• No results found

Autonomous vehicles in last-mile logistics

N/A
N/A
Protected

Academic year: 2021

Share "Autonomous vehicles in last-mile logistics"

Copied!
53
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Autonomous vehicles in last-mile

logistics

a design science approach

March 25th, 2019

Jorrit Brolsma

Studentnumber: 2704196

E-mail: j.brolsma.2@student.rug.nl

Supervisor University:

Prof. Dr. K.J. Roodbergen

Co-Assessor University:

Dr. Ir. S. Fazi

University of Groningen

(2)

Abstract

Autonomous vehicles are an emerging concept and could offer significant opportunities within last-mile delivery. Especially since last-mile delivery is suffering from both a growing e-commerce industry and increasing urbanisation. This paper serves therefore as an introduction to autonomous vehicles in last-mile delivery and how the logistical network should be altered to accommodate autonomous vehicles. In order to investigate this, a design science approach has been used. To begin with, it was required to identify the most feasible application for autonomous vehicles within the last-mile. It turned out that this would be an autonomous parcel locker. The main reasons for this are: a reasonable capacity with subsequent efficiency advantages, more flexibility considering the location in comparison with ordinary parcel lockers and a fixed route which complies with the current capability of the technology. Due to primarily the technology, autonomous vehicles are not a realistic option in last-mile delivery at this moment. The capabilities are too limited and the current delivery methods, such as a conventional van, would perform better in terms of: costs, flexibility and efficiency. The recommendation therefore would be to start with pilots in order to stimulate the development of autonomous vehicles, because there is potential in the future.

(3)

Table of contents

Acknowledgements 4

Chapter 1 - Introduction 5

Chapter 2 - Theoretical Framework 7

2.1 - Last-mile logistics 7 2.2 - Stakeholders 9 2.3 - Autonomous vehicles 13 Chapter 3 - Methodology 15 3.1 - Overarching method 15 3.2 - Stepwise description 16

Chapter 4 - Empirical Background 20

4.1 - Autonomous vehicles in non-academic literature 20

4.2 - Interviews 23

Chapter 5 - Design Criteria 27

5.1 - Technological perspective 27

5.2 - Legal perspective 28

5.3 - Logistical perspective 29

Chapter 6 - The design 31

6.1 - Technological perspective 32

6.2 - Legal perspective 33

6.3 - Logistical perspective 33

Chapter 7 - Validation & Benchmark 36

7.1 - Validation by a panel of experts 36

7.2 - Benchmark 36

Chapter 8 - Conclusion 39

References 41

Appendix I - Interview Protocol Autonoom Vervoer Noord 47

Appendix II - Interview Protocol DHL (Depot) 48

Appendix III - Interview Protocol DHL (Engineer) 49

Appendix IV - Interview Protocol MotracLinde 51

Appendix V - Interview Protocol Wehkamp 52

(4)

Acknowledgements

(5)

Chapter 1 - Introduction

The world’s population living in urban areas is growing and The Netherlands is no exception for this (UN, 2018a; UN, 2018b). Simultaneously with this trend, e-commerce is growing as well and logistical challenges arise. More freight needs to be transported into a smaller area and this results in a tremendous pressure for carriers of parcels (AD, 2018; Parool, 2017; Zembla, 2016; ZF, 2016). Especially since it becomes more difficult to recruit sufficient workers (Berenschot, 2017). This has been proven during the peak season in The Netherlands. Black Friday and Sinterklaas followed each other in rapid succession, which resulted in a serious lack of capacity for carriers (Emerce, 2018; RTLZ, 2018).

Nevertheless, the last-mile is going to change in the coming years (Dijkhuizen, 2018). The last-mile is the final leg in a business-to-consumer delivery service where the product is delivered at the consument or at a collection point (Gevaers et al., 2011). Autonomous vehicles are a reason for the changing last-mile, since they will become a commonly used concept for this last-mile delivery (Dijkhuizen, 2018; ZF, 2016). Autonomous vehicles are in this research: “Vehicles that can drive themselves on existing roads and can navigate many types of roadways and environmental contexts with almost no direct human input” (Fagnant & Kockelman, 2015, p167). The implementation of autonomous vehicles in the last-mile would moreover revolutionize the logistical network (James & Lam, 2018). This logistical network is used within last-mile delivery to perform freight movements and factors in this are to optimally plan, manage and control the movements (Amaral & Aghezzaf, 2015; Dolati Neghabadi et al., 2018; Morfoulaki et al., 2015, In Dolati Neghabadi et al., 2018). This paper serves as an introduction to autonomous vehicles in last-mile delivery. Especially how the logistical network, should be altered to accommodate autonomous vehicles.

