time horizon. If not enough space is available in Zone 1, the
excess amount is assigned to Zone 2. On the other hand, if extra space is available in Zone
1 more than what is required from the product with the lowest index, the next lowest
index product is also placed in Zone 1 in the appropriate amount up to the capacity of
Zone I (see Figure i 1.3). This process is repeated until all products are assigned in the
proper zones (Kallina and Lynn, 1976).

Dedicated and randomized storage methods In this section, we present two major
methods of storage policies, namely: dedicated storage policy and randomized storage
policy. Under the dedicated storage policy, a particular number of storage locations is
reserved for a class of products. In other words, the same product is found in the same
location during the planning horizon. The dedicated storage policy is sometimes referred
to as a fixed slot policy, implying that a fixed location is always allocated to a specific
product. The exact assignments of products to locations may be achieved based on the
CPO index. It should be noted that the total number of assigned locations to products
must accommodate the sum of the maximum inventory levels of these products.
182 Technologies

The dedicated storage policy, if based on activity, will maximize throughput of the
AS/RS. However, the implementation of such a storage policy requires more informa-
tion and accurate data to plan the system for maximum efficiency (White and Kinney,
1982).
Under the randomized storage policy, products are stored in the closest available loca-
tion, and are retrieved on a first-in-first-out (FIFO) basis. Thus, the product's location
changes according to its use. In fact, items of the same product may exist in more than
one location. However, applying the FIFO service discipline for item retrievals ensures
uniform rotation of stock.
The main advantage of randomized storage policies when compared with dedicated
storage policies is that they require less space to meet the same demand requirements.
However, the randomized storage policies have lower throughput and require com-
plicated storage and retrieval policies which can keep track of each product, the number
of locations at which it exists, the arrival date of the product to the inventory system, and
dynamic scheduling of the S/R machine.

Mathematical programming formulation of the storage assignment problem Math-

ematical programming can be used as an alternative method to CPO for assigning
storage spaces to products. The following are two mathematical programming formula-
tions for illustrative purposes only: linear and integer programming. Other formulations
of the problem can be developed according to the objective function and the required
constraints.

LINEAR PROGRAMMING FORMULATION

One of the first linear programming models for order picking (or storing) operations is
reported in Ballou (1967). The problem is to find the optimal assignments of N types of
products to A/total locations in the storage structure of which L locations are in the over-
flow or reserve area (an area which is reserved when there is an overflow of products
from their pre-assigned locations Af). Let x,j be the quantity of product / to be assigned
(stored) in location^ and C,j be the cost of storing or retrieving an item of product / to or
from location^ measured from the DP point. Assume that the capacity of storage loca-
tiony is bj. The objective function is expressed as a linear equation of the travel cost from
the DP point to all possible locations in tbe storage structure, i.e.

N M
Minimize ^ 2 ^u -^t/ ^2)

Subject to 2 ^ij ~ ^J j = I, . . .M (3)

i= \,...N (4)

/ = 1, . . . N (5)

(6)
 Automated Storage Systems 183

where constraint (3) ensures that any storagey cannot contain more than bj items, con-
straint (4) ensures that the total items of product / cannot reach beiow a reorder level r,
otherwise orders of this product cannot be retrieved. The purpose of constraint (5) is to
ensure that the total inventory level of product / at all locations is greater or equal to a
specified quantity QI. The objective function ignores the cost of transporting items
between pre-assigned storage locations and the locations in the reserve or overflow area.
Solution of this problem can be easily attained by using a standard software package
for linear programming. However, for storage structures with a large number of loca-
tions (1,000,000) and a large number of products (100,000>), optimal solution of this
problem may prove to be impossible to obtain. In this case, reformulation of the
problem may be necessary to reduce its size.

INTEGER PROGRAMMING FORMULATION

Goetschalchx (1983) reported an integer programming formulation of the storage
assignment problem as described in Francis and White (1974). It is assumed that items
have equal size, there are no overflow areas, and one or more DP stations. The DP sta-
tions are indexed by k = 1, . . . , K, the products are indexed by / = 1 A', and the
storage locations byj= I, . . . , A/.
Define: t^f, as the travel time from DP station k to storage location /.
(7,j as the quantity of product / to be handled through the DP
station k.
' 1 if an item of product / is assigned to location 7

0 otherwise.
The cost of assigning an item of product / to location y is given by:

*=l

where Q, is the total quantity of product / to be assigned to the storage locations.

