Download as pdf or txt
Download as pdf or txt
You are on page 1of 25

International Journal of Logistics Management, The

An investigation of the effects of storage assignment and picker routing on the occurrence of picker
blocking in manual picker-to-parts warehouses
Torsten Franzke, Eric H. Grosse, Christoph H. Glock, Ralf Elbert,
Article information:
To cite this document:
Torsten Franzke, Eric H. Grosse, Christoph H. Glock, Ralf Elbert, "An investigation of the effects of storage assignment and
picker routing on the occurrence of picker blocking in manual picker-to-parts warehouses", International Journal of Logistics
Management, The, https://doi.org/10.1108/IJLM-04-2016-0095
Permanent link to this document:
https://doi.org/10.1108/IJLM-04-2016-0095
Downloaded on: 17 July 2017, At: 05:10 (PT)
References: this document contains references to 0 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 13 times since 2017*
Access to this document was granted through an Emerald subscription provided by emerald-srm:333301 []
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service
information about how to choose which publication to write for and submission guidelines are available for all. Please
visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of
more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online
products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication
Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.

*Related content and download information correct at time of download.


An investigation of the effects of storage assignment and picker routing on
the occurrence of picker blocking in manual picker-to-parts warehouses
Abstract
Purpose: Order picking is one of the most costly logistics processes in warehouses. As a result, the
optimization of order picking processes has received an increased attention in recent years. One
potential source for improving order picking is the reduction of picker blocking. This paper
investigates picker blocking under different storage assignment and order picker route combinations
and evaluates its effects on the performance of manual order picking processes.
Design/methodology/approach: This study develops an agent-based simulation model (ABS) for
order picking in a rectangular warehouse. By employing an ABS, we are able to study the behaviour
of individual order pickers and their interactions with the environment.
Findings: The simulation model determines shortest mean throughput times when the same routing
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

policy is assigned to all order pickers. In addition, it evaluates the efficiency of alternative routing
policies-storage assignment combinations.
Research limitations/implications: The paper implies that ABS is well suited for further
investigations in the field of picker blocking, for example with respect to the individual behaviour of
agents.
Practical implications: Based on the results of this paper, warehouse managers can choose an
appropriate routing policy that best matches their storage assignment policy and the number of order
pickers employed.
Originality/Value: This paper is the first to comprehensively study the effects of different
combinations of order picker routing and storage assignment policies on the occurrence of picker
blocking.
Keywords: picker blocking, congestion, manual order picking, routing policies, storage assignment,
agent-based simulation
Paper type: Research paper

1. Introduction

Order picking is one of the most labour- and time-consuming processes in warehouses (Frazelle, 2000;
Tompkins et al., 2010). It can be defined as the process of retrieving Stock Keeping Units (SKUs)
from storage locations to fulfil customer orders (de Koster et al., 2007), and it is a critical process
within supply chains with a direct influence on customer satisfaction. If order picking is not organized
adequately, the picking of wrong or damaged SKUs or long order picking process times can have a
negative influence on customer satisfaction (Gue et al., 2006; Parikh and Meller, 2008). Thus, efficient
order picking processes are a prerequisite for a high-performing supply chain (Chen et al., 2013).
Due to the fact that order picking is performed manually in many cases, researchers have estimated
that it accounts for about 50% of the operating costs of warehouses (Frazelle, 2000; Tompkins et al.,
2010; Grosse et al., 2015). To reduce the cost of order picking, researchers and practitioners have tried
to minimize travel times for order pickers in picker-to-parts systems to increase the throughput of the
warehouse (Chen et al., 2014; de Koster et al., 2007). Travelling, which can account for up to 55% of
the total work time of the order picker (Tompkins et al., 2010), can be influenced by assigning SKUs
to storage locations, by grouping or splitting up incoming orders, or by defining routes for the order
picker.
Another aspect that is of high importance in warehousing is space utilization. Especially in situations
where it is not possible, or, too expensive to make additional storage space available, warehouse
managers have to store as many products as possible in the existing storage facility. To increase space
utilization for order picking warehouses, so-called narrow-aisle warehouses are commonly employed
(Chen et al., 2013; Chen et al., 2014; Gue et al., 2006). Narrow-aisle warehouses have one major
disadvantage, however: aisles in such warehouses are usually too small to permit order pickers to pass
each other. The consequence is that in narrow-aisle warehouses, order pickers working in the same
zone may block each other (Davarzani and Norrman, 2015), which can put the operational advantages
of narrow-aisle warehouses at stake (Hong, 2014). “Picker blocking” or “picker congestion”, in this
context, refers to a situation where one order picker is disturbed by another order picker, such that
he/she is unable to pass the other order picker or reach a pick-column (e.g. Parikh and Meller, 2009).
Picker blocking can result in longer travel times due to idle or waiting times, and hence may reduce
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

order picking efficiency (Chen et al., 2013; Chen et al., 2014; Mowrey and Parikh, 2014; Parikh and
Meller, 2008). Due to the complex and stochastic nature of picker blocking, its influence on the
performance of order picking is difficult to quantify (Heath et al., 2013).
Although picker blocking has received some attention in the literature, there are still several research
gaps that need to be addressed. A closer look at the literature shows that most studies on picker
blocking focused mainly on the s-shape routing policy and on random storage assignment (e.g.
Mowrey and Parikh, 2014). Existing studies thus neglected the effects of the combination of storage
assignment and routing policies on the occurrence of picker blocking (e.g. Chen et al., 2013;
Kłodawski and Żak, 2013; Pan and Shih, 2008). However, as we hypothesize, different routing
policies can positively or negatively influence additional throughput times resulting from picker
blocking, depending on how they are combined with storage assignment rules. Thus, it is necessary to
quantify the effects of picker blocking on mean order throughput times for different routing policies
and storage assignment rules. Another important aspect is that most studies on routing policies and
storage assignment rules analysed the performance of alternative picker routing-storage assignment
combinations in single picker-environments only. Multi-picker investigations that study the effects of
route combinations for multiple order pickers on picker blocking, and the resulting order picking
efficiency, are still lacking.
The paper at hand aims to close these research gaps by studying the effects of routing and storage
assignment on the occurrence and the performance-impact of picker blocking. The results of the paper
support practitioners in selecting suitable combinations of picker routing and storage assignment
policies that reduce picker blocking and guarantee shorter order throughput times. In our simulation
analysis, we consider varying the number of order pickers and picks per order, as well as several
combinations of routing policies and storage assignment rules. To analyse the system’s performance in
terms of mean order throughput time, an agent-based simulation model (ABS) is developed.
The remainder of this paper is organized as follows: The next section reviews the related literature,
and Section 3 describes the developed simulation model. Section 4 presents the results obtained in the
simulation study, and Section 5 concludes the paper.