Last-mile delivery is currently a relevant research topic in the academic debate as well, because the number of challenges in the near future is primarily increasing, according to Savelsbergh & Van Woensel (2016). A cause of these challenges is the conflict of interests among involved stakeholders (Dolati Neghabadi et al., 2018). Consumers have for instance a growing ´desire for speed´ regarding the delivery of their parcel, but this increases the logistical costs for both the webshop and the carrier (Savelsbergh & Van Woensel, 2016). However, the webshop and carrier are industrial stakeholders in last-mile delivery and are aiming for minimizing costs (Ballantyne et al., 2013; Harrington et al., 2016). This illustrates the relevance of discussing stakeholder interests within this research.

IT is also an important driver in the current debate, because it offers both challenges and opportunities (Speranza, 2018). Autonomous vehicles are frequently mentioned as a promising trend for last-mile delivery (Dolati Neghabadi et al., 2018; Savelsbergh & Van Woensel, 2016; Speranza, 2018). James & Lam (2018) state even that autonomous vehicles could fulfill the challenges regarding the the growing volumes completely. Nevertheless, the academic debate is not unanimous about the best concept for autonomous vehicles, so multiple concepts have been investigated.

(6)

combination with trucks (Agatz et al., 2018). Trucks are furthermore used for a network design combined with driving autonomous vehicles as third option (Boysen et al., 2018).

Despite these previously mentioned concepts, autonomous vehicles are a concept that is not been investigated much yet by academic literature, while it offers a variety of advantages (Dolati Neghabadi et al., 2018). There is already evidence for the advantage that drivers could use their time differently, which could release some pressure from the labour market (Fagnant & Kockelman, 2015; Logistiek, 2018). The logistical sector in particular as mentioned before, suffers from significant labour shortage (Berenschot, 2017). Fagnant & Kockelman (2015) presume that there is no doubt that the current network should be altered in order to make autonomous vehicles useful. This goes for both the entire logistical network and the last-mile in particular (Dolati Neghabadi et al., 2018; Fagnant & Kockelman, 2015)

The aim of this paper is to explore future challenges regarding autonomous delivery vehicles within last-mile logistics. Ultimately in order to design a logistics network in which these vehicles could operate. This will be done by answering the following research question:

“How should the design of the last-mile logistics network be altered to accommodate autonomous vehicles?”

In order to answer this main research question, five subquestions are formulated:

1. What are important interests for involved stakeholders, regarding last-mile delivery?

2. Which opportunities for autonomous vehicles are available in the last-mile?

3. For which activity are autonomous vehicles the most, feasible regarding last-mile logistics, at this moment?

4. Which alterations to the current logistical network design are required to improve the accommodation of autonomous vehicles in the last-mile?

5. How could the designed network be improved, based on the validation?

(7)

Chapter 2 - Theoretical Framework

This chapter starts with an overview of the last-mile environment, because the autonomous vehicles will become a part of this environment. Subsequently, the different stakeholders and their interest will be presented since these interests are relevant for the future design criteria in Chapter 5. The final section discusses autonomous vehicles themselves, because literature describes already some capabilities of these vehicles and this will serve as an introduction towards Chapter 4. Chapter 4 describes the non-academic literature and the empirical interviews. The structure of this chapter is closely related to the six clusters of research identified by Dolati Neghabadi et al. (2018). They observe within the field of city logistics the following research clusters: definition & perimeters, policy, innovative solutions, sustainability, methods & stakeholders. Section 2.1 discusses the definitions, perimeter and methods considering the last-mile. Section 2.2 does the same with the stakeholders, policies that apply and sustainability. In the end, there is Section 2.3 that discusses the innovative solution of autonomous vehicles. 2.1 - Last-mile logistics