The problem is formulated as:

Minimize: V V C, X; (8)
i=l >=l

,M

Subject to: ^ -^t/ = ^' / = 1, . . ., N (9)

2 X, ^1 y=l,...,A^ (10)

;f, = Oor 1 (II)

Constraint (9) ensures that Q, items of product ( are to be assigned to the storage loca-
tions while constraint (10) states that at most one item is assigned to each location.
184 Technologies

This problem is referred to as an 'assignment problem' which can be optimally solved

using transportation or network flow algorithms.

Travel (ime calculations As mentioned earlier, the S/R machine is capable of travelling
in the horizontal and vertical directions at the same time. The travel time between loca-
tions / and j with horizontal coordinates x^ and Xj and vertical coordinates y, and Vj
respectively is determined by the Chebyshev norm:
l^j = max [/,-, tj]
where /"„ = travel time between locations / andy
/,- = time to travel a horizontal distance {x^ - Xj\
tj = time to travel a vertical distance {y^ - yj\
The travel time /, is calculated by determining the time /, (acceleration period) required to
reach the maximum horizontal speed of the S/R machine, time /; during the constant
speed period of the S/R machine, and time f, (deceleration period) required to bring the
S/R machine from maximum speed to zero speed, as shown in Figure 11.4. The travel

Time

fi acceleration time
f2 constant speed time
^3 deceleration time

Figure 11.4 Travel time of S/R machine in the horizontal direction between locations'\ and].
 Automated Storage Systems 185

time t, is the sum of r,, /; and r,. The trave! time /, is calculated following the same pro-
cedure of t, and using appropriate vertical speed and acceleration. The shuttle subcycle
time is negligible when compared with /, or tj.
The throughput of the AS/RS is greatly affected by the travel time of the S/R machine
(time to perform storage and retrieval tasks). The time that the S/R machine consumes
starting from the DP station, performs the required storage and retrieval tasks and
returns to the DP station is referred to as a 'command cycle'. There are two types of
command cycle. The first is single-command multiple-address cycle, which consists of
either storage or retrieval but not both. A single storage command multiple-address cycle
starts by sending a loaded S/R machine from DP station to the first location where some
items are to be stored; it deposits these items, travels to the next location where items are
to be stored, deposits these items, and continues this process until all items are deposited
in their corresponding locations. The cycle ends by directing the S/R machine to return
to the starting point of the cycle. The single-command multiple-address cycle is referred
to as a single-command cycle when the S/R machine performs either one storage or one
retrieval during the cycle.
The second type of cycle is the dual-command multiple-address cycle (or interleaving
cycle) in which storage and retrievals are performed in the same cycle. Generally, the
dual-command multiple-address cycle begins with the S/R machine at the DP station; it
picks up items of different products, deposits items in their appropriate locations,
retrieves items from other locations, and repeats this sequence until all items needed to be
stored are deposited in their locations and those needed to be retrieved are picked up. The
cycle ends when the S/R machine completes the last storage or retrieval process and
returns to the DP station with the retrieved items. In other words, the S/R machine starts
with a load to be stored and returns with a load of items that are needed to be picked up.
When the S/R machine performs one storage and one retrieval, other than those per-
formed at the DP station, in one cycle, this cycle is called a dual-command cycle.
The travel times, as published by the Material Handling Institute (1977), of the S/R
machine for single- and dual-command cycles, regardless of the storage policy being
used, are given below:
Tsc = 1 (13)
7~Dc=l-5 (14)
where r^c is the travel time of the S/R machine for the single-command cycle
Tpt is the travel time for the double-command cycle
In other words, the travel time of the dual-command cycie, on average, is 50 per cent
longer than that of the single-command cycle.
Bozer and White (1984) developed expressions for the travel times for both single-
command and dual-command cycles. The following assumptions are required for
deriving these expressions:
(a) The rack is considered to be a continuous rectangular pick face where the
input/output point is located at the lower left-hand corner. A randomized storage
policy is used, i.e. any point within the pick face is equally likely to be selected for
storage or retrieval.
(b) The S/R machine is assumed to move simultaneously in horizontal and vertical
directions, so that travel time is the maximum of horizontal travel time and vertical
travel time.
186 Technologies

