PI Cross-Docking: Applying AutoStore Design Principles

(1)

1

PI Cross-Docking: Applying AutoStore Design Principles

Willem R. Slump

Supervisor: Dr. N.B. (Nick) Szirbik

Second supervisor: Dr. J.C. (Hans) Wortmann

Faculty of Economics and Business University of Groningen

(2)

2

Abstract

(3)

3

Introduction

Over the past decades, the transportation of goods has increased in volume substantially. From 2001 to 2016 global freight throughput grew by 313%, to a figure of over 701 million twenty-foot

equivalent units (TEU) (Gharehgozli, Zaerpour & De Koster 2019). In order cope with increasing demand, it is imperative to maintain the effectiveness of the supply chain with the increase in volume. In response, facility managers are adopting new techniques to improve efficiency and throughput. Although the concept of cross-docking is not new, novel technologies are enabling the next generation of cross-docking facilities to be more space efficient with higher throughput.

Cross-docking is a supply chain logistics procedure that involves moving goods directly from the inbound receiving bay to the outbound bay. This implies that in the ideal case, goods would move directly from the inbound truck to the outbound truck by a single party. However, in practice goods may be placed into a staging area for sorting. Cross-docking facilities are often used in supply chain networks as a node where shipments are consolidated or broken up, depending on the final destination of the individual units in the shipment. This can cut down on total distance driven and other operational costs associated with shipping.

The Physical Internet concept has been developed as an answer to the sustainability challenge coupled with the growing need for goods transport (Montreuil 2011). The author’s proposal for achieving this relies on the standardization and modularization of transportation units, called PI-containers, and the consolidation of supply chain networks and operations. The potential impact of the consolidation of supply chain operations is a 23.05% reduction in average total cost and 93.97% reduction in average inventory cost (Venkatadri, Krishna & Ülkü 2016). In addition to contributing to the economic sustainability of supply chain operations, the Physical Internet also can improve environmental sustainability. Especially when paired with environmentally friendly energy sources, the reduction in emissions from the Physical Internet can reach up to 90% (Ballot 2019). The Physical Internet is an exciting concept in logistics system design which increases returns on improved emissions efficiency as the scope of the system gets larger (Mangina, Narasimhan, Saffari & Vlachos 2020).

(6)

6 and vehicles that are capable of lifting loads rather than simply carrying them around. These classes of automated vehicle are called Automated Lifting Vehicles (ALVs), as opposed to Automated Guided Vehicles (AGVs) which are not able to directly lift the load they carry. While research focusing on AGVs is prevalent, ALVs are among the vehicle types that are less studied (Carlo, Vis & Roodbergen 2014). This paper will focus on the application of design principles loaned from the AutoStore system to create a cross-docking system in the Physical Internet (PI) context based on ALVs.

In order for the Physical Internet to function, all of its components need to be specified and designed. Cross-docking facilities can be used to route containers through the Physical Internet network, behaving similarly to routers used in the digital internet today. This project discusses how the cross-docking can be applied in the context of the Physical Internet and will yield a design for an automated cross docking procedure. Using input from extant literature, a baseline of the design context will be given. This paper will also discuss the lessons learned from designs and research into similar systems.

Research Questions

In order to design an automated cross-docking system, one must first understand the design of current cross-docking procedures implemented in practice. It is not expected that every

implementation of cross-docking is operationally identical, however the similarities could be used to draw insight into the critical processes and functions of a cross-docking system. Furthermore, this understanding can be used as a basis from which the new design will be considered. This leads to the following research questions:

1. What is the current model of PI-containers defined in the body of Physical Internet literature?

(7)

7

Theoretical Background

Cross Docking

Cross-docking is a supply chain network activity in which shipments are transferred between vehicles without intermediate long-term storage (Van Belle, Valckenaers & Cattrysse 2012). One of the main benefits that this offers customers and supply chain partners is a reduction in order lead time. From the perspective of the company responsible for shipping the goods, this practice can also reduce shipment costs by cutting down on the total miles driven to ship goods. In addition, cross-docking helps to ensure that long distance shipments are fuller than they otherwise would be.

This section will give background information on cross-docking facilities from perspectives both inside and outside the facility. From inside the facility, the cross-docking procedure can be described at a finer level of detail. From outside the facility, the cross-docking facility is important for the function that it provides the supply chain at large.

Cross Docking Types

The absence of long-term storage in cross-docking facilities does not mean that goods travel directly from the inbound to the outbound vehicles without intermediate handling. In terms of intermediate handling, cross-docking only requires that the goods are not put away in storage or picking shelves (Apte & Viswanathan 2000). In fact, the cross-docking process of bringing materials from the

inbound vehicles to their destination outbound vehicles can be classified according to the number of staging areas utilized at the facility (Van Belle et al. 2012). Pure cross-docking is characterized by the absence of intermediate staging areas – where goods flow directly from the inbound to the

outbound vehicle. Single-stage cross-docking is characterized as using a single staging area between the inbound and outbound vehicles. The purpose of this staging area is often to sort the goods before they are placed into the outbound vehicles. In some cases, two staging areas are used - called two-stage cross-docking.

Cross-docking facilities can be classified into three main categories according to the type of service provided: Consolidation, Break-bulk, and Mixed (Apte & Viswanathan 2000):

(8)

8 destinations of the goods in the FTL shipments are not all the same, they all need to travel in the same direction for a long distance.

Break-bulk – This type of cross-docking facility is breaks FTL shipments into many LTL shipments destined for separate final destinations. For this reason, this process can be viewed as the opposite of consolidation cross-docking.

(9)

9

Network Structures

A key advantage of cross docking facilities is the effect they can have on supply chain network structure when they are properly incorporated into the supply chain. The power of cross-docking comes in fully when considering aggregating supply chain activities across previously closed

networks. The issue identified by Montreuil, Ballot, and Fontane (2012) is that supply chain networks often overlap, creating the effect that the overall network to operate inefficiently. Therefore,

network structure is relevant to the performance of the supply chain activities holistically and to ensuring that the transportation of goods is as efficient as possible. Supply chain nodes where shipments are efficiently routed play a role in the streamlining of supply chain networks (Fergani, El Bouzekri El Idrissi, Marcotte & Hajjaji 2019).

As an example of the aforementioned principle, if we consider a set of suppliers and destinations, a traditional supply chain may include a considerable degree of overlap. In the figure below, the color of the arrows distinguish between supply chain flows from a particular supplier to the destinations. If it were possible to aggregate the flow of goods from both suppliers, the trucks used to transport the goods would likely be utilized to a greater extent and the total distance driven would decrease. This presents an opportunity for the introduction of cross-docking.