First of all the definitions will be discussed and in particular ´city logistics´, because this term is both used by Dolati Neghabadi et al. (2018) and has many similarities with ´last-mile logistics´. This makes that these two terms have been used interchangeable, but they are not fully the same. Cardenas et al. (2017) concluded that city logistics is at a macro level and last mile logistics at a micro level. This agrees with Dolati Neghabadi et al. (2018) that uses the term: ´last mile delivery´. All in all, city logistics has multiple relations with last-mile logistics which implies that literature with both terms could be used for this research. This has been done previously by papers, such as Lim et al. (2018).

The second point is the environment, or perimeter (Dolati Neghabadi et al., 2018), of the last-mile. The environment has been drawn in a flowchart by Aized & Srai (2014) and will be explained step by step. The flowchart includes the most frequently used methods for last-mile delivery, but lacks some details due to the relatively high aggregation level. These details are however required to identify a feasible process for autonomous vehicles in Chapter 5. In order to complement the flowchart, additional papers have been used and this results in Figure 1. Aized & Srai (2014) start at the final transit point in the supply chain, the point that receives parcels from for instance a webshop. These parcels are transported via logistics to the drop point and subsequently to the customer.

(8)

The second step in last-mile delivery, is the logistics between the transit point on the one hand and the drop point on the other hand. Logistics is an example of a previously mentioned missing process, because Aized & Srai (2014) consider this as a black-box, while different methods of transportation are possible. This is relevant, because autonomous vehicles could perform this part of the last-mile (Boysen et al., 2018; Fagnant & Kockelman, 2015). The most common way currently and therefore the biggest competitor, is by commercial vehicles such as vans (Anand et al., 2015). These vehicles might also drive on electricity in order to reduce emissions (Pelletier et al., 2016). A third option that is used on a regular basis, are bicycles intended for delivery (Oliveira et al., 2017). The final solution for logistics in the last-mile are drones, which is an innovative and relatively new concept just as autonomous vehicles (Agatz et al., 2018; Murray & Chu, 2015). Nevertheless, there are undescribed modes of transportation left, but the mentioned modes are the most important (Oliveira et al., 2017; Savelsbergh & Van Woensel, 2016; Speranza, 2018).

The third step is the drop point, the final point in the supply chain and also the location where an autonomous vehicle might deliver the parcels. Aized & Srai (2014) mention: collection points, customer home and neighbours. A relevant addition to this, are parcel lockers as it might be a solution for many challenges mentioned in Chapter 1 (Deutsch & Golany, 2018). Parcel lockers have multiple designs, but are mainly grouped in a wall and the customer is able open its specific locker with a code (Deutsch & Golany, 2018; Faugere & Montreuil, 2017). This makes it a kind of automated collection point. The distance to drop points is ideally less than 500 meters, but this is currently not achieved (Harrington et al., 2016).

Aized & Srai (2014) also describe express delivery in their model, which has multiple similarities with same-day delivery. It is an upcoming trend with consequences for last-mile logistics, because it is almost by definition inefficient such that the last-mile becomes even more expensive (Aized & Srai, 2014; Lin et al., 2018).

(9)

2.2 - Stakeholders

As mentioned in Chapter 1, conflicting interests of stakeholders are a relevant topic in last-mile logistics. These interests are therefore an important input for the design criteria. Last-mile logistics is namely a complex problem which implies that multiple perspectives should be considered (Dolati Neghabadi et al., 2018). Each perspective is presented by a stakeholder and these stakeholders interact with each other (Anand et al., 2015; Harrington et al., 2016).