2. Literature review

To increase the performance of manual order picking, researchers have developed various planning
approaches that support managing order picking in practice. The objective of planning approaches in
this area usually is to reduce the time that is required for completing a single order, or, a set of orders,
for example by implementing order picking policies that specify the warehouse layout, the storage
assignment, picker routes, or order batches (de Koster et al., 2007). Warehouse layout planning
defines the configuration of the warehouse, including the number of aisles and cross aisles as well as
their size (Meller and Gau, 1996). A typical warehouse layout that can frequently be found in practice
has a rectangular shape and consists of one or more blocks (Chew and Tang, 1999; Roodbergen et al.,
2008). Alternative warehouse layouts that have recently been introduced are the U-shaped (Glock and
Grosse, 2012) and fishbone layouts (Çelk and Süral, 2014). Storage assignment rules define how
products should be assigned to storage locations (Glock and Grosse 2012), where the assignment can
be random or follow a certain pattern. A typical storage assignment rule is class-based storage, which
first divides SKUs into classes (e.g., A, B, C) based on their turnover, and then assigns fast-moving
classes to zones near the depot, which is also often referred to as packaging and despatch area, loading
bay, pick-up/drop-off point or input/output point (Le-Duc and de Koster, 2005). Routing policies
determine the way the order picker walks through the storage area (Petersen and Aase, 2004). In
practice, heuristic routing policies are most often used as they are easy to implement and reduce the
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

possibility of confusing the order picker by a complex tour (Gademann and van de Velde, 2005).
Finally, order batching combines or splits up customer orders when restructuring the orders will lead
to better capacity utilization (Grosse et al., 2014).
One major drawback of research on order picking is that researchers have frequently neglected
operational problems that may occur when implementing order picking policies in practice. If multiple
order pickers work in the same zone of a narrow-aisle warehouse, then they may block each other
when entering the same aisle. In such a situation, one of the order pickers either has to wait until the
other order picker leaves the aisle, or, he/she has to deviate from the original route to pick other SKUs
first, before returning to the aisle in question. Clearly, both alternatives would lead to increased
throughput times.
Picker blocking is a phenomenon that quite frequently occurs in practice, and thus it has attracted the
attention of researchers. Works in this area generally distinguish between in-the-aisle blocking and
pick-column blocking (Chen et al., 2014; Hong, 2014; Mowrey and Parikh, 2014; Parikh and Meller,
2009). In-the-aisle blocking occurs in order picking warehouses with narrow aisles where order
pickers are unable to pass one another within an aisle. Pick-column blocking, in contrast, occurs if two
or more order pickers intend to collect SKUs from the same pick-column (storage location), and it can
occur both in narrow-aisle and in wide-aisle warehouses. For narrow-aisle warehouses, Gue et al.
(2006) showed in an analytical model that a high pick density (i.e., when orders consist of many SKUs
and workers need to stop more often to perform picks) leads to less blocking. Pan and Shih (2008) and
Pan et al. (2012) proposed a performance measure in terms of travel time and blocking-caused delay
and an heuristic algorithm to balance the occurring trade-off. Furmans et al. (2009) studied picker
blocking in a rectangular warehouse using a queuing model and presented solution algorithms for s-
shape and return routing that can easily be applied in practice. Parikh and Meller (2008) considered
picker blocking in a comparative cost analysis of batch and zone picking systems and concluded that
workload-imbalance is greater in zone picking systems, even though such systems eliminate picker
blocking. In a follow-up work, Parikh and Meller (2009) developed analytical models for pick-column
blocking in wide-aisle warehouses and showed that blocking increases monotonically when the
number of SKUs at the same pick-column increases. This paper was extended by Parikh and Meller
(2010a) for the case of narrow-aisle warehouses. The authors showed that underestimating picker
blocking may put the efficiency (in terms of order picking time per order) of the process at stake. The
authors concluded that blocking is relevant when pick density increases. Parikh and Meller (2010b)
also confirmed that picker blocking becomes more frequent when fast-moving SKUs are stored near
the depot.
Hong et al. (2010, 2012a, 2012b) developed order batching algorithms that reduce picker blocking
with the intention to maximize throughput in narrow-aisle warehouses with s-shape routing. Pan and
Wu (2012) compared s-shape and return routing and found that an across-aisle horizontal storage
policy that balances the workload over all aisles leads to less frequent picker blocking. Chen et al.
(2013) developed a routing algorithm for two order pickers working simultaneously in the same
warehouse zone that minimizes picker blocking based on an ant colony optimization meta-heuristic.
The proposed algorithm can be extended to the case of more than two order pickers (Chen et al. 2014).
Heath et al. (2013) employed an agent-based simulation model and showed that extra walking time
caused by picker blocking is one of the most important drivers of travel time in the case of picker
blocking. The authors modelled random and a modified s-shape routing as well as random and
turnover-based storage assignment. They concluded that agent-based simulation is well suited to
quantify the effects of picker blocking. Hong et al. (2013) estimated increasing order processing times
that result from picker blocking in wide- and narrow-aisle warehouses. This paper was extended by
Hong (2014) to the case of a picking line. Recently, Hong et al. (in Press) extended their analysis to
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

quantify the level of picker blocking in a bucket brigade order picking system. Finally, Mowrey and
Parikh (2014) proposed a mixed-aisle layout that uses wide and narrow aisles and derived the optimal
aisles configuration for this layout taking account of both space requirement and travel time and
including picker blocking.
Our overview of the literature indicates that most studies on picker blocking focused on individual
decision problems, such as picker routing for a given storage assignment or layout planning where
routes are not optimized. Works addressing more than a single decision problem are rather rare. With
respect to routing policies, most studies employed the traversal policy (e.g., Hong et al., 2012a; Pan
and Shih, 2008; Pan et al., 2012; Parikh and Meller, 2010a; 2010b; Ruben and Jacobs, 1999). Existing
works frequently studied circular picking areas (where order pickers walk in circular shape) (e.g. Gue
et al., 2006, Hong, 2014, Hong et al., 2015; Skufca, 2005) or rectangular warehouses (e.g. Chen et al.,
2013; Chen et al., 2014; Furmans et al., 2012; Mowrey and Parikh, 2014). In addition, the assumption
that products are assigned randomly to the storage locations of the warehouse was frequently made
(e.g. Parikh and Meller, 2010a; Parikh and Meller, 2010b). The number of order pickers working in
the warehouse in parallel was also frequently varied in works on picker blocking (from 2 up to 90; see
e.g. Hagspihl and Visagie, 2014; Ruben and Jacobs, 1999); the same is the case for the number of
picks per order (e.g. up to 50 in Pan et al. 2012).
To further our understanding of how picker blocking affects the performance (in terms of mean order
throughput time) of order picking systems, and to gain insights into how the specification and
combination of operational policies may help to avoid picker blocking, this paper proposes an agent-
based simulation model that investigates picker blocking under different routing policies and storage
assignment rules. By employing an ABS model, it is possible to study the dynamic phenomenon of
congestion (blocking) in a typical manual order picking warehouse based on the single entities
(agents) to gain insights into system behaviour (bottom-up approach).