(10)

10

Cross-Docking Automation

There are several different approaches to cross-docking automation designs with varying degrees of adoption in practice and maturity in design. Literature based on shipping container terminals may provide insight into the cross-docking paradigm presented in this paper since the handling

operations will be on standard c container sizes. These designs include conveyor systems, automated guided vehicles (AGV), overhead grid-rail (GR), and automated storage and retrival systems

(AS/RS)(Liu, Jula & Ioannou 2002). Research into automated lifting vehicles (ALVs), which are capable of lifting the containers as well as transporting them, has shown systems based on ALVs to

outperform systems based on AGVs (Bae, Choe & Park 2011). Some approaches in practice use two-stage AGV systems with racks to buffer the inbound and outbound side of the cross-dock (Dematic Retrotech 2010). Other systems used in practice employ conveyor systems to automate the sorting procedure in cross-docks (Vanderlande 2020).

Physical Internet

The Physical Internet as a topic grew out of observations of the inefficiencies in current logistics networks and a proposal to drastically change the way we approach the design of logistics systems – on both a global and local scale. Fundamentally, the current state of supply chain logistics is such that often times supply networks will overlap geographically, without coordinating between the networks (Montreuil et al. 2010). The key observation made was that supply networks are often specific to one company operating or a small group of companies working together to realize the supply network operations. This results in many more miles driven than would theoretically be necessary if the supply chain operations were aggregated to a higher level.

(11)

11 is the perfect example to use as the basis for a global supply chain network through which anyone and everyone can transport their goods.

Core Tenants of the Physical Internet

The Physical Internet (PI) is the name given to the conceptual redesign of global logistics networks. The name is meant to suggest similarity of the PI logistics system and the mechanisms of the Digital Internet. The core tenants of the Physical Internet include modularity, standardization, and the use of automation where possible (Montreuil, Meller & Ballot 2012).

The modularity aspect of the PI is seen in the concept for the PI-containers. As originally conceived, the PI-container would be an object capable of storing objects inside, storing information about the shipment, and compose with other containers on the outside dimensions – much like how Lego pieces can be used to compose larger bricks. Due to the intention of the PI designers to have the ability to compose the PI-containers in shipments, the architecture of the PI and all of its facilities is dependent on the container sizing scheme. This means that one globally consistent convention for PI-container sizing must be chosen. Again, as a clear constraint, the containers must be able to be added together to create the same outside dimensions as other larger containers in the

specification.

Automation is another key consideration of the PI, which is heavily driven by the level of

standardization and modularity in the PI. These two factors allow automation to take over repetitive and frequent tasks in the PI. The more standardized the PI-containers are in terms of size, the simper the handling mechanisms can be across the PI network. That is, the fewer PI-container sizes that one system needs to be capable of handling, the simpler the handling mechanism can be. Luckily, driven by the modular aspect of the PI, the PI-containers can be aggregated and combined to produce the dimensions required by the handling system.

The implementation of the Physical Internet will simplify distribution networks as PI-hubs are used efficiently by multiple suppliers, resulting in fewer direct connections between the source of goods and their destination (Fergani, El Bouzekri El Idrissi, Marcotte, Hajjaji 2019). Ballot (2019) also shows how this effect will manifest on a larger scale by making use of the network of networks concept to consolidate distribution on a national level. France was chosen as the case shown in the article, but the principle could be applied to any nation or region.

Cross-docking in the Physical Internet

The analogy to the Digital Internet means that design of the Physical Internet should take

(12)

12 has been done to envision the various network layers in the Physical Internet (Montreuil, B., Ballot, E. & Fontane, F. 2012). This paper also included a comparison to the network layer protocols which are in place for the Digital Internet, namely the Open Systems Interconnection (OSI) and the Transmission Control Protocol / Internet Protocol (TCP/IP). The particular system layer that PI-hubs such as cross-docking facilities find themselves in is the network layer. This layer is found in the OSI and TCP/IP protocols as well.

Recently, study into the trade-offs between conventional and Physical Internet logistics systems has interesting implications for PI-hubs. Among the key points of the research, was that the Physical Internet excels in simplifying the driving logistics, however this requires more work to be done in the transit centers themselves (Fazili et al. 2017). Additionally, this paper concludes that the efficiency of the PI depends in great part on the efficiency of its transit centers. In this case, the transit centers are the same as PI-hubs. Since a PI-cross-dock facility is a type of PI-hub, this would imply that a proper design of PI-cross-docking facilities is one aspect of an efficient and effective design of the Physical Internet.

Physical Internet Containers

The PI-container is the fundamental transportation unit used in the Physical Internet. They were first described as coming in modular dimensions and being able to stack together based on outside dimensions (Montreuil, Meller & Ballot 2010). The initial proposal a set of 11 side length dimensions which ranged from 120 cm to 12 meters in size. The full set of initially proposed sizes includes: 0.12m, 0.24m, 0.36m, 0.48m, 0.6m, 1.2m, 2.4m, 3.6m, 4.8m, 6m, and 12m. Since the idea is to compose the containers along the outside dimensions, it was important that many of the proposed side lengths are multiples of smaller side lengths. A PI-container should have outside dimensions that are picked from a set of defined modular dimensions, note that the actual definition of allowed modular dimensions is subject to change. In fact, work has been done to define a set of modular dimensions that are suitable given a range of products and their respective sizes (Landschützer, Ehrentraut & Jodin 2015). Alternative physical designs for PI-containers have also been given (Harald et al. 2016). The containers elaborated in this design, called SmartBoxes, come in five sizes and are designed to work in a last mile parcel locker called a SmartTerminal.

Developments in PI-Container Design

(13)

13 2014). In the paper, the authors explain that these categories of containers form a hierarchy in which groups of smaller containers are encapsulated in the larger containers.

According to the definition, packaging containers (P-container) would be grouped together and encapsulated into handling containers (H-container). By doing this, the number of individual handling operations goes down. In order to ship freight in even larger units, the transport container (T-container) will encapsulate a grouping of H-containers. The concept is also expanded upon and can allow for a T-container to be comprised of H-containers as well as a container group consisting of a grouping of a large P-containers and a group of P-containers (Sallez, Montreil & Ballot 2015). Elaborating on packaging containers, the name of this container is exactly descriptive of the purpose of the container itself: to serve as packaging for the goods held inside. It is important to note that according to the article, the P-container is not a generic container the contents of which remain anonymous. Rather, the P-container is a container which serves as the packaging for a specific goods. As described in the article, P-containers have the following characteristics (Montreuil, Ballot & Tremblay 2014):

1. Lower privacy of the contained goods since goods owners want to expose the product to potential buyers

2. Lowest amount of robust physical protection of the contained goods, they are the lightest and thinnest PI-containers

3. Fastest handling/sorting speed, accuracy, and efficiency since they encapsulate individual product units

(14)

14

Container Handling Systems

Frame Bridge

Figure 1: An illustration of the Frame Bridge Automated Container Terminal System (Zhen et al. 2012)

The Frame Bridge Automated Container Terminal (FB-ACT) pictured above provides an automated system for performing handling operations on containers coming in on maritime vessels. Containers are offloaded from the ships using quay cranes (QCs), and are placed onto frame trolleys (FTs) that move parallel to the ship in rows 𝑟1 to 𝑟5, in the figure above. The FTs move along either an upper

story or lower story along guide rails. Transfer platforms (TPs) provide the means to rotate the container 90 degrees to transfer the container to the ground trolley. The ground trolley is then responsible for bringing the container to a yard crane in blocks 𝑏1 to 𝑏10, in the figure above.