The expected travel times using F C F S (First-Come First-Service) for single- a n d dual-
command cycle a r e :

Sh = horizontal travel speed [ft/mini

5v = vertical travel speed [ft/min]
L - rack length [ft]
H = rack height [ft]
L
ti, = " ^ ; time t o reach the end of the rack [min]
LJ
^ = -^ y time to reach the end of the rack [min]
S
T = travel time; max (/"h, O [min]
b = shape-factor of the rack; min ( ^

Expected travel time of a single-command cycle, E(SC), is:

E(SC)= [1 + \b']T(min) (15)
Expected travel time between any two randomly selected points, E(TB), in a dual-
command cycle is:
E(TB) = [i+ib'-^b']T[mm] (16)
The expected travel time of a dual-command cycle, E(DC), is obtained by adding
equations (15) and (16). Thus:
E(DC) = {*j+ \h' - ^ 6 ' ] r [ m i n l (17)
where T is referred to as the normalization constant, since equations (15), (16)
and (17), without T, are dimensionless
b is referred to as the shape-factor, since the normalized rack is 1.0 x b
(01 b X I.O)
Clearly, estimation of travel times in single- or dual-command multiple-address cycles
is a difficult task, since the relationships between the distribution of the number of pick-
ups and deposits, the distribution of the quantities to be picked up or deposited, and the
distribution of the locations to be visited in a cycle are difficult to determine. Therefore,
simulation and approximation models may be used in the estimation of the travel times.

Order retrieval policies

Sequencing and scheduling order retrievals (storages) is a complex and important pro-
blem. The fact that 60 per cent reduction in travel between locations can be achieved with
reasonable effort (Han et al., 1987) shows the importance of having efficient sequencing
and scheduling solutions. Earlier, we discussed how order retrievals can be efficiently
achieved. The following terminologies relating to order retrievals now need to be
defined:
• Order or request: An order consists of a set of items to be stored
or retrieved consecutively.
 Automated Storage Systems 187

• Order batch or order block: A set of orders that must be picked and/or
stored in one cycle of the S/R machine. This
can oniy be achieved if items of several orders
can be mixed during picking or storing, sorting
of the orders at the DP station is required to
separate items by order.
• Batching policy (sequencing): The policy which determines the orders that can
be picked/stored in one cycle.
• Routeing of the S/R machine: The travel path of the S/R machine to perform
order storage/retrievals.

The objective of the order retrieval/storage policy is to determine the optimum batch-
ing policy (sequencing of orders) and the optimum routeing of the S/R machine for each
order batch such that the total travel time is minimized. There are two components of
any order retrieval/storage policy: order batching and S/R routeing.

Order batching Order batching is affected by the application and the environment in
which it is being employed. For example, in typical flexible manufacturing systems the
list of retrievals/storage changes through time as old requests are filled and new requests
arrive. Order batching may be performed as (1) select an order batch, when its processing
is complete, select another, etc. - this is analogous to the static job arrival pattern in
scheduling theory - or (2) select an order batch, when its processing is complete, revise
the list of orders as new requests arrive, select another batch, revise the list of orders,
etc. - this is analogous to the dynamic job shop scheduling problem.
There are two questions that need to be addressed to be able to solve the order
retrieval/storage problem:
(a) How should orders be grouped to form a batch?
(b) What is the optimal routeing path of the S/R machine to process this batch?