3. Simulation study
3.1. Agent-based simulation (ABS)

Investigating the causes and consequences of picker blocking in real-world environments is often very
difficult due to the many factors that could influence picker blocking – traffic or possibly the
unnoticed decisions of order pickers, especially if several blocking situations occur at the same time,
to name just two examples. Quantifying the effects of picker blocking on warehouse efficiency based
on practical observations involves the risk that important influencing factors or consequences are
ignored or misinterpreted (Heath et al., 2013). Developing analytical models is a possible remedy to
this problem. Analytical models, however, usually cannot fully capture congestion in a natural and
realistic way, which would lead to high model complexity and difficulties in solving the model.
Simulation appears to be better suited to investigate such complex and dynamic phenomena as picker
blocking.
For the reasons outlined above, an agent-based simulation model was developed. In general,
simulation is an appropriate method to investigate human behaviour, and it is especially useful in cases
where systems are complex and/or subject to uncertainty (Law, 2013). An important advantage of
simulation is the possibility to deliberately vary individual system conditions and to analyse their
influence on the model’s performance. In practice, such comprehensive analyses are often not
possible, especially since individual system conditions often cannot be varied in isolation.
For ABS, there is no generally accepted definition (Borshchev and Filippov, 2004; Law, 2013). ABS
models can best be characterized along three model components (Borshchev and Filippov, 2004;
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Macal and North, 2010): a set of agents, their relationships (e.g. interactions between agents), and their
environment. Agents are single entities that can act autonomously, and they are also able to learn and
adapt their behaviour. Agents can represent people, animals, companies, or other entities such as cells.
They are modelled by a set of rules that define their behaviour, i.e. they can act autonomously
according to their implemented behaviour. Another feature of ABS is the possibility to interact with
other agents and/or their environment. ABS is a bottom-up approach where the system under study is
investigated via single rules for each entity, instead of implementing the behaviour of the whole
system in advance. This helps to capture emergent phenomena, i.e. properties or the behaviour of the
system cannot be attributed to properties of the single entities; the behaviour emerges over time due to
the interactions of the single agents.
Law (2013) defined three criteria for situations where ABS should be used, namely I) situations where
the single entities naturally interact, II) where it is important to learn or adapt their behaviour, or III)
where movements of entities depend on the situation or the environment.
In accordance with the objective of this paper, which is to study the effects of picker routing and
storage assignment policies on the occurrence and the performance-impact of picker blocking, single
entities/agents (here: order pickers) working on individual pick lists have to be created in the ABS
model. In addition, each agent has to follow a determined routing policy. During the order picking
process, blocking can occur. In such situations, the agents have to decide autonomously whether they
want to wait until the other order picker has finished his/her picking task, whether they prefer to
follow the other agent until the path is clear, or if they wish to walk out of the aisle to clear the way.
During the order picking process, blocking situations occur dynamically depending on several
parameters, such as the number of agents or picks per order. Furthermore, picker blocking depends on
the environment, i.e. the characteristics of the order picking warehouse (e.g. storage assignment or
width of aisles).
Several software solutions support the development of ABS (Heath et al., 2009; Robinson and Ding,
2010). We decided to use the software tool AnyLogic, as it is appropriate to simulate persons (active
objects) (Borshchev and Filippov, 2004) and offers a high level of flexibility due to a set of predefined
libraries that can be customized with additional classes and functions based on Java-Code (Borshchev
and Filippov, 2004; Macal and North, 2010). The behaviour of each agent can be implemented with
state charts, which describe the behaviour of agents in states and transitions. State charts are
comparable to UML (Unified Modelling Language) diagrams and offer a good trade-off between
simplicity (predefined functionality) and complexity (extensions in Java) (Borshchev and Filippov,
2004).
3.2. Problem description and parameters (conceptual model)

We assume a standard rectangular warehouse, which is a layout that is common in practice and that
has often been studied in the literature (Bassan et al., 1980; Jarvis and McDowell, 1991; Hsieh and
Tsai, 2006). The warehouse under study consists of 10 narrow aisles that are 1m wide (see also Grosse
et al., 2014). Each storage position (width: 0.5m; depth: 1m) in the shelf contains a single type of
SKU, and 1000 SKUs (100 in each aisle) are stored in the warehouse in total. Additional areas at the
upper and lower ends of the aisles are added where the order pickers can wait if the aisle entry is
blocked. We refer to this areas as the “waiting zone”.
The replenishment of shelves is excluded from the analysis, which is a common assumption as
replenishments are often made during idle times (de Koster et al., 2007). It is assumed that order
pickers start at the depot and return to the depot after all SKUs contained on the pick list have been
collected. Travel speed is set constant at 0.75 meters per second, and the time to pick an SKU is 20
seconds (Gue et al., 2006; Pan et al., 2012). The number of order pickers working in the warehouse is
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

varied (2, 5, 10, and 15 agents). To reduce congestion in front of the first aisle, the next agent starts
collecting SKUs when the order picker that started picking earlier arrives at the first aisle (i.e., after
18.5 seconds) (Chen et al., 2013). The path on which order pickers walk through the warehouse is
defined by the routing policy used. Seven routing policies are implemented: return (Figure 1a), s-
shape (1b), midpoint (1c), largest gap (1d), composite (1e), combined (1f), and optimal (1g). These
routing policies are either common in practice or have frequently been studied in the literature on
manual order picking (e.g., Ratliff and Rosenthal, 1983; Caron et al., 1998; Petersen, 1999; Petersen
and Schmenner, 1999; Petersen and Aase, 2004; Roodbergen and de Koster, 2001).

Figure 1: Overview of the implemented routing policies

The following four storage assignment rules are studied: random (Figure 2a), class-based (diagonal
(2b), vertical (2c), and horizontal (2d)) (Pan and Wu, 2012). For class-based storage assignment, the
storage area is divided into three zones: Zone A for SKUs accounting for 80% of the demand (fast-
moving SKUs), zone B for SKUs accounting for 15% of the demand, and zone C for SKUs accounting
for only 5% of the demand (slow-moving SKUs) (Petersen and Aase, 2004; Ruben and Jacobs, 1999).
A-SKUs are located near the depot to keep travelling distances short (20% of all products). The area
for B-SKUs contains 20% of all SKUs and the area for C-SKUs the remaining 60%. The only
difference exists for the diagonal ABC storage assignment rule, where the distribution is 20% (A),
30% (B), and 50% (C). Within a class, SKUs are stored randomly (Petersen et al., 2004).
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Figure 2: Overview of the implemented storage assignment rules (a: random; b: ABC diagonal; c: ABC vertical; d:
ABC horizontal; A-SKUs are located in black areas, B-SKUs in dark grey areas and C-SKUs in light grey areas)

Customer orders contain 10 or 20 picks, and they are assigned to a single order picker. A list of 1000
orders for each agent is randomly generated and kept constant during all simulation runs for each
storage assignment policy and each specified amount of picks per order.
Agents, which are entities of the simulation model, represent order pickers who retrieve SKUs on
assigned order lists. Each SKU is identified via a part number which also determines the storage
location of the SKU. For each agent, 1000 orders, each order with 10 or 20 of those part numbers, are
created, assigned to an order picker and written in a data sheet. When order lists already exist, the
existing order lists are used (read out from a data sheet) instead of creating new ones during every
replication.
The order pickers interact with the environment by walking on predefined paths through the
warehouse (see dotted lines in Figure 2), where the paths depend on the order list and routing policy.
Picker blocking occurs if the distance between two order pickers becomes less than 0.5 meters. Picker
blocking thereby only occurs within aisles, i.e. if pickers walk into opposite directions, in the same
direction with another worker performing picks, or in case several order pickers try to approach the
same storage location. If pickers walk into the same direction without retrieving SKUs within that
aisle, no picker blocking occurs. Agents can traverse all aisles in both directions (according to their
routing policy and depending on potential blocking situations that may occur), and blocking
exclusively occurs in aisles (i.e. passing within an aisle is not possible). However, passing in the cross
aisle (front or back) is possible. For solving a blocking situation and to avoid deadlocks, the order
pickers are able to leave the aisles on both sides. If blocking occurs, the order picker with the least
distance to the next pick in an aisle or to the aisle exit gets the highest priority. All other order pickers
either have to clear the way and leave the aisle on one side (and wait outside the aisle), or they have to
follow the order picker with highest priority until s/he arrives at the next target location (the next pick
in an aisle or the aisle exit). After the blocking situation has been resolved, all other order pickers
pursue their initial targets (in other words: during a blocking situation, the order picking processes of
all order pickers with lower priority is interrupted to enable the order picker with the highest priority
to complete his/her order. Once the blocking situation has been resolved, the order pickers continue
their original order picking process). Figure 3 illustrates this order of events as implemented in the
simulation model.
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Figure 3: Order of events as implemented in the simulation model with the focus on the occurrence of blocking