The FB-ACT system comprises of many working pieces that need to work together in order to realize the goal of efficient container handling operations. To provide a framework for understanding the performance of FB-ACT systems, analytical methods were provided by (Zhen et al. 2012). Further analysis into the performance of the system, focusing on the model of the transfer platform concluded that a careful balance between the arrival times of the FT and GT is necessary to ensure optimal throughput of the transfer platforms (Hu et al. 2013). The performance and space efficiency of the FB-ACT can also be seen as a function of the vehicles and systems used to implement

(15)

15 decreases (Jiang et al. 2018). However, the FB-ACT system provides an important datapoint in understanding the design of systems that handle shipping container sized loads.

AutoStore Warehouse System

Figure 2: The AutoStore System (Motrac Linde 2020)

Vertical expansion seems to be a promising design choice made by facility designers due in part to making better use of land (Gharehgozli, Zaerpour & De Koster 2019). While there are many systems that can accommodate vertical expansion (Roodbergen & Vis 2009), it seems most logical to use a design which incorporates an ALV. An example design of an ALV used to implement a system that is implemented in large part due to its benefit on the usage of space is the AutoStore system. The system, pictured above makes use of robots which traverse a grid above columns of bins contained within. When a human operator gives the signal to retrieve an item from storage, the AutoStore system sets the robots into motion to retrieve the desired item. When the proper container is picked, the robot proceeds to the drop-off point where the human operator can freely pick or place items from the bin provided to them by the AutoStore system. Interest in ALV-based systems over AGVs is based in the evidence that ALVs can outperform systems based on AGVs (Azadeh, Roy & De Koster 2019a).

Goods Retrieval and Inventory Distribution System

(16)

16 While there are many systems used for handling containers in terminals (Carlo, Roodbergen & Vis 2014), an advance of container handling technology is the Goods Retrieval and Inventory

Distribution (GRID) is gaining research interest. The GRID system, developed by BEC Industries LLC, is an autonomous container storage and retrieval system which is capable of handling 40-foot ISO containers up to 55,000 lbs. Designs which incorporate GRID into terminal operations often involve employing them to mediate handling and storage of containers coming off of ships in port

transhipment terminals (Jiang, Chew & Lee 2015). The authors highlight that the benefits of this method include making maximum use of land for storage, higher unloading throughput rate, eliminating unnecessary transfers between vehicles in the storage facility itself, and that it is powered by electricity.

While there are other systems for storing goods based on grids (Gue, Furmans, Seibold & Uludag 2014), systems resembling the AutoStore and GRID will be chosen as the design inspiration for the reasons given above.

Methodology

Scope

The focus of this paper will be to provide a design for a PI-hub, a PI component (Pan, Ballot, Huang & Montreuil 2017). As there are many open topics when it comes to automating cross docking systems in an PI context, it is important to clearly define the scope of the project. The research and proposed design will focus on the cross-docking operation itself. The research should cover consolidation and break-bulk cross-docking, focusing on the functional design of the procedure. Aspects of the physical design of the cross-docking facility such as the number of container levels able to be cross-docked at a single facility will not be treated as part of the current scope. The design problem will consider that the inbound and outbound trucks have already arrived as scheduled and are ready for cross-docking.

Research Design

(17)

17 a starting point for the design. In this case, the need to provide efficient routing of PI-containers in the Physical Internet to ensure effective global transportation of goods is that context. The design cycle draws upon knowledge gained in the Rigor cycle, the area of rigorous scientific inquiry and research. The rigor cycle provides designers with the required information to design systems based on knowledge of the underlying working principles. After the design is proposed, it must be

validated with evidence to show that the design would function as required if it were implemented in practice.

Design Validation

The validation plan for this project consists of two parts. First, a series of design guidelines will be established with the help of literature. To ensure consistency within the design process, part of the validation will be dedicated to discussing how these guidelines were addressed. Second, to ensure that the design could function as specified, evidence must be provided that the required physical components and designs exist. This will be substantiated by research showing that the key elements of the design can be realized in practice.

Design Guidelines

The following design guidelines were collected with respect to the design of the cross-docking facility (Walha et al. 2014, Montreuil et al. 2018):

Source Index Guideline Walha et al.

2014

1 Only PI-containers can be used in a PI-hub

2 The hub will not be restricted to a specific set of users

3 The hub must be designed with economic, environmental, and societal constraints in mid

4 The number of sources and destinations of containers will be less than in traditional networks

Montreuil et al. 2018

5 Hubs should receive and ship modular containers encapsulating shipments based on common next destination within the PI

(18)

18

As-Is Design

Montreuil et al. (2012) provide a design of a cross-docking PI-hub and provided a feasible functional concept. The authors proposed a set of functional elements that when put together could provide cross-docking services in a PI setting. This source was reverse engineered to provide a baseline for the design of a PI-cross-docking hub.

Provide Cross-Docking in a PI Context

In the context of the PI cross-docking services there are three main functions: the arrival routine, the driver services, and the final security check. The arrival routine covers all aspects of accepting a new truckload into the cross-docking facility. The incoming truckload is recognized and registered with the PI-hub and crucial security checks are performed. The goal of these security checks is to ensure that the load currently being carried matches what left the previous PI-facility. Once the security check is performed, the driver is assigned to a PI-InGate. If the driver indicated that they want to make use of the driver services available at the facility, this is registered at this time.

The final security check is performed to ensure that the driver, truck, truckload, and next destination matches the work order accepted by the driver when entering the facility. If all of the data matches, then the truck is cleared to leave the facility and the driver can carry out their work order.

(19)

19 The above figure shows the specification of the various functions in the context of the cross-docking procedure in greater detail. The “Perform Arrival Routine” function takes a truck, driver, and a trailer to perform the necessary routine when the truck approaches the cross-docking facility. Triggered on the arrival of a new truck, this function is performed and many physical outputs and signals are generated. The physical outputs include passing the truck, trailer and the truck route to the next function, and passing the drivers who intend to use the driver services to the “Provide Driver Services” function. The signals produced are those to signify that the truck has arrived in the correct PI-dock bay so that cross-docking can start, and to communicate that the driver of this truck has requested driver services.

The “Provide Cross-Docking Services” function performs the core cross-docking functionality of the system. This function is conceptualized as such since the original model formulated the cross-docking procedure as part of the core of the facility, around which the rest of the functionality takes place. Once the truck is in the proper PI-dock, cross-docking is performed to produce a loaded truck and trailer. Additionally, the system is made aware that the cross-docking procedure is complete for that given truck and a signal is passed to the final security check function.

(20)

20

Provide Cross-Docking Services

When examined more closely, the “Provide Cross-Docking Services” function can be broken down into a number of sub-functionalities that contribute to the overall cross-docking process. These four functionalities can be found in the above figure. The “Prepare Load” function performs functions that make sure that the new containers to be loaded onto the trailer are grouped together properly so as to allow for a time efficient loading process. The “Reconfigure Trailer” function is performed to ensure that the trailer has enough space so that new groupings can be inserted in between the containers that are left on the trailer after unloading.