GROUPING ORDERS TO FORM A BATCH

When a list of retrieval requests is given and a set of these requests is to be formed, these
requests need to be grouped according to some criteria. The list of orders is scanned to
find a 'seed' order to start the batch (e.g. the order with largest number of storage loca-
tions to be visited). Other orders are compared with the seed order and if an order has
common locations with the seed order, it is added to this batch, provided the capacity of
the S/R machine is not exceeded. (If more than one order has common locations with the
seed, the order that has the maximum number of common locations is selected provided
that the capacity of the S/R machine is not exceeded.) When an order is added to the
seed, the locations (common and not common) of the two orders are compared with the
remaining orders of the retrieval list. If an order is found to have common locations with
these two orders, it is added to the batch, provided that the capacity of the S/R machine
is not exceeded. If more than one order is found to have common locations with the first
two orders of the batch, the order with the largest number of common locations with
them is chosen, provided that the capacity of the S/R machine is not exceeded. The steps
are repeated until no more orders can be added to this batch of orders (Elsayed and
Stern, 1983).
188 Technologies

ROUTEING THE S/R MACHINE

The S/R machine is then routed to retrieve the items requested in the batch. Routeing of
the S/R machine can be accomplished by using the well-known travelling salesman pro-
blem (TSP). The TSP is one of the class of so-called 'NP-compIete' problems, with
strong evidence that no efficient algorithm can be developed for its solution (Han et al.,
1987). Another procedure for routeing the S/R machine is the 'nearest neighbour'
heuristic.

The above procedures for order batching and routeing of the S/R machine is illustrated
by the following example.
Consider an eighty location storage structure, as shown in Figure 11.5, with forty loca-
tions on each side of the aisle. Each location has dimensions of 4 ft x 4 ft x 4 ft. The
length of each storage rack is 40 ft and the height is 16 ft. The horizontal and vertical
accelerations of the S/R machine are equal and have values of 1.5 ft/sec'^ while the
horizontal and vertical speeds are 8.3 ft/sec and 2.5 ft/sec respectively. A set of orders is
randomly generated and tabulated in Table 11.1. The capacity of the S/R machine is 350
items. Note that each order consists of several retrievals. The first number of the retrieval
represents the location number of the item and the second number represents the quan-
tity to be retrieved from that location. For example, order no. 10 requires the retrieval of
9 items from location 34, 15 items from location 73, and 18 items from location 26.
Scanning through these orders, order no. 4 has the largest number of locations to visit,
thus order no. 4 is selected as 'seed'. Scanning the remaining orders for common loca-
tions with the seed, order no. 14 has the most number of common locations with order
no. 4 and the total items to be retrieved for both orders no. 4 and no. 14 is 329, which is
less than the capacity of the S/R machine, therefore order no. 14 is added to the batch
which includes order no. 4. Scanning the remaining orders for common locations with
orders no. 4 and no. 14, no order is found to have common locations with these two

Figure 11.5 A two-rack storage system.

 Automated Storage Systems 189

Table 11.1 A randomly generated set of orders

Order no. (Location, quantity); (Location, quantity); etc. Total quantity Total locations

1 (23,18) 18 1
2 (35,13); (18,8); (73,1); (64,9); (29,19); (6,10);
(26,10); (8,13); (61,19); (46,5); (7,8); (63,6);
(70,16); (60,19); (59,10); (71,4); (52,9); (57,18); 197 18
3 (4.9) 9 1
4 (10,7); (58,9); (55,13); (36,15); (1,6); (73,8);
(14,17); (76,6); (56,4); (20,4); (46,18); (32,2);
(11.14); (19,9); (77,15); (17,15); (57,8);
(33,10); (69,9); (49,14); 203 20
5 (77,9); (32,12); (7,6); (47,5); (56,14); (78,16);
(12,7); (50,8); (27,10); (33,15); (16,10);
(53,15); (67,19); (34,6); (15,12); (37,1);
(71,3); (79,13); (80.6); 187 19
6 (7,16); (37,7); (10,7); (38,16); (3,8); (78,8);
(64,14); (14,15); (4,16); (34,18); (25,10);
(79,16); (55,17); (36,9); (67,1); (74,1); (44,13); 192 !^
7 (37,3); (61,16); (66,5); (50,13); (74,17);
(64,5); (19,15); (67,19); (42,6); (13,1); (48,6); 106 1]
8 (28,19); (51,1); 20 2
9 (28,10); (6,8); (43,19); (75,19); (5,17);
(29,3); (18,15); (44,18); (64,12); (27,16); (77.12); 149 11
10 (34,9); (73,15); (26,18); 42 3
11 (36,1); (5,16); (44.11); (26,3); (7.11);
(67,13); (57,11); (24,10); (69,5); (1,10);
(70,17); (71,1); 109 12
12 (54,3); (5,11); (44,11); (46,1); (72,7);
(78,9); (55,6); (14.12); (3,8); (6,18);
(65,2); (17.19); 107 13
13 (54,18); (76.1); (74,16); (19,6); (43,12);
(65,3); (75,1); (10,11); (1.16); (64.17);
(45,6); (79,1); (77,18); (8.1); 127 14
14 (54,1); (79,4); (32,5); (14,18); (1,4);
(39,15); (15,12); (55.11); (78,1); (16,10);
(11,18); (19,11); (36,6); (12,10); 126 14