As performance measures, the simulation model determines the mean throughput times needed to
fulfil a single order for each agent (e.g. Davarzani and Norrman, 2015; Chen et al., 2013). In this
context, order throughput times consist of waiting/idle times, pick times as well as travel times. The
time count for a simulation run that is required for calculating the mean throughput times starts with
the despatch of the first order picker, and it stops once all agents arrive at the depot (see, for a similar
setting, Chen et al., 2013). Subsequently, the next simulation run is initiated until a predefined number
of cycles (1000) has been reached. Mean order throughput times are determined for the case with
blocking and also for the case where blocking is not considered. In case blocking is neglected, each
order picker is able to retrieve its own order list without interacting with other workers (i.e., it is
assumed that aisle and pick-columns are wide enough to permit order pickers to pass each other). The
simulation study first derives mean order throughput times for the case where all agents use the same
routing policy. Afterwards, route combinations (i.e., different routing policies for different order
pickers during a single simulation run) are analysed. Table 1 summarizes all factors considered in the
simulation studies.
[Include Table 1 here]
3.3. Validation
Simulation models need to be validated along their entire lifecycle (Balci, 1998; Klügl, 2008; Law,
2013; Sargent, 2013). Validation is an individual task and depends on the data availability or purpose
of the simulation study. Hence, there is no single framework or predefined order of activities that can
be employed in the same sequence for every simulation study (Sargent, 2013). Balci (1998) provided a
comprehensive overview including 77 methods to increase validity of a simulation model. These
methods are frequently divided into qualitative or subjective assessments and quantitative or statistical
testing (e.g. Heath et al., 2009; Klügl, 2008; Sargent, 2013). Additionally, validation can be divided
into validation of a conceptual model, a computer model, and simulation results. The process is
iterative, which leads to adaptions at different stages whenever necessary (e.g. Balci, 1998; Heath et
al., 2009; Klügl, 2008; Law, 2013; Sargent, 2013).
For the paper at hand, the conceptual model, as described in Section 3.2, was validated in a first step.
Thus, the authors developed and discussed the conceptual model and based its assumptions on relevant
literature in the field of manual order picking and picker blocking. Structured walk-throughs were
performed to ensure the validity of the assumptions. During all stages, warehousing experts or,
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

respectively, managers and domain experts (Klügl, 2008), were consulted. We discussed the
simulation model as well as the underlying assumptions with warehousing experts during site walks at
four companies where the storage area for the distribution of goods was visited. The visited companies
produce and sell: chemical products, coatings, measurement technologies, and tools. Formal and
informal talks were held with logistics managers (6 persons) and order pickers (9 persons).
Additionally, a workshop was organized to present and discuss the output data and results. 33
representatives (e.g., head of logistics, warehouse manager, and plant manager) from different
companies and industrial sectors, such as automotive, food, and pharma industries as well as logistics
service providers or consultants, participated in the workshop. The warehousing experts agreed with
the conceptual and computer model and only minor changes were proposed. If appropriate, they were
included in the simulation model.
For validation of the computer model, techniques described in the literature were employed.
Qualitative methods, such as animation or structured walk-throughs, as well as quantitative methods
(i.e., statistical methods) were used to improve the validity of the simulation model. To improve the
modelling accessibility, a graphical user interface was implemented, including a graphical
representation of the storage area with different agents and picking positions. The implementation of
intuitive routing policies, such as the s-shape or return policy, could easily be analysed this way. Each
agent and the products on the associated order list were assigned a colour to follow the order picking
process. Furthermore, the trace function in AnyLogic was used to guarantee the structural validity with
the correct execution and calculation of the simulation model and to track the correct execution of the
internal structure of the model implemented via the individual functions.
Finally, also the output data was validated. For this purpose, several parameters were changed to
verify if the simulation model behaved as intended (e.g., increase the number of agents or picks per
order). Thereby, the input-output behaviour of the simulation model could be evaluated. As in the
results sections, means with confidence intervals were used to check differences in mean throughput
times under varying parameters. Furthermore, the throughput times for different routing policies were
compared (e.g., optimal policy vs. combined policy to proof if the throughput times for the optimal
policy are always shorter or at least as good as the throughput times for the combined policy).
Some other validation techniques were not considered relevant to this study, e.g. Turing test, or in
general the comparison between model and real-world system results, as the model was not based on a
specific real-world system. However, for the paper at hand, the results obtained using the simulation
model can indirectly be compared to real-world systems by referring to the results of other studies
with similar assumptions that are based on the observation of real-world systems.
4. Results
4.1 The same routing policy is assigned to all order pickers

In a first step, we analysed the case where all order pickers use the same routing policy to travel
through the warehouse. This setup can often be observed in practice, as it is much easier to assign the
same routing policy to all order pickers than to decide which policy to assign to which order picker. At
first, the results for mean order throughput times are presented; afterwards, the average number of
blockings for a single simulation run and the comparison of parameter combinations with a
benchmarking case follow.
Evaluation of mean order throughput times
The simulation model determined the shortest mean order throughput times for the optimal routing
policy with vertical ABC storage and random storage assignment. The optimal routing policy was
originally designed only for single picker operations (Ratliff and Rosenthal, 1983), but it can also lead
to shortest mean throughput times in comparison to the other six investigated routing heuristics when
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

several order pickers work in the same storage area and in case picker blocking occurs. This is the case
for 10 as well as 20 picks per order. Only in one case (for 15 order pickers and 10 picks per order), the
random storage assignment led to shorter mean throughput times than vertical ABC storage
assignment in combination with the optimal routing policy. Results suggest that using only a single
routing policy is beneficial even when the number of order pickers or the size of order lists are
changed. Table 2 gives an overview of mean order throughput times for varying parameters, where the
number of picks per order is 10 or 20 and the number of agents picking simultaneously in the
warehouse is varied between 2, 5, 10, and 15. Note that all confidence intervals (two-sided, 95%) are
not higher than 0.4% of the respective mean throughput times.
[Include Table 2 here]
The vertical ABC storage assignment (in combination with the optimal routing policy) led to the
shortest mean throughput times for almost all considered cases, except for the case with 10 picks per
order and 15 order pickers, where the random storage assignment performed best. ABC storage
assignment rules typically lead to busy areas (e.g., in areas with fast moving products), in which
articles are stored that appear with a high probability on most of the order lists. Compared to other
combinations of routing policies and storage assignment rules investigated in this study, the negative
influence of such busy areas on the mean order throughput time seems to be low for a vertical ABC
distribution when the order pickers follow the optimal routing policy. This is due to the highly
individual design of routes created with an optimal policy. The highest number of blockings occurred
in the area with A-products; for the other zones, the order pickers spread over the remaining storage
area and the additional number of blockings was low. As the simulation progresses, the initially high
number of blockings decreases. At the same time, also short travel distances are obtained as the
optimal routing algorithm is used. Hence, a low number of blockings as well as short mean throughput
times result for the optimal policy. Only in one case, the mean order throughput times were lower for
random assignment combined with the optimal routing policy as compared to the combination of
vertical ABC storage assignment and the optimal routing policy. This case occurred for 10 picks per
order and 15 order pickers. The possibly shorter ways, which result for the vertical ABC storage
assignment in comparison to a random storage assignment, lead to a higher number of blockings
because of the busy areas with fast moving products (see black areas in Figure 2). This changes the
relative advantageousness of the combination with optimal routing policy and vertical ABC storage
assignment in favour of the combination with optimal routing policy and a random storage
assignment. In case of 20 picks per order and 10 or 15 order pickers, the mean number of blockings
also decreases for a random storage assignment in comparison to a vertical ABC storage assignment
for the optimal policy. Nevertheless the lowest mean throughput times were determined for the latter
case.
Mean order throughput times for picking an average order with 10 or 20 picks are illustrated in
Figures 4 and 5. As can be seen, the mean order throughput times increase in the number of agents.
Not all combinations with the optimal routing policy led to the shortest mean throughput times (e.g.,
optimal routing policy combined with diagonal ABC storage assignment (713 sec) vs. largest gap
policy combined with random storage assignment (546 sec) for 10 picks per order and 15 order
pickers). As can be seen in Figures 4 and 5, some routing policies should not be combined with
particular storage assignment rules (e.g., vertical ABC storage assignment should not be combined
with the return policy, and the s-shape policy should not be combined with horizontal ABC storage
assignment) to avoid longer mean throughput times. In some cases, e.g. for the combination of a
horizontal ABC storage assignment rule and the s-shape policy, it could be expected that this
combination would not be useful if the majority of products were spread evenly over the lower entries
of all aisles and the order pickers still have to traverse the complete aisle. A better combination seems
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