First, the trailer must be unloaded to provide the cross-docking system with the container groups that need to be routed onto another trailer. This function is triggered as soon as the truck is present at the scheduled PI-dock. The function takes the full trailer and the next destination of the truck as input, producing a partially empty trailer. Additionally, a signal indicating that the trailer is unloaded is created and is distributed to the system.

(21)

21 information and decides how to reconfigure the grouping. This produces a reconfigured trailer, which is used in the loading process.

In order to prepare the trailer load, the system needs to take into account all containers registered in the system as well as the next destination of the truck to decide which containers must be assigned to which trucks. This function produces the container groups that will be loaded onto the trailer in a given PI-dock bay as well as a signal indicating that the trailer is ready to be loaded. Once this signal is given, the system begins processing the assigned container groups and loading them onto the trailer. When the loading process is finished, the fully loaded trailer and a signal indicating that cross-docking is finished are produces as output.

Prepare Load

The “Prepare Load” function can be understood as encapsulating the container pooling, sorting, and grouping functionalities. These three functionalities work together to ensure that the proper

(22)

22 When the trailer is unloaded, the containers are first pooled together in central buffer zones to await sorting and the grouping of containers. The pooling process takes the collection of containers in the system as well as the next destination of the truck into account.

The containers are then sorted according to common next destination required to meet their individual final destinations. This process begins when the containers are pooled together. The outcome of this process is a list of containers sorted according to their next destinations. Then the containers are grouped according to the input sorted list of containers available. This process begins when the “Containers Sorted” signal is received. The output of this function is the list of container groups that are assigned to a particular trailer and the signal that the trailer is ready for loading.

Reconfigure Trailer

(23)

23 First, the new container groups must be determined based on the containers left over after the trailer is unloaded. Thus, the reconfiguration process is triggered when the “Trailer Unloaded Signal” is received. At which point, the truck route and the partially loaded trailer are used to figure out the new grouping of containers, should regrouping be necessary. The outputs of this process are a signal that indicates that a new grouping is determined along with the specification of this grouping. After which, the spacing requirements of the groups which will later be loaded onto the trailer must be calculated.

Finally, the container groups are moved according to the provided spacing specification to make room for the incoming container groups. Once this is finished, the signal is sent that the trailer is reconfigured.

Unload Trailer

(24)

24 is characterized as the process that occurs when the groups are marked for unloading, unlatched, and finally physically removed from the trailer.

When the truck is located in the proper PI-dock bay, the unloading processes is triggered. First, the groups which need further processing at the PI-cross-docking facility are determined. In order to do this, the truck route and the routing information of the groups located on the trailer are used to work out which groups need unloading. Additionally, a “Grouping Determined” signal is produced by this process which signals the subsequent process to be performed.

At this point, the groups indicated in the previous process need to be unlatched from their

surroundings so that they can be manipulated by other PI-movers to be brought into the main part of the facility. The system must determine, based on the borders of the respective container group, which containers to send which unlatching signals to. This process is complicated since you only want the containers to unlatch from their outside connections, and to maintain their connections to other containers of the same group. However, once this process is successfully completed, the groups are now able to be moved by the PI-movers provided by the facility.

(25)

25

To-Be Design

The design which will be outlined in the following section builds upon previous research in the PI field by showing the design implication of using nested containers. In order to continue with the design of the cross-docking system, a framework for the containers which will be handled must be given. Once a detailed high-level specification is given, the design incorporating the new system of containers can be given.

Another Look at PI-containers

As discussed in the Theoretical Background, the specification of PI-containers has been expanded upon since the first papers written about the Physical Internet (Montreuil, Ballot & Tremblay 2014). The paper elaborates on applying the concept of encapsulation to PI-containers and specifies three levels. Transportation containers are meant to act functionally as current shipping containers, handling containers are meant to function similarly to boxes, and packaging containers should function as packaging for products. The general dimensions of transport containers are given to be 1.2m or 2.4m in height and width, and from 1.2m to 12m in multiples of 1.2m. These dimensions would make them incompatible with ISO 40-ft containers.

The author would like to propose another system of encapsulating containers. Rather than assigning a purpose to each of the encapsulation layers, the levels of PI-containers should be indexed with numbers. This ensures that purpose of each PI-container is separated from its dimensions. One of the key advantages of PI-containers is the fact that it encapsulates the contents inside (Osmolski, Voronina & Kolinski 2019). Referring to PI-containers with a number index will also ensure that the definition of the container implies nothing about where in the PI it could be used and handled in the same way as any other PI-container.

The foundation of the system will be the 40-ft and 20-ft units as they are internationally used and recognized standards for container sizes. These containers are the foundational sizes of container and are called Level-0 containers. The other guiding principle of the new system of containers is that a container can either encapsulate a single product, or item which needs handling or it must be comprised of smaller containers that can compose to fit securely within the internal dimensions of the higher-level container. Thus, Level-1 containers are containers that are of such size and dimensions as to compose to fit within a Level-0 container. Generally speaking, a container at any arbitrary level N (Level-N or 𝐿𝑁) is designed to compose with containers at the same level, contain

(26)

26 (Montreuil, Ballot & Tremblay 2014;

Sallez et al. 2016)

New container classification

Example size, longest side length

Transportation container 𝐿0 container 40-ft, 20-ft

Handling Container 𝐿1 container 10-ft, 5-ft

𝐿2 container 2.5-ft

Packaging Container 𝐿3 container 1.25-ft

Process 𝐿

_𝑛

Container

The function 0 in this design is considered to be a drop-in replacement for the function 0 in the design previously described. Therefore, rather than repeating the diagrams, the specification for the new design begins at this level.

(27)

27 The initial sorting procedure is started when a new 𝐿𝑛 container arrives. This sorting process figures

out where to send the containers which need to be unloaded at the current facility. For the groups whose final destination is the current facility, the group must be disbanded and reformed. Thus, they are sent to the “Process 𝐿𝑛+1 Groups” function. The groups which needed to be unloaded, but not

taken apart, are sent to the “Sort and Group” function. Once a complete set of groups are available for loading at the 𝐿𝑛 level, they are loaded and a signal is sent that the 𝐿𝑛 group is ready to be

loaded. The loading process puts the container top back on and the container is ready to be shipped.

Sort and Group

The “Sort and Group” function is necessary to reform 𝐿𝑛+1 container groups so that they can be

(28)

28 At the beginning of this process, 𝐿𝑛+1 level groups are received and passed on to the function that

sorts the incoming containers on travel distance in the network. The sorting process also takes into account any 𝐿𝑛+1 containers left in the system that have yet to be included in a grouping. The

sorting starts when the sorting start signal is received and results in a sorted list of 𝐿𝑛+1 container

groups. This function also sends the “Perform Grouping Signal” signal when finished.