orders without causing the capacity of the S/R machine to be exceeded. Therefore order
no. 1 is added to the batch which brings the total number of items on the S/R machine to
347 (< 350). Since no more orders can be added to orders no. 4, no. 14 and no. 1 without
exceeding the capacity constraint, then the first order batch has been formed and its
orders are no. 4, no. 14, and no. 1 (total quantity = 347) with 29 locations to be visited.
The scheduling (routeing) of the S/R machine involves solving the 'travelling salesman
problem' for the 29 locations to be visited. The path of the S/R machine which defines
the location and the quantity of items to be picked from the location is:
Tour no. 1 (1,10); (46,18); (49J4); (9,11); (14,35); (54,1); (17,15); (57,8); (69,9);
(73,8); (33.10); (77,15); (78,1); (79,4); (39,15); (36.21); (76.6); (32,7);
(23,18); (19.9); (58,9); (55,24); (15,12); (20,4); (56,4); (16,10); (12,10);
(11,32); (10,7)
The total time required to retrieve orders no. 4, no. 14 and no. 1 when the S/R machine
follows the above path is 67 units of time.
We now scan the remaining orders to form a new order batch. Among the remaining
orders, the seed order (order having the largest number of locations) is order no. 5.
190 Technologies

Table 11.2 Summary of batches formed

Order batch no. Orders included in the batch Total quantity Total locations to visit Time

\ 4, 14, 1 347 29 67
2 5,7, 10,3 344 30 76
3 2, 11.8 326 26 83
4 6. 12 299 24 72
5 13.9 276 21 73

Orders that can be added to the same batch as order no. 5 without violating the capacity
constraint of the S/R machine are orders no. 7, no. 10, and no. 3. The total quantity for
these orders is 344, thus, the second order batch includes orders no. 5, no. 7, no. 10 and
no. 3. The routeing of the S/R machine for the second batch can be obtained by solving
the TSP for the locations to be visited for all orders in the batch. The total travel time to
retrieve the second batch is 76 units of time.
The above procedure is repeated to form other batches. Table 11.2 is a summary of the
formed batches. Thus the totai travel time needed to retrieve all fourteen orders is 371
units of time.
If the criterion of selecting the 'seed' order is changed from the selection of the order
with the largest number of locations to the selection of the order with the maximum
quantity to be retrieved, different order batching is formed, as shown in Table 11.3. The
total travel time needed to retrieve all fourteen orders in this case becomes 358, which is
less than that obtained earlier.

Table 11.3 Summary of batches according to maximum quantity criterion

Order batch no. Orders included in the batch Total quantity Total locations to visit Time

1 4. 14. I 347 29 67
2 2. 11. 10 348 25 75
3 6. 12.3. 328 26 70
4 5.7 293 27 73
5 9. 13 276 21 73

Clearly, the order batching policy as well as the routeing of the S/R machine has a
significant effect on the throughput of the AS/RS.

S/R machine routeing It was mentioned in the previous section that the routeing of the
S/R machine can be obtained using the well-known TSP. A heuristic of the TSP is the
'nearest neighbour' algorithm, which is also referred to as the 'closest unvisited city'
algorithm. It is summarized as follows.
When a set of locations is to be visited by the S/R machine, the S/R machine travels to
the closest location that has not been visited. That is, in the matrix of the travel time
(representing the time of travel from any location to the other locations in the set of loca-
tions to be visited by the S/R machine), the S/R machine starts from any row k of the
travel time matrix and travels to the location denoted by the column that contains the
minimum cost element, excluding locations already visited. There are two possible
alternatives for selecting the origin (row):
 A utomated Storage Systems \ 91