to be the s-shape policy with a vertical ABC storage assignment, where the majority of products are
distributed over a smaller number of aisles. The results indicate, however, that this only holds for up to
5 order pickers. The average number of blockings for the vertical ABC storage assignment rule in
combination with the s-shape policy is generally higher as compared to a horizontal ABC storage
assignment, but the mean order throughout times for a vertical ABC storage assignment increase in
cases of 10 and 15 order pickers (both for 10 and 20 picks per order). Thus, the mean order throughput
times for a horizontal ABC storage assignment rule combined with the s-shape policy and for 10 or 15
order pickers are lower as compared to the vertical ABC storage assignment when blocking is
considered.

Figure 4: Overview of mean throughput times for 10 picks per order


Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Figure 5: Overview of mean throughput times for 20 picks per order

Figure 6 illustrates different options for order pickers for retrieving products from a single aisle. For
easy-to-follow heuristics, such as s-shape, only options (a) and (e) can be realised. For the return
policy, only options (b) or (c) and (e) are suitable. Especially for the return policy, a bottleneck may
occur because aisles can only be left on the side where the aisles were entered. Consequently, some
routing heuristics are more flexible than others and therefore possibly result in lower mean throughput
times. If an order picker who intends to leave an aisle has to follow a worker who has to pick several
SKUs on his/her way to the aisle exit, mean order throughput times may increase especially when
following an s-shape policy. Even if it is possible to skip aisles when following an s-shape policy, this
can lead to a new blocking situation if another worker in the aisle to be entered walks in the opposite
direction of the order picker in question. Flexible policies, such as the optimal routing policy that
contains components of all other investigated routing policies, provide more opportunities for
individual routes (see Figure 6 configurations (a) – (e) are possible for the optimal policy). Hence,
more flexible routing policies (e.g. combined or optimal policy) seem to be more robust against
additional picking times due to picker blocking compared to other policies, such as the return or s-
shape policies, with only few exceptions for higher numbers of order pickers (15), picks per order
(20), and the s-shape policy.
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Figure 6: Pick configurations for one aisle (a: traverse completely; b: enter and exit from the upper side; c: enter and
exit from the lower side; d: enter and exit from the upper and lower side; e: skip aisle) (Ratliff and Rosenthal, 1983)

Evaluation of the average number of blockings


The average number of blockings, which are depicted in Figures 7 and 8, show that blockings strongly
increase in the number of order pickers and picks per order. The simulation model determined the
highest number of blockings for vertical and diagonal ABC storage assignment regardless of the
routing policy used (for 10 and 20 picks per order). With two exceptions (largest gap policy, random
storage assignment and 10 or 15 order pickers), the optimal routing policy led to the lowest mean
number of blockings. The lowest mean number of blockings mostly occurred in combinations with the
random storage assignment rule, which utilizes all areas of the warehouse with the same intensity.
Nevertheless, the combinations with a low mean number of blockings do not necessarily lead to the
shortest mean throughput times. For example, this is the case for the combination of midpoint policy
with a random storage assignment rule. The midpoint policy seems to be more robust as compared to
other investigated routing policies against negative effects of blocking in some cases (e.g. for 10 picks
per order). The reason for the robustness of the midpoint policy is that it divides the storage area into
two halves (e.g. upper and lower) where the order pickers can retrieve SKUs, and due to the random
storage assignment the SKUs are almost evenly distributed over the whole storage area. However, this
routing policy usually leads to longer mean throughput times (also when blocking is neglected) as
compared to other heuristics, such as the composite or combined policy. Thus, the mean throughput
times for the midpoint policy with a comparatively low number of blockings are higher in comparison
to the composite or combined policy, even though these policies mostly caused a larger number of
blockings. Our results indicate that a low number of blockings is not the only objective for achieving
shortest mean throughput times. Choosing a routing policy which leads to short mean throughput
times for single picker operations (without blocking) also seems to be important.
Figure 7: Average number of blockings during a single simulation run with 10 picks per order
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Figure 8: Average number of blockings during a single simulation run with 20 picks per order

Evaluation of mean throughput times in comparison to a benchmarking case


The benchmarking case considered in this section assumes a combination of the optimal routing policy
and vertical ABC storage assignment (with one exception: for 15 order pickers and 10 picks per order,
random storage assignment is used). These are the combinations which led to the lowest mean
throughput times. Figures 9 and 10 illustrate the percentage increase in mean order throughput times
for different combinations of storage assignment and routing policies in comparison to this
benchmarking case. As can be seen, the selection of a routing policy should depend on the storage
assignment rule to achieve shorter throughput times. If a routing policy that does not match the
requirements of a given storage assignment is used, this can lead to a dramatic increase in mean order
throughput times when blocking is considered (for the effects of combining routing policies with
different storage assignment rules without blocking see Petersen, 1999). This is the case even for
combinations involving the optimal routing policy, which cannot guarantee shortest mean throughput
times if used together with several storage assignment rules. The increase in mean throughput times
can be up to 103% (return policy, vertical ABC storage assignment, 15 order pickers, and 20 picks per
order) in comparison to the benchmarking case.
These results indicate that the impact of picker blocking on mean order throughput times varies
considerably if important parameters, such as the size of the order list, the number of order pickers, the
routing policy or storage assignment in use, are altered. For example, the horizontal ABC storage
assignment seems to be quite robust to picker blocking, especially for the case with 10 picks per order
(for varying numbers of order pickers). This result changes if order lists contain 20 picks. In case of 10
order pickers, the random storage assignment rule would be recommended. Therefore, it seems
important to determine important warehouse parameters (e.g., by analysing those parameters by
employing ABS approaches) before implementing a routing policy or storage assignment rule to
reduce the extent of blocking.
The s-shape policy, which has frequently been used as a benchmark in the order picking literature
(Chen et al., 2013) and which is frequently used in practice (Hwang et al., 2004), led to longer mean
throughput times for most of the combinations investigated in our study, particularly for 10 picks per
order and in comparison to the optimal routing policy (see Table 2). Hence, it was not considered as a
benchmarking case in the study at hand. If order pickers are supposed to use the s-shape policy (e.g., if
determined by the warehouse manager), the percentage increase in mean order throughput times is
between 14.82% and 48.96% (for 10 picks per order) as well as between 14.18% and 57.13% (for 20
picks per order) in comparison to the benchmarking case.
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Figure 9: Percentage increase in mean order throughput times for different combinations of routing policies and
storage assignments in comparison to the benchmarking case for 10 picks per order

Figure 10: Percentage increase in mean order throughput times for different combinations of routing policies and
storage assignments in comparison to the benchmarking case for 20 picks per order

4.2 Different routing policies are assigned to the order pickers

In a second step, the effect of assigning different routing policies to different order pickers during the
same simulation run (route combinations) was evaluated to investigate the effects of picker blocking
when route combinations are employed. In the simulation study, the number of agents in the
warehouse was limited to two and three, as considering more order pickers would dramatically
increase the number of picker-route combinations, which would make it difficult to present the results
in a concise way. Additionally, the focus is on random storage assignment, which best utilizes the
storage area, and 20 picks per order (with 1000 randomly created orders). All seven routing policies
are investigated: return (R), s-shape (S), midpoint (M), largest gap (L), composite (Cs), combined
(Cb), and optimal (O).