The containers are then assigned to groups according to the furthest PI-hub node that they have in common with the other groups. In this process, care should be taken to create complete groups while keeping the number of times that a container is opened as low as possible. The outcome of this function is a list of the groupings based on the lower level groups available. A signal is also sent to trigger the next function to be performed.

In order to create an 𝐿𝑛 group, changes to both the software system and the physical arrangement

of the group have to be made. In the software system, the final destination of the 𝐿𝑛 group is

determined by the last common node in the optimal path through PI network that all the sub-groups in the group share. The unload destination, the next destination where this group is to be unloaded and processed, is determined by matching the optimal path of the group with the routes of the available trucks. The group will be assigned to the higher-level group which can carry it the farthest along the optimal path towards the final destination. This concludes the “Sort and Group” function, the result is a full 𝐿𝑛 group and a signal to the system to physically load this group.

(29)

29 This function is crucial in determining what must be done to each incoming 𝐿𝑛+1 group upon the

opening of its 𝐿𝑛 container group. This function will decide which container groups need to be

opened further at the cross-docking location and which can be directly transferred onto another 𝐿𝑛

group. This function also takes care of signalling the system to move the container groups in question and notifying the “Process 𝐿𝑛+1” and “Sort and Group” functions of the oncoming

container groups.

When the 𝐿𝑛 trailer arrives, the container top is removed, exposing the contents of the container

and signalling that the container’s contents will be processed. Then, the unload destination is read from each of the constituent 𝐿𝑛+1 groups. All groups for which the unloading destination is the

current facility are unloaded from the 𝐿𝑛 group and sent to make another determination crucial to

how they will be processed. If the final destination of an 𝐿𝑛+1 group is the current facility, then it is

sent for further processing through the “Move Containers to Processing” function. In this case, further processing means the unloading and reassembly of a complete 𝐿𝑛+1 container. If the final

destination of the container is a facility other than the current one, the 𝐿𝑛+1 group is sent to be

transferred directly to the “Sort and Group” function, where the optimal 𝐿𝑛 group to carry it will be

(30)

30

Validation

Design Guidelines

1 – Only PI-containers can be used in a PI-hub

This guideline is met by the system as the only kind of containers considered for handling are the containers described in the Design section. Given the high level of container standardization

required for this design, it is assumed that PI-hubs using the design outlined above would be realized in a mature stage of PI development and global adoption.

2 – The hub will not be restricted to a specific set of users

The design outline in this paper makes no mention of any restrictions as to who may use the facility. Other than information about a container’s final destination, the cross-docking system operates without information of any of the parties using the system.

3 – The hub must be designed with economic, environmental, and societal constraints in mind

The GRID system lowers economic costs by making efficient use of facility space and lowering operating costs and environmental costs by running on electricity (Jiang, Chew & Lee 2015). Any implementation of this design will use electricity as a power source since it is validated to work as a power source to drive the system when lifting the heaviest loads (20 and 40-ft containers). The societal implications of this design are few, because the PI-hub design is fully automated and the hub does not need to interact with any people in the performance of its cross-docking task. The only impact is the positive societal impact that playing a role in the implementation of PI as goods transport is made more sustainable and reliable.

4 – The number of sources and destinations will be less than in traditional networks

(31)

31

5- Hubs should receive and ship modular containers encapsulating shipments based on common next destination within the PI

The containers considered in this design encapsulate smaller containers by design and the function 21 “Generate Container Groupings” performs the task of ensuring that lower level containers get grouped together properly.

6 – Hubs should exploit pre-consolidation and avoid sorting when possible

When a container is unloaded from the higher-level container, a check occurs to determine whether the unloaded container can be immediately cross-docked and grouped with other containers of the same size. This eliminates unnecessary sorting and handling operations.

7 – Hubs should be increasingly multi-party and multi-modal

As discuss previously, the system is inherently multi-party since no restrictions are made as to who may use the system. The design also supports intermodal cross-docking of containers since the inbound and outbound vehicles are general PI-vehicles. Multi-modality with more than two transportation modes available may require an extension to the presented functional design if and only if the various modes transport PI-containers of differing sizes.

Open Top Containers

Open top containers are containers which have only 5 attached sides with the top open. They can be made in 20-ft and 40-ft variants (CBOX Containers 2019). The containers are available in both hard top (shown on the left) and soft top (shown on the right). This design is chosen so that the

(32)

32 standard containers. Although this design is not yet the subject of academic interest, it provides the physical component needed in our design to facilitate top loading and unloading via the overhead grid system.

GRID System

Beyond providing a feasible mechanism through which PI-containers can be handled and sorted, the GRID system may also be employed in high-capacity scenarios where performance is critical (Zhou, Chew, Lee & Liu 2016). The authors highlight that ALVs are favorable to AGVs since they have improved performance when compared. This is an advantage for GRID-like systems which integrate the lifting procedure as a fundamental part of the design. The hybrid GRID design uses several GRID systems together with one another to mediate container storage and handling in the shipyard. Overhead container handling systems can be used effectively at the level of 40-ft containers through the use of the GRID system and at the other end of the spectrum through the AutoStore system. Therefore, it must be assumed that overhead container handling is feasible for containers in between these two extremes.

Control Algorithms and Procedures

The design incorporates a hybrid control system. The material handling agents have their own control systems geared towards ensuring the agent is capable of performing the task of moving PI-containers around the facility. While on a higher level there is a scheduling system keeping track of the tasks that need to be performed and assigning jobs to each agent. The central system assigns tasks to agents with limited knowledge of what it takes to perform these tasks and the agents themselves dutifully perform their tasks as instructed, with little insight into the overall plan. This design pattern is known in software engineering as Separation of Concerns (Hürsch & Lopes 1995). Additional benefits of this approach of decentralized control are decreased dependence on one system for the reliability of the facility (Gue, Furmans, Seibold & Uludag 2014) and decentralized algorithms can exhibit improved performance above centralized algorithms (Borrajo & Fernandez 2019).

(33)

33 based on heuristic, solving the schedule to optimality for small problems and near-optimality larger problems (Nossack, Briskorn & Pesch 2018). However, research into blocking scenarios in the GRID system suggest using the Single Direction Traffic Rule (Zhou, Li, Lee & Chew 2018). The Single

Direction Traffic Rule allowed the performance of the GRID system to remain high, even when under high load. This principle can be used in the PI-hub to mitigate blocking scenarios.

Discussion

Russian Dolls and AutoStore

The design presented in this paper made use of two key concepts to define the containers and procedures used to perform cross-docking. The foundational concept of this design was the goal to mimic the AutoStore container handling system. This meant that cross-docking should be performed by agents that traverse the facility on an overhead grid and that the contents of a PI-container should be standard in dimension. Encapsulating containers was found to be an established part of the PI literature (Montreuil, Ballot & Tremblay 2014; Sallez et al. 2016). When the concept of encapsulating containers is combined with the AutoStore-inspired container concept the result is a Russian Doll-like system of nested containers. The main purpose of nesting containers is to provide a physical grouping to consolidate handling operations that must be performed on the group. Groups were therefore formed on the basis of common final destination.