(a) the all-origin alternative which requires that all locations be successively chosen as
the starting point and the best of the resuhing tours is chosen; or
(b) the arbitrary choice of the origin, that one location be designated as the start (or end)
of the tour.
To illustrate this algorithm, consider the travel matrix shown below, where the
element ij of the matrix represents the travel time from location / to location y:

To location
1 7 20 25 9 6 15 37

I 00 4 9 11 6 5 8 20
7 4 00 10 12 1 2 8 15
20 9 10 00 4 11 13 6 12
From 25 11 12 4 00 19 17 9 8
location 9 6 I 11 19 00 4 9 21
6 5 2 13 17 4 00 10 23
15 8 8 6 9 9 10 OO 15
37 20 15 12 8 21 23 15 00

Origin 1: Tour is I--•7-9 - 6 - 1 5 - 2 0 - 25 - 3 7 time = 57

Origin 7: Tour is 7--'8—6 - 1 — 15 —2 0 - 25 - 3 7 - 7 , time = 51
Origin 20: Tour is 20-- 2 5 - *37 - 7 - 9 — 6 - -»! - 1 5 - 2 0 time = 51
Origin 25: Tour is 25-- 2 0 - ••15 - 1 - 7 - 9 - -»6 - 3 7 - 2 5 time = 58
Origin 9: Tour is 9-^•7-6 — 1 - 1 5 — 2 0 - 25 - 3 7 - 9 time = 55
Origin 6: Tour is 6 ^ •7-9 - 1 - 1 5 - 2 0 - 25 - 3 7 - 6 time = 58
Origin 15: Tour is 15--20- *25 — 3 7 - 7 - S ( — • 6 - 1 - 1 5 time = 51
Origin 37: Tour is 37 - 2 5 - *20 — 15 —I—•;' — 9 - 6 - 3 7 time = 58

Any of the tours with travel time of 51 is an appropriate tour for the S/R machine.
There is a close relation between the routeing of the S/R machine and the well-known
vehicle routeing problem in the distribution area. The vehicle routeing problem generally
covers the assignment of truck(s) to different route(s) on which cities are visited and
served without exceeding the capacity of the truck(s). From one point of view, the vehicle
routeing problem is similar to the operation of storage/retrieval processes. For instance,
storage locations can be assumed as nodes of a network and the solution method for this
network is a part of the research work on AS/RS. Some strategies for vehicle routeing
problems are also similar to the strategies of routeing the S/R machine, e.g. cluster-first
route-second, savings or insertion, and interactive optimization.
On the other hand, there are dissimilarities between the routeing of S/R machines and
the vehicle routeing problem. The objective in the AS/RS is to solve the order batching
and S/R machine routeing problems. Orders cannot be assumed as nodes of a network
due to the fact that some of them overlap with others. Another difference is that, unlike
the nodes in a distribution network, orders cannot be divided into different routes (tours)
(Elsayed and Unal, 1987). Therefore, it is important to analyse the AS/RS under study
and utilize the appropriate vehicle routeing algorithm if applicable.
192 Technologies

Physical design

The physical design, configuration or layout of an automated warehouse and AS/RS

have been studied since the early 1960s. The early research is focused on the basic design
concept of a warehouse, including such factors as length, height and the number of
racks.
In the beginning of the 1970s German researchers and engineers studied the design and
application of high-rise parallel configurations as well as carousel conveyors.
Researchers have found that a square-in-time rack configuration (in square-in-time
configurations the S/R machine travels simultaneously from the I/O point along the two
axes of the rack, till the farthest point of the rack, at the same time duration) is verified to
be optimal. This is in agreement with Budill (1983) who states that building the aisles too
long may cause the S/R machines to operate at too high a percentage of their capabilities.
The cost/performance trade-off may require a higher number of shorter aisles.
Systems may be applied at heights from 50 ft to 85 ft without noticeable financial
penalty. Systems in the 90 ft to 100 ft range impose certain penalties in structure and per-
formance. Designing the system with a number of different load heights or load sizes
may ensure efficiency in the structural design of the system. However, as the mix of loads
changes, there may be too few large loads and too many small openings. This lack of
flexibility may result in a higher cost in the future, since some of the desired openings
may be in the least desirable locations which will certainly affect the throughput of the
system.