Route combinations for two order pickers


First, route combinations for two order pickers are analysed. When blocking is not considered (i.e.,
wide aisles are assumed; see Table 3), combinations with the return policy mostly led to longest mean
throughput times (this could probably change if a horizontal class-based storage assignment was
implemented). Combinations with the optimal policy (for one of the order pickers) led to shortest
mean throughput times (except for the combinations optimal – return, return – optimal, optimal – s-
shape, s-shape – optimal).
Additionally, the simulation model determined the mean order throughput times for route
combinations for two agents with picker blocking (see Table 4). In this case, the combination where
both order pickers follow the optimal policy led to the shortest mean throughput times (ca. 730 sec;
confidence interval = ca. 2 sec). All combinations where at least one of both order pickers travels
according to the optimal policy resulted in shorter mean throughput times as compared to other route
combinations (see dark grey areas in Table 4). In the case of two order pickers, the results indicate no
need to employ route combinations because the shortest mean throughput time is determined for the
combination: optimal – optimal. However, route combinations are well suited if only heuristics are
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

used to guide the order pickers through the storage area. For example, using the combined policy
together with the largest gap policy led to shorter mean throughput times in comparison to the case
where both order pickers follow the combined or the largest gap policy. It made at least a slight
difference if order picker #1 or order picker #2 followed a certain policy (e.g. largest gap – midpoint
performs better than midpoint – largest gap). The combination of routing policies that led to the
longest mean throughput time when blocking is considered is the combination where both agents
follow a return policy (ca. 968 sec; confidence interval = ca. 4 sec). Furthermore, it can be seen that
most combinations that employ the return policy (for agent #1 or agent #2) led to longer mean
throughput times as compared to route combinations without return policy.
Comparing the differences between the mean throughput times with and without blocking, it can be
shown that combinations with the smallest differences (e.g. combination return – optimal with ca. 8
sec) do not lead to shortest mean throughput times. Although the combination with the optimal routing
policy for both order pickers led to shortest mean throughput times (for both cases with and without
blocking), the differences between mean throughput times with and without blocking is relatively high
(ca. 23 sec).
[Include Tables 3 and 4 here]
Route combinations for three order pickers
In the case of three order pickers with blocking, the simulation model determined the shortest mean
order throughput time for the combination with optimal (agent #1), combined (agent #2), and optimal
policy (agent #3) (see Figure 11). Between the top ten lowest mean throughput times, there is only one
combination that employs only a single routing policy, which is the optimal routing policy. Thus,
employing route combinations is, in most cases, favourable. However, the increase in mean
throughput times between the route combination that led to lowest mean throughput times (optimal –
combined – optimal) and the route combination with only a single route policy (optimal) is only about
0.5%. Decreasing the mean throughput times only about 0.5% is negligible in comparison to the
possible decrease of mean throughput times by changing the routing policy or storage assignment rule
(see Figures 9 and 10). Hence, only a slightly improved mean throughput time may be achieved by
employing more than one routing policy for three order pickers in comparison to the best performing
route combination consisting only of a single routing policy. Additionally, the first 10 route
combinations showed only an increase in the mean throughput times of less than 1% as compared to
the route combination with the lowest mean throughput times. Again, the longest mean throughput
times were determined for route combinations employing the return policy (e.g. return – return – s-
shape or return – return – return). The combination that led to the longest mean throughput times is
about 41% higher as compared to the combination leading to shortest mean throughput times.
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Figure 11: Overview of mean throughput times for the top ten route combinations considering the case of three order
pickers (O – optimal policy; Cb – combined policy; Cs – composite policy; L – largest gap policy)

Note that all confidence intervals (two-sided; 95%) differ not more than 0.53% (two order pickers) and
0.59% (three order pickers) from the mean with blocking and 0.51% (two order pickers) and 0.4%
(three order pickers) from the mean without blocking.

5. Discussion and Conclusion

This paper investigated the effects of picker blocking on mean order throughput times in a rectangular
manual order picking warehouse. Seven routing policies and four storage assignment rules as well as
different numbers of order pickers and picks per order were analysed to quantify the effects of picker
blocking on the mean order throughput time in an agent-based simulation model. The results of the
paper indicate that only the optimal policy led to shortest mean throughput times when one routing
policy is assigned to all order pickers. Furthermore, the effects of picker blocking for route
combinations, where different routing policies are assigned to different order pickers during one
simulation run, were analysed. While in case of two order pickers, both should follow an optimal
routing policy, results show that for three agents, numerous route combinations outperform (e.g.
optimal – combined – optimal; optimal – optimal – combined; combined – optimal – optimal) route
combinations where a single routing policy is implemented for all order pickers (including the optimal
one).
As an implication for practice, it is important to consider picker blocking when planning order picking
operations, as its impact on order throughput time is significant. For example, it was found that the
combined routing policy minimizes the mean throughput time without blocking as compared to other
investigated policies (except the optimal routing policy), but it led to longer mean throughput times,
e.g. compared to the largest gap or the composite policy, when picker blocking is considered. The
paper also showed that more complex policies, such as the optimal routing policy, are recommended
also in situations where picker blocking occurs. The longer mean throughput times determined with
the s-shape policy without considering blocking increase if picker blocking is taken into account. To
further decrease throughput times for order picking, route combinations should be employed in case of
three order pickers. Here, route combinations with the optimal and the combined policy led to the
lowest mean throughput times. The simulation model also determined shortest mean throughput times
when both order pickers follow the optimal routing policy for two order pickers as compared to other
route combinations for the same scenario.
As for managerial insights, this paper supports warehouse managers in choosing a suitable routing
policy in light of the employed storage assignment and the number of order pickers that minimizes
order throughput time. Practitioners should concentrate on implementing the optimal routing policy
combined with vertical ABC storage assignment to reduce the negative effects of picker blocking
especially for a varying number of order pickers in the storage area (2 up to 15). Even if the optimal
routing policy is not very common in practice today (de Koster et al., 2007), because it is not
commonly known (de Koster and van der Poort, 1998) and can possibly result in some problems such
as confusion of order pickers (e.g. Gademan and van de Velde, 2005; Petersen, 1999; Petersen and
Schmenner, 1999), this routing policy seems to be well suited to achieve lowest mean throughput
times even if picker blocking occurs. Furthermore, warehouse manager have to decide if the
investment for implementing a new storage assignment rule should be done or if only the routing
policy should be adapted to suit the special warehouse parameter configuration (size of order lists,
number of order pickers, etc.). At first, a detailed analysis on warehouse parameters is required before
changing a routing policy or storage assignment rule due to the highly variable results that fluctuate
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

for altered warehouse configurations.


This study has several limitations. First, we investigated only a standard warehouse layout. Thus,
analysing alternative layouts, e.g. with cross aisles, would be interesting to investigate whether other
layouts reduce picker blocking. Furthermore, the analysis of route combinations could be extended to
a higher number of agents, picks per order, or storage assignment rules. This would dramatically
increase computational effort due to a high number of possible picker-route combinations.
Additionally, a sensitivity analysis could quantify the influence of other important factors (e.g., speed,
acceleration/deceleration) on performance measures, such as the mean order throughput time.
The results of this study have important implications for future research. Other storage assignment
rules could be implemented to study the effects of picker blocking under route combinations. Such
strategies usually lead to busy areas where blocking could become a big operational problem. An
agent-based simulation approach allows studying several other influencing factors such as blocking at
the depot or the individual behaviour when two or more order picker are involved in a blocking
situation. Finally, also different sizes of order lists or shelf replenishment could be considered in future
studies. A simulation study could also be advanced by including real-world data. These limiting
factors can be addressed through further research.