(34)

34 this stage what precise sizes of container will be assigned to each container level and how many levels should be implemented at a PI-hub cross-docking facility.

Standard Handling Containers

Another way in which the design presented above draws inspiration from the AutoStore system is the standardization of container sizes allowed in the system. Both the AutoStore system and the GRID system discussed previously use relatively simple mechanisms for picking and handling the containers, taking advantage of the standardization of container sizes. In general, the greater the diversity in shape and size of the objects that need to be handled, the more complex the handling system must become. In fact, systems for recognizing and picking objects in a traditional warehouse setting has become the subject of academic and industry interest.

Amazon hosted a design challenge for picking systems which could be applied in their warehouses, which included many teams from universities across the world (Correll et al. 2018). In other research efforts, hybrid suction-gripper mechanisms have been developed such that the two methods

account for the weaknesses in the other (Hasegawa, Wada, Niitani, Okada & Inaba 2017). Another interesting approach to the design of general picking mechanisms is that of the soft gripper, which is capable of wrapping around an object (Zhang et al. 2016). While these systems were developed to handle general objects in a warehouse picking scenario, the discrepancy between the complexity of these picking systems and that of the AutoStore demonstrates the principle that picking mechanisms can be simpler if the units being handled are standardized.

(35)

35 Another interesting aspect of the PI system is how to structure last mile goods transport. That is, the method through which goods are delivered to end consumers. The Last Mile Problem is the term for the problem of getting goods transported to their final destinations. An interesting approach to this problem is that of the parcel locker network (Deutsch & Golany 2018). The parcel locker network is a network of terminals in which goods are securely stored until the person for whom a delivery is intended interacts with the terminal to retrieve their package. Applications of parcel lockers this method in a PI context have been explored (Harald et al. 2016). The authors describe the

SmartTerminal, their design for parcel lockers which make use of SmartBoxes, the series of boxes designed in the article. Since the containers presented in this paper encapsulate smaller containers, the PI container itself could be used as a parcel locker. Along with the AutoStore-like handling system, a full-fledged high-capacity parcel locker could be deployed even in densely populated areas.

Transportation Modality

One of the design guidelines taken into account is the emphasis on incorporating multiple modes of transport in the design of PI-hub (Montreuil et al. 2018). In fact, the design does not specify that the inbound and outbound vehicles must both be trucks. In this way, the design lends itself to the interpretation that it could specify how an intermodal cross-docking facility could work. Intermodal PI terminals have been shown to have favorable performance in terms of total operation time in moving unit loads, lowering storage cost, and storage capacity requirements (Wickramage & Ferrel 2016).

As an example of type of facility, the design could be adopted as a rail to road cross-docking facility. Simulation based analysis and mathematical models have been proposed in this area of research (Sun, Zhang, Dong & Lang 2018; Chargui, Bekrar, Reghioui & Trentesaux 2019b).

(36)

36 More research could be done into investigating how the PI could be used as a mechanism for

integrating supply chain flows across carriers to make efficient usage of alternative transportation modalities (Pan, Trentesaux, Ballot & Huang 2019).

Network Routing and Scheduling

When taking a broader perspective, the cross-docking facility can be seen as a single node in a supply chain network. At this level, the cross-docking facility provides the supply chain users the possibility of transferring goods from one truck to another midway through the journey of the individual unit loads. Analysis at this level is concerned with generating optimal vehicle routing schedules based on a set of available trucks and the set of goods that need to be transported. This is the level at which the cross-docking facility interacts with the larger context of the supply chain logistics system of which it is a part.

The vehicle routing problem focuses on the optimal vehicle routing that minimizes total transportation cost. This problem is primarily focused on the supply chain activities around the cross-docking facility, as seen by the metric used to determine the value of the solution. Heuristic approaches to the solution of such models have been successful (Lee & Yung 2006; Wen et al. 2009). The variables considered in these models are the number of vehicles in the system, the route of each vehicle, and the arrival times of the vehicles to the cross-docking facility.

This type of problem places the focus on the cross-docking facility itself by capturing the utility of the inbound and outbound truck scheduling assignment. Mathematical studies of this problem

formulate generally their utility functions to reduce material handling times and to improve material throughput in the facility (Yu & Egbelu 2008, Vahdani & Zandieh 2009, Boysen 2010). Studies using metaheuristics to approach the vechicle scheduling problem have been successful for real-world complexity problems (Vahdani & Zandieh 2009). For certain markets, such as food transportation, the goal is often to move materials through the supply chain as fast as possible. In these cases, it is helpful to match the arrival times of the inbound and outbound trucks such that no intermediate handling is required. A dynamic programming model tackling this topic has been successful in providing the solution to this extension of the vehicle scheduling problem (Boysen 2010).

Conclusion

(37)

37 has certainly been successful. The initial literature search defined cross-docking and described how it can be used in the consolidation of supply chain operations and the simplification of supply chain networks.

The background on the Physical Internet provided the backdrop needed to properly discuss the context of the design, especially the containers used in the design. A four-level system of containers was proposed, named Level-0 through Level-3. This provided the ability to generalize the cross-docking process and show how the process for a large container will be the same for a small, nested container. This also provided the opportunity for the design to be dynamic in its definition of cross-docking. Using the design specified in this paper as a template, one can design a new facility which can process different levels of encapsulation (container levels) by adding or removing levels to the diagrams.

To show that the working principles of the design were valid, an exploration into the physical and software components of the design were provided. Open top containers can be used to provide the ability to load and unload the container from an overhead system. The currently designed and implemented GRID system provides a proof of concept design that makes it possible to envision a design for cross-docking which includes 40-ft containers as allowable sizes. The myriad control algorithms and solutions provide the confidence to say that the control system would be possible to implement.

(38)

38

References

Alice (2020). Sustainable, Safe and Secure Supply Chain. Research & Innovation Roadmap. Retrieved from http://www.etp-logistics.eu/?page_id=24

Ambra, T., Caris, A., & Macharis, C. (2019). Towards freight transport system unification: reviewing and combining the advancements in the physical internet and synchromodal transport research.

International Journal of Production Research, 57(6), 1606-1623.

Andriansyah, R. (2011). Order-picking workstations for automated warehouses. Doctoral diss.,

Technische Universiteit Eindhoven.

Apte, U.; Viswanathan, S. (2000). Effective Cross Docking for Improving Distribution Efficiencies. International Journal of Logistics: Research and Applications, vol 3, no 3 pp. 291-302.

Azadeh, K., Roy, D., & De Koster, R. (2019a). Design, modeling, and analysis of vertical robotic storage and retrieval systems. Transportation Science, 53(5), 1213-1234.

Azadeh, K., De Koster, R., & Roy, D. (2019b). Robotized and automated warehouse systems: Review and recent developments. Transportation Science, 53(4), 917-945.

Ballot, E. (2019). The Physical Internet.Operations, Logistics and Supply Chain Management.