Demand requirements

The throughput of AS/RS is greatly affected by the demand requirements which

represent the order requests that need to be stored or retrieved to meet the required
production schedule. Demand can be defined by using severaJ parameters, for instance
order arrival rate (constant or variable), order arrival pattern (static or dynamic),
quantity of items to be stored or retrieved per order, weight and size of items, and due
date constraints.
Investigating the effect of demand on the throughput of AS/RS has not been exten-
sively studied by researchers. Few papers have dealt with the effect of demand such as
Elsayed (1985) and Freeman et al. (1983). Elsayed (1985) concludes his study by stating
that as the variance of the quantity per order increases, the effect of rack configuration
on the AS/RS performance becomes insignificant and the differences of performance in
the operating policies are reduced.
The planner of automated storage/retrieval systems must perform detailed analyses of
such systems during the planning stage by considering many design alternatives that can
meet different demand requirements. The analyses will enable the planner to predict and
measure the system's performance as the demand requirements change. The planner will
also be able to design robust AS/RS which can achieve high throughput for high varia-
tions in the demand requirements.
 A utomaled Storage Systems 193

SUMMARY

Automated storage and retrieval systems have entered new areas of complex manu-
facturing, distribution and service systems. Their design and performance are affected
by many factors such as the operating policies, demand requirements and physical struc-
ture. Techniques, such as those presented in this chapter, can be used during the planning
and operation phases of the AS/RS to predict and evaluate different design alternatives
of the system. Presented also are storage and operating policies that can be implemented
to improve the performance of the AS/RS.

REFERENCES

Adams, W. (1983) AGVS horizontal transportation systems. Advanced Institute for Material
Handling Teachers. Alabama: Auburn University. (June 12-17.)
Ballou, R. H. (July 1967) Improving the physical layout ofmerchandise in warehouses. Journal of
Marketing, iX, 60-64.
Bozer, A. Y. and White, A. J. (1984) Travel-time models for automated storage/retrieval
systems. HE Transactions, 16 (4). 329-338.
Budill, E. J. (1983) Automated storage and retrieval systems. Proceedings of Advanced Institute
for Material Handling Teachers, pp. 106-115.
Elsayed, E. A. (1985) Design and Scheduling Rules for Automated Storage/Retrieval Systems.
Working Paper 85-122. Rutgers University: Department of Industrial Engineering.
Elsayed, E. A. and Stern, R. G. (1983) Computerized algorithms for order processing in
automated warehousing systems. International Journal of Production Research, 21 (4),
579-586.
Elsayed, E. A. and Unal, O. (1987) A Review of Automated Storage and Retrieval Systems.
Working Paper 87-103. Rutgers University. Department of Industrial Engineering.
Francis, R. L. and White A. J. (1974) Facility Layout and Location: An Analytic Aproach.
Englewood, N.J. Prentice-Hall.
Freeman, D. R., Hoover, S. V. and Perry, R. F. (!983) Developing Design Heuristics for an
Automated Storage/Retrieval System Using Simulation Modeling. Working Paper.
Northeastern University, Boston, MA.
Goetschalchx, M. P. (1983) Storage and Retrieval Policies for Efficient Order Picking
Operations. PhD Dissertation, School of Industrial Engineering, Georgia Institute of
Technology.
Han, M. H., McGinnis, L. F., Shieh, J. S. and White, A. J. (1987) on sequencing retrievals in
automated storage retrieval systems. HE Transactions, 19 (3), 56-66.
Heskett, J. L. (1963) Cube-per-order index - a key to warehouse stock locations. Transportation
and Distribution -Management, 3, 27-31.
Kallina, C. and Lynn, J. (1976) Application of the cube-per-order index rule for stock location in
distribution warehouses. Interfaces, 7 (I), 37-46.
Material Handling Institute, Inc. (\977) Considerations for Planning and Installing an Automated
Storage/Retrieval System, AS/RS Document - 100 7M.
White, J. A. and Kinney, H. D. (1982) Storage and warehousing. Handbook of Industrial
Engineering. New York, NY: Wiley and Sons. pp. 10.4.1-10.4.32.