Acknowledgements
The authors wish to thank the guest editor Mr. Peter McCullen and two anonymous reviewers for their
valuable comments and suggestions that helped to improve an earlier version of this paper.

References

Balci, O. (1998), “Verification, validation, and accreditation”, paper presented at 30th conference on
Winter Simulation, 1998, Washington DC , available at: http://www.informs-
sim.org/wsc98papers/006.PDF (accessed 07 July 2016).
Bassan, Y., Roll, Y. and Rosenblatt, M. J. (1980), “Internal layout design of a warehouse”, AIIE
Transactions, Vol. 12 No. 4, pp. 317-322.
Borshchev, A. and Filippov, A. (2004), “From system dynamics and discrete event to practical agent
based modeling: reasons, techniques, tools”, paper presented at the 22nd International Conference
of the System Dynamics Society, 2004, Oxford, available at:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.511.9644&rep=rep1&type=pdf
(accessed 07 July 2016).
Caron, F., Marchet, G. and Perego, A. (1998), “Routing policies and COI-based storage policies in
picker-to-part systems”, International Journal of Production Research, Vol. 36 No. 3, pp. 713-
732.
Chen, F., Wang, H., Qi, C. and Xie, Y. (2013), ”An ant colony optimization routing algorithm for two
order pickers with congestion consideration”, Computers & Industrial Engineering, Vol. 66 No.
1, pp. 77-85.
Chen, F., Wang, H., Xie, Y. and Qi, C. (2014), “An ACO-based online routing method for multiple
order pickers with congestion consideration in warehouse”, Journal of Intelligent Manufacturing,
Vol. 27 No. 2, pp. 389-408.
Chew, E. P. and Tang, L. C. (1999), “Travel time analysis for general item location assignment in a
rectangular warehouse”, European Journal of Operational Research, Vol. 112 No. 3, pp. 582-
597.
Çelk, M. and Süral, H. (2014), “Order picking under random and turnover-based storage policies in
fishbone aisle warehouses”, IIE Transactions, Vol. 46 No. 3, pp. 283-300.
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Davarzani, H. and Norrman, A. (2015), “Toward a relevant agenda for warehousing research:
literature review and practitioners’ input”, Logistics Research, Vol. 8 No. 1, pp. 1-18.
De Koster, R., Le-Duc, T. and Roodbergen, K. J. (2007), “Design and control of warehouse order
picking: a literature review”, European Journal of Operational Research, Vol. 182 No. 2, pp.
481-501.
De Koster, R. and van der Poort, E. (1998), “Routing orderpickers in a warehouse: a comparison
between optimal and heuristic solutions”, IIE transactions, Vol. 30 No. 5, pp. 469-480.
Furmans, K., Huber, C. and Wisser, J. (2009), ”Queueing models for manual order picking systems
with blocking”, Logistics Journal, Vol. 1 No. 1, pp. 1-16.
Frazelle, E. H. (2000), Trends in Transportation, Logistics Management Series, Georgia.
Gademann, N. and van de Velde, S. (2005), “Order batching to minimize total travel time in a parallel-
aisle warehouse”, IIE transactions, Vol. 37 No. 1, pp. 63-75.
Glock, C. H. and Grosse, E. H. (2012), “Storage policies and order picking strategies in U-shaped
order-picking systems with a movable base”, International Journal of Production Research, Vol.
50 No. 16, pp. 4344-4357.
Grosse, E. H., Glock, C. H. and Ballester-Ripoll, R. (2014), “A simulated annealing approach for the
joint order batching and order picker routing problem with weight restrictions”, International
Journal of Operations and Quantitative Management, Vol. 20 No. 2, pp. 65-83.
Grosse, E. H., Glock, C. H., Jaber, M. Y. and Neumann, W. P. (2015), “Incorporating human factors
in order picking planning models: framework and research opportunities”, International Journal
of Production Research, Vol. 53 No. 3, pp. 695-717.
Gue, K. R., Meller, R. D. and Skufca, J. D. (2006), “The effects of pick density on order picking areas
with narrow aisles”, IIE Transactions, Vol. 38 No. 10, pp. 859–868.
Hagspihl, R. and Visagie, S. E. (2014), “The number of pickers and stock-keeping unit arrangement
on a unidirectional picking line”, South African Journal of Industrial Engineering, Vol. 25 No. 3,
pp. 169-183.
Heath, B., Hill, R. and Ciarallo, F. (2009), “A survey of agent-based modeling practices (January 1998
to July 2008)”, Journal of Artificial Societies and Social Simulation, Vol. 12 No. 4, pp. 9-44.
Heath, B., Ciarallo, F. and Hill, R. (2013), “An agent-based modeling approach to analyze the impact
of warehouse congestion on cost and performance”, International Journal of Advanced
Manufacturing Technology, Vol. 67 No. 1, pp. 563-574.
Hong, S., Johnson, A. L. & Peters, B. A. (2010) “Analysis of picker blocking in narrow-aisle batch
picking”, paper presented at the International Material Handling Research Colloquium (IMHRC),
2010, Charlotte, available at: http://www.mhi.org/downloads/learning/cicmhe/colloquium/
2010/peters.pdf (accessed 03 November 2015).
Hong, S., Johnson, A. L. and Peters, B. A. (2012a), “Large-scale order batching in parallel-aisle
picking systems”, IIE Transactions, Vol. 44 No. 2, pp. 88-106.
Hong, S., Johnson, A. L. and Peters, B. A. (2012b), “Batch picking in narrow-aisle order picking
systems with consideration for picker blocking”, European Journal of Operational Research,
Vol. 221 No. 3, pp. 557-570.
Hong, S., Johnson, A. L. and Peters, B. A. (2013), “A note on picker blocking models in a parallel-
aisle order picking system”, IIE Transactions, Vol. 45 No. 12, pp. 1345-1355.
Hong, S. (2014), “Two-worker blocking congestion model with walk speed m in a no-passing circular
passage system”, European Journal of Operational Research, Vol. 235 No. 3, pp. 687-696.
Hong, S., Johnson, A. L. and Peters, B. A. (2015), “Quantifying picker blocking in a bucket brigade
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