Ballot, E., Montreuil, B., & Thivierge, C. (2012). Functional design of physical internet facilities: a road-rail hub.

Bae, H. Y., Choe, R., Park, T., & Ryu, K. R. (2011). Comparison of operations of AGVs and ALVs in an automated container terminal. Journal of Intelligent Manufacturing, 22(3), 413-426.

BEC Industries (2020). Material Handling: GRID. [JPG Image] Retrieved from http://www.becind.com/GRID/grid.html

(39)

39 Borrajo, D., & Fernández, S. (2019). Efficient approaches for multi-agent planning. Knowledge and

Information Systems, 58(2), 425-479.

Boysen, N. (2010). Truck scheduling at zero-inventory cross docking terminals. Computers & Operations Research, 37, pp. 32-41.

Carlo, H. J., Vis, I. F., & Roodbergen, K. J. (2014). Transport operations in container terminals: Literature overview, trends, research directions and classification scheme. European journal of

operational research, 236(1), 1-13.

CBOX Containers (2019, 4 October) Open top containers, voor een extra grote lading. Retrieved from https://cboxcontainers.nl/nieuws/7/open-top-containers-voor-een-extra-grote-lading

Chargui, T., Bekrar, A., Reghioui, M., & Trentesaux, D. (2019a). Proposal of a multi-agent model for the sustainable truck scheduling and containers grouping problem in a Road-Rail physical internet hub. International Journal of Production Research, 1-25.

Chargui, T., Bekrar, A., Reghioui, M., & Trentesaux, D. (2019b). Multi-Objective Sustainable Truck Scheduling in a Rail–Road Physical Internet Cross-Docking Hub Considering Energy Consumption.

Sustainability, 11(11), 3127.

Colin, J. Y., Mathieu, H., & Nakechbandi, M. (2016, May). A proposal for an open logistics

interconnection reference model for a Physical Internet. In 2016 3rd International Conference on

Logistics Operations Management (GOL) (pp. 1-6). IEEE.

Correll, N., Bekris, K. E., Berenson, D., Brock, O., Causo, A., Hauser, K., ... & Wurman, P. R. (2016). Analysis and observations from the first amazon picking challenge. IEEE Transactions on Automation

Science and Engineering, 15(1), 172-188.

Dematic Retrotech (2010, Feb 16). Improve Your Cross Docking Throughput|Retrotech [Video]. YouTube. https://www.youtube.com/watch?v=ubGEt9V09bo

(40)

40 Ekren, B. Y., Heragu, S. S., Krishnamurthy, A., & Malmborg, C. J. (2010). Simulation based

experimental design to identify factors affecting performance of AVS/RS. Computers & Industrial

Engineering, 58(1), 175-185.

Fazili, M.; Venkatadri, U.; Cyrus, P.; Tajbakhsh, M. (2017). Physical Internet, conventional and hybrid logistic systems: a routing optimisation-based comparison using the Eastern Canada road network case study. International Journal of Production Research, vol 55, no 9, pp. 2703-2730.

Fergani, C., El Idrissi, A. E. B., Marcotte, S., & Hajjaji, A. (2019). Framework and characterization of Physical Internet. International Journal of Computer and Information Engineering, 13(11), 599-608.

Gharehgozli, A., Zaerpour, N., & de Koster, R. (2019). Container terminal layout design: transition and future. Maritime Economics & Logistics, 1-30.

Gue, K. R., Furmans, K., Seibold, Z., & Uludağ, O. (2013). GridStore: a puzzle-based storage system with decentralized control. IEEE Transactions on Automation Science and Engineering, 11(2), 429-438.

Gupta, A., Roy, D., de Koster, R., & Parhi, S. (2017). Optimal stack layout in a sea container terminal with automated lifting vehicles. International Journal of Production Research, 55(13), 3747-3765.

Harald, K., Graf, H., Hoertenhuber, S., Widmann, R., & Stadlmann, B. (2016). SmartBox–An Austrian PI solution leading to small loads mobility 4.0. 3rd international physical internconference.

Hasegawa, S., Wada, K., Niitani, Y., Okada, K., & Inaba, M. (2017, September). A three-fingered hand with a suction gripping system for picking various objects in cluttered narrow space. In 2017 IEEE/RSJ

International Conference on Intelligent Robots and Systems (IROS) (pp. 1164-1171). IEEE.

Hevner, A. R. (2007). A three cycle view of design science research. Scandinavian journal of

information systems, 19(2), 4.

Homayouni, S. M., & Tang, S. H. (2016). Optimization of integrated scheduling of handling and storage operations at automated container terminals. WMU Journal of Maritime Affairs, 15(1), 17-39.

Hu, H., Lee, B. K., Huang, Y., Lee, L. H., & Chew, E. P. (2013). Performance analysis on transfer platforms in frame bridge based automated container terminals. Mathematical Problems in

(41)

41 Hürsch W.L., Lopes C.V. (1995) Separation of Concerns. Technical Report NU-CCS-95-03, College of Computer Science, Northeastern University, Boston,MA, February 1995.

Iris, Ç., & Lam, J. S. L. (2019). A review of energy efficiency in ports: Operational strategies, technologies and energy management systems. Renewable and Sustainable Energy Reviews, 112, 170-182.

Jiang, X. J., Xu, Y., Zhou, C., Chew, E. P., & Lee, L. H. (2018). Frame trolley dispatching algorithm for the frame bridge based automated container terminal. Transportation Science, 52(3), 722-737.

Ladier, A. L., & Alpan, G. (2016). Cross-docking operations: Current research versus industry practice.

Omega, 62, 145-162.

Landschutzer, C.; Ehrentraut, F.; Jodin, D. (2015). Containers for the Physical Internet: requirements and engineering design related to FMCG logistics. Logistics Research, 8, 8.

Lai, M., & Cai, X. (2019, September). Less-than-Truckload Shipper Collaboration in the Physical Internet. In Proceedings of SAI Intelligent Systems Conference (pp. 138-151). Springer, Cham.

Lee, Y.; Jung, J.; Lee, K. (2006). Vehicle routing scheduling for cross-docking in the supply chain. Computers & Industrial Engineering, 51, no 2, pp. 247-256.

Li, W., Xiaoning, Z., & Zhengyu, X. (2020). Efficient container stacking approach to improve handling: efficiency in Chinese rail–truck transshipment terminals. SIMULATION, 96(1), 3-15.

Liu, C. I., Jula, H., & Ioannou, P. A. (2002). Design, simulation, and evaluation of automated container terminals. IEEE Transactions on intelligent transportation systems, 3(1), 12-26.

Mangina, E., Narasimhan, P. K., Saffari, M., & Vlachos, I. (2020). Data analytics for sustainable global supply chains. Journal of Cleaner Production, 255, 120300.

(42)

42 Meller, R. D., Montreuil, B., Thivierge, C., & Montreuil, Z. (2012). Functional design of physical internet facilities: a road-based transit center.