order picking system.”, in International Journal of Production Economics, Vol. 170 Part C, pp.
862-873.
Hsieh, L. F. and Tsai, L. (2006), “The optimum design of a warehouse system on order picking
efficiency”, The International Journal of Advanced Manufacturing Technology, Vol. 28 No. 5-6,
pp. 626-637.
Jarvis, J. M. and McDowell, E. D. (1991), “Optimal product layout in an order picking warehouse”,
IIE Transactions, Vol. 23 No. 1, pp. 93-102.
Kłodawski, M. and Żak, J. (2013), “Order picking area layout and its impact on the efficiency of order
picking process”, Journal of Traffic and Logistics Engineering, Vol. 1 No. 1, pp. 41-46.
Klügl, F. (2008), “A validation methodology for agent-based simulations”, paper presented at the 2008
ACM Symposium on Applied Computing, 2008, New York, available at:
http://dl.acm.org/citation.cfm?id=1363696 (accessed 20 October 2015).
Law, A. M. (2013), Simulation Modeling and Analysis, McGraw-Hill Education, New York.
Le-Duc, T. and de Koster, R. (2005), “Travel distance estimation and storage zone optimization in a 2-
block class-based storage strategy warehouse”, International Journal of Production Research,
Vol. 43 No. 17, pp. 3561-3581.
Macal, C. M. and North, M. J. (2010), “Tutorial on agent-based modelling and simulation.”, Journal
of Simulation, Vol. 4 No. 3, pp. 151-162.
Meller, R. D. and Gau, K. Y. (1996), ”The facility layout problem: recent and emerging trends and
perspectives”, Journal of Manufacturing Systems, Vol. 15 No. 5, pp. 351–366.
Mowrey, C. H. and Parikh, P. J. (2014), “Mixed-width aisle configurations for order picking in
distribution centers”, European Journal of Operational Research, Vol. 232 No. 1, pp. 87-97.
Pan, J. C. H. and Shih, P. H. (2008), “Evaluation of the throughput of a multiple-picker order picking
system with congestion consideration”, Computers & Industrial Engineering, Vol. 55 No. 2, pp.
379-389.
Pan, J. C. H. and Wu, M. H. (2012), “Throughput analysis for order picking system with multiple
pickers and aisle congestion considerations”, Computers & Operations Research, Vol. 39 No. 7,
pp. 1661-1672.
Pan, J. C. H., Shih, P. H. and Wu, M. H. (2012), “Storage assignment problem with travel distance and
blocking considerations for a picker-to-part order picking system”, Computers & Industrial
Engineering, Vol. 62 No. 2, pp. 527-535.
Parikh, P. J. and Meller, R. D. (2008), “Selecting between batch and zone order picking strategies in a
distribution center”, Transportation Research Part E: Logistics and Transportation Review, Vol.
44 No. 5, pp. 696-719.
Parikh, P. J. and Meller, R. D. (2009), “Estimating picker blocking in wide-aisle order picking
systems”, IIE Transactions, Vol. 41 No. 3, pp. 232-246.
Parikh, P. J. and Meller, R. D. (2010a), ”A note on worker blocking in narrow-aisle order picking
systems when pick time is non-deterministic”, IIE Transactions, Vol. 42 No. 6, pp. 392-404.
Parikh, P. J. and Meller, R. D. (2010b), “A travel-time model for a person-onboard order picking
system”, European Journal of Operational Research, Vol. 200 No. 2, pp. 385-394.
Petersen, C. G. (1999), “The impact of routing and storage policies on warehouse efficiency”,
International Journal of Operations & Production Management, Vol. 19 No. 10, pp. 1053-1064.
Petersen, C. G. and Schmenner R. W. (1999), ”An evaluation of routing and volume‐based storage
policies in an order picking operation”, Decision Sciences, Vol. 30 No. 2, pp. 481-501.
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Petersen, C. G. and Aase, G. (2004), “A comparison of picking, storage, and routing policies in
manual order picking”, International Journal of Production Economics, Vol. 92 No. 1, pp. 11-19.
Petersen, C. G., Aase, G. R. and Heiser, D. R. (2004), “Improving order-picking performance through
the implementation of class-based storage”, International Journal of Physical Distribution &
Logistics Management, Vol. 34 No. 7, pp. 534-544.
Ratliff, H. D. and Rosenthal, A. S. (1983), “Order-picking in a rectangular warehouse: a solvable case
of the traveling salesman problem”, Operations Research, Vol. 31 No. 3, pp. 507-521.
Robinson, W. N. and Ding, Y. (2010), “A survey of customization support in agent-based business
process simulation tools”, ACM Transactions on Modeling and Computer Simulation (TOMACS),
Vol. 20 No. 3, pp. 14:1-14:29.
Roodbergen, K. J. and de Koster, R. (2001), “Routing methods for warehouses with multiple cross
aisles”, International Journal of Production Research, Vol. 39 No. 9, pp. 1865-1883.
Roodbergen, K. J., Sharp, G. P. and Vis, I. F. (2008), “Designing the layout structure of manual order
picking areas in warehouses”, IIE Transactions, Vol. 40 No. 11, pp. 1032-1045.
Ruben, R. A. and Jacobs, F. R. (1999), “Batch construction heuristics and storage assignment
strategies for walk/ride and pick systems”, Management Science, Vol. 45 No. 4, pp. 575-596.
Sargent, R. G. (2013), “Verification and validation of simulation models”, Journal of Simulation, Vol.
7 No. 1, pp. 12-24.
Skufca, J. D. (2005), “k workers in a circular warehouse: a random walk on a circle, without passing”,
SIAM Review, Vol. 47 No. 2, pp. 301-314.
Tompkins, J. A., White, J. A., Bozer, Y. A. and Tanchoco, J. A. (2010), Facilities Planning, Wiley,
New York.
Table 1: Overview of factors varied in the simulation study

Factor Specification
Routing policy return, s-shape, midpoint, largest gap, composite, combined, optimal
Storage assignment rule random, class-based (ABC vertical, horizontal, diagonal)
Number of order pickers 2, 5, 10, 15
Number of picks per order 10, 20
Blocking on, off
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Table 2: Overview of mean throughput times [sec] (shortest mean throughput times highlighted)

Random ABC diagonal ABC vertical ABC horizontal

Picker
Return
Return
Return
Return

S-shape
S-shape
S-shape
S-shape

Picks per order


Optimal
Optimal
Optimal
Optimal

Midpoint
Midpoint
Midpoint
Midpoint

Combined
Combined
Combined
Combined

Composite
Composite
Composite
Composite

Largest gap
Largest gap
Largest gap
Largest gap

2 558 515 472 465 474 468 430 399 465 432 431 391 391 380 453 397 395 384 383 377 346 386 500 432 432 383 383 377

5 633 569 505 496 517 515 458 488 539 499 498 472 474 456 603 512 477 461 491 478 408 447 552 473 473 443 442 427

10

10 726 640 536 526 569 569 495 610 648 582 587 573 586 580 820 668 584 570 633 620 484 517 625 525 525 510 509 508

15 799 691 557 546 607 611 530 708 727 654 663 653 678 713 1020 787 680 676 746 731 556 571 681 560 561 562 560 586

2 969 830 815 789 804 795 730 736 786 756 755 708 709 657 776 677 729 685 651 663 592 710 817 738 736 702 700 660

5 1180 952 909 869 916 915 789 995 971 966 972 918 937 788 1025 887 976 869 845 880 661 904 939 889 889 887 883 783

20
113
10 1443 1105 1014 965 1055 1065 880 1323 1189 1214 1257 1154 1237 1020 1375 1134 1261 1138 1060 1135 756 1103 1073 1076 1100 1105 961
6

133 114
15 1654 1226 1097 1039 1166 1189 961 1616 1331 1402 1497 1330 1485 1243 1737 1330 1545 1409 1233 1347 854 1239 1223 1238 1277 1286
0 1
Table 3: Mean throughput times [sec] (without blocking) for routing combinations with two agents (white: longer
mean throughput times; dark grey: shortest mean throughput times; bold: combinations with same routing policies;
underlined: shortest mean throughput time)

Order picker 1
Largest
Return S-shape Midpoint Composite Combined Optimal
gap
Return 865 823 816 807 810 802 787

S-shape 821 779 772 763 766 758 743

Midpoint 814 771 764 756 758 751 735


Order
Largest
picker 805 763 756 747 750 742 727
gap
2
Composite 808 766 759 750 753 745 730
Downloaded by Cornell University Library At 05:10 17 July 2017 (PT)

Combined 800 757 750 742 745 737 722

Optimal 785 742 735 727 729 722 706

Table 4: Mean throughput times [sec] (with blocking) for routing combinations with two agents (white: longer mean
throughput times; dark grey: shortest mean throughput times; bold: combinations with same routing policies;
underlined: shortest mean throughput time)

Order picker 1
Largest
Return S-shape Midpoint Composite Combined Optimal
gap
Return 968 911 846 839 879 867 795

S-shape 885 829 794 787 820 817 755

Midpoint 826 794 810 799 778 771 745


Order
Largest
picker 819 790 795 784 773 764 738
gap
2
Composite 857 818 783 777 803 798 741

Combined 848 818 778 772 802 794 734

Optimal 793 754 745 737 741 733 730

You might also like