Motrac Linde (2020). AutoStore. [JPG Image] Retrieved from https://www.motraclinde.nl/warehouse/automatisering/autostore

Montreuil, B.; Ballot, E.; Fontane, F. (2012). An Open Logisitics Interconnection model for the Physical Internet. IFAC Proceedings Volumes, 45(6), 327-332.

Montreuil, B., Ballot, E., & Tremblay, W. (2014). Modular structural design of physical internet containers. Progress in Material Handling Research, 13.

Montreuil, B., Buckley, S., Faugere, L., Khir, R., & Derhami, S. (2018). Urban parcel logistics hub and network design: The impact of modularity and hyperconnectivity.

Montreuil, B.; Meller, R.; Ballot, E. (2010). Towards a Physical Internet: the Impact on Logistics Facilities and Material Handling Systems Design and Innovation.

Montreuil, B.; Meller, R.; Ballot, E. (2012). Physical Internet Foundations. IFAC Proceedings Volumes, 45(6) 26-30

Montreuil, B.; Meller, R.; Thivierge, C.; Montreuil, Z. (2012). Functional Design of Physical Internet Facilities: A Unimodal Road-Based Crossdocking hub.

Nossack, J., Briskorn, D., & Pesch, E. (2018). Container dispatching and conflict-free yard crane routing in an automated container terminal. Transportation Science, 52(5), 1059-1076.

Osmólski, W., Voronina, R., & Koliński, A. (2019). Verification of the possibilities of applying the principles of the Physical Internet in economic practice. LogForum, 15.

(43)

43 Pan, S., Trentesaux, D., Ballot, E., & Huang, G. Q. (2019). Horizontal collaborative transport: survey of solutions and practical implementation issues. International Journal of Production Research, 57(15-16), 5340-5361.

Qiao, B., Pan, S., & Ballot, E. (2019). Dynamic pricing model for less-than-truckload carriers in the Physical Internet. Journal of Intelligent Manufacturing, 30(7), 2631-2643.

Rohrer, M. (1995). Simulation and Cross Docking. Winter Simulation Conference Proceedings, 1995, pp 846-849. IEEE.

Roodbergen, K. J., & Vis, I. F. (2009). A survey of literature on automated storage and retrieval systems. European journal of operational research, 194(2), 343-362.

Roy, D., Krishnamurthy, A., Heragu, S. S., & Malmborg, C. J. (2013). Blocking effects in warehouse systems with autonomous vehicles. IEEE Transactions on Automation Science and Engineering, 11(2), 439-451.

Sallez, Y.; Berger, T.; Bonte, T.; Trentesaux, D. (2015). Proposition of a hybrid control architecture for the routing in a Physical Internet cross-docking hub. IFAC-PapersOnLine, 48(3), 1978-1983.

Sallez, Y., Montreuil, B., & Ballot, E. (2015). On the activeness of physical internet containers. In

Service orientation in holonic and multi-agent manufacturing (pp. 259-269). Springer, Cham.

Schaefer, M., Čáp, M., Mrkos, J., & Vokřínek, J. (2019, November). Routing a Fleet of Automated Vehicles in a Capacitated Transportation Network. In 2019 IEEE/RSJ International Conference on

Intelligent Robots and Systems (IROS) (pp. 8223-8229). IEEE.

Smolic-Rocak, N., Bogdan, S., Kovacic, Z., & Petrovic, T. (2009). Time windows based dynamic routing in multi-AGV systems. IEEE Transactions on Automation Science and Engineering, 7(1), 151-155.

Sternberg, H., & Norrman, A. (2017). The Physical Internet–review, analysis and future research agenda. International Journal of Physical Distribution & Logistics Management.

(44)

44 Vahdani, B.; Zandieh, M. (2010). Scheduling trucks in cross-docking systems: Robust meta-heuristics. Computers & Industrial Engineering, 58, pp. 12-24.

Van Belle, J.; Valckenaers, P.; Cattrysse, D. (2012). Cross-docking: State of the art. Omega, 40(6), pp. 827-846.

Vanderlande (2020). Crossdocking. Retrieved from

https://www.vanderlande.com/warehousing/innovative-systems/sortation/cross-docking/

Venkatadri, U., Krishna, K. S., & Ülkü, M. A. (2016). On Physical Internet logistics: modeling the impact of consolidation on transportation and inventory costs. IEEE Transactions on Automation

Science and Engineering, 13(4), 1517-1527.

Vo, N. V., Berger, T., Bonte, T., & Sallez, Y. (2018). Control of Rail-Road PI-Hub: The ORCA Hybrid Control Architecture. In Service Orientation in Holonic and Multi-Agent Manufacturing (pp. 291-302). Springer, Cham.

Walha, F.; Chaabane, S.; Bekrar, A.; Loukil, T. (2014). The Cross docking under uncertainty: state of the art. 2014 International Conference on Advanced Logistics and Transport, pp. 330-335. IEEE.

Wang, N., Chang, D., Shi, X., Yuan, J., & Gao, Y. (2019). Analysis and Design of Typical Automated Container Terminals Layout Considering Carbon Emissions. Sustainability, 11(10), 2957.

Wen, M.; Larsen, J.; Clausen, J.; Cordeau, J. (2009). Vehicle Routing with Cross-Docking. The Journal of the Operational Research Society. vol 60, no 12 pp. 1708-1718.

Wickramage, N., & Ferrell, B. (2016). Intermodal Hubs in the Physical Internet.

Yan, W., Zhu, Y., & He, J. (2014). Performance analysis of a new type of automated container terminal. International Journal of Hybrid Information, 7(2), 237-248.

Yu, W.; Egbelu, P. (2008). Scheduling of inbound and outbound trucks in cross docking systems with temporary storage. European Journal of Operational Research, 184, pp. 377-396.

(45)

45 Zhang, H., Long, P., Zhou, D., Qian, Z., Wang, Z., Wan, W., ... & Chen, Y. (2016, August). DoraPicker: An autonomous picking system for general objects. In 2016 IEEE International Conference on

Automation Science and Engineering (CASE) (pp. 721-726). IEEE.

Zhen, L., Hu, H., Wang, W., Shi, X., & Ma, C. (2018). Cranes scheduling in frame bridges based automated container terminals. Transportation Research Part C: Emerging Technologies, 97, 369-384.

Zhen, L., Lee, L. H., Chew, E. P., Chang, D. F., & Xu, Z. X. (2012). A comparative study on two types of automated container terminal systems. IEEE Transactions on automation science and engineering,

9(1), 56-69.

Zhou, C., Chew, E. P., Lee, L. H., & Liu, D. (2016). An introduction and performance evaluation of the GRID system for transshipment terminals. Simulation, 92(3), 277-293.

Zhou, C., Chew, E. P., & Lee, L. H. (2018). Information-based allocation strategy for GRID-based transshipment automated container terminal. Transportation Science, 52(3), 707-721.

Zhou, C., Li, H., Lee, L. H., & Chew, E. P. (2018, August). Single Direction Traffic Rule for GRID System-An Innovative Automated Material Handling System. In 2018 IEEE 14th International Conference on

PI Cross-Docking: Applying AutoStore Design Principles