This paper will use the classification of Harrington et al. (2016) to distinct different stakeholder perspectives. There are different classification options, as displayed in Table 1, but the classification of Harrington et al. (2016) is appropriate for this research, because it minimizes the potential overlap of a specific entity. Kiba-Janiak (2016) for instance uses six perspectives, such that an entity might fit in multiple perspectives which is affects the understandability of the research. Moreover, the more perspectives are included, the more extensive the research would be. This is basically an advantage, but not feasible due to limited resources. This makes the classification with five perspectives not appropriate as well (Macário et al., 2008, In Maggi & Vallino, 2016). Nevertheless, two perspectives such as Lindholm & Behrends (2012) use, would results in a less detailed research. So the classification of Harrington et al. (2016) offers a decent balance considering detail, reliability and feasibility. This is supported by the number of times that the article has been cited by other papers (e.g. Dolati Neghabadi et al., 2018).

Stakeholders (Lindholm

& Behrends, 2012)

Stakeholders (Harrington et al., 2016)

Stakeholders (Macario

et al., 2008, IN Maggi & Vallino, 2016)

Stakeholders (Kiba-Janiak, 2016)

Public stakeholders Institutional Authority & public service

Local authority

Private stakeholders Consumers Residents Residents / consumers

Industrial Retailers Receivers

Suppliers Shippers

Carriers Transport companies

Public transport operators

Table 1 - Overview stakeholders classification - (Harrington et al., 2016; Kiba-Janiak, 2016; Lindholm & Behrends, 2012; Macario et al., 2008, IN Maggi & Vallino, 2016)

(10)

The first perspective in Table 2 are the institutional stakeholders, which is the government in this research. The structure of Kiba-Janiak (2016) has been used for this perspective and gives the overarching clusters: strategy & operations, logistics infrastructure, innovation & ideas, marketing, people, regulations, environment and finance. It appears that institutional stakeholders are primarily interested in policy measures that subsequently could result in an improved quality of life (i.e. Ballantyne et al., 2013). Important aspects for this, are first of all the accessibility, which includes a proper infrastructure and a minimum amount of traffic congestion (i.e. Lindholm & Behrends, 2012). Secondly, there are environmental issues. These environmental issues consist on the one hand of a minimization of emissions and on the other hand by keeping noise disturbance as low as possible (i.e. Kiba-Janiak, 2016). In the end, institutional stakeholders aim to identify the interests of all different stakeholders in order to come to proper policies (i.e. Harrington et al., 2016).

Consumers are the second stakeholder perspective and they have their own specific interests. Although literature does not provide a structure, such as for the other stakeholders, an overarching structure with clusters has been made based on literature. Consumers are first of all interested in receiving their products and services with the highest possible service level (Harrington et al., 2016). These products and services have preferably the lowest price (Ballantyne et al., 2013). Furthermore, environmental issues are for this stakeholder a concern as well, especially from a residential point of view (Ballantyne et al., 2013; Macharis et al., 2014. Therefore, the environmental interests show a close relationship between the interests of institutional stakeholders and consumers.

Finally, the interests for industrial stakeholders, which are the webshop and carrier in this research. These interests are structured by the overarching clusters of Awasthi et al. (2016): maximizing benefits, minimizing costs, maximizing opportunities and minimizing risks. Industrial stakeholders are in general pursuing profitability. This should be achieved by first of all maximizing the benefits, such as service quality and trip reduction (i.e Allen et al. 2018; Awasthi et al., 2016). Secondly, the costs should be minimized by improving the efficiency such as reducing traffic congestion (i.e Harrington et al., 2016). This point is mentioned by the other stakeholders as well (Ballantyne et al., 2013) . Furthermore, opportunities are maximized by for instance embracing technological developments as autonomous vehicles (Awasthi et al., 2016). Fourthly, risks should be avoided as much as possible, because this could harm the profitability (Awasthi et al., 2016).

(11)

Stakeholder interests Institutional stakeholders

Strategy and operations

- Targets for environmental friendly transport ( Kiba-Janiak, 2016; Lindholm & Behrends, 2012)

- Departments responsible for transportation ( Kiba-Janiak, 2016)

- Involvement of representatives of local government (Kiba-Janiak, 2016)

- Cooperation with other stakeholders (Kiba-Janiak, 2016; Lindholm & Behrends, 2012)

- Experts’ involvement (Kiba-Janiak, 2016; Lindholm & Behrends, 2012)

- Reputation (Ballantyne et al., 2013; Harrington et al., 2016)

- Freight data (Lindholm & Behrends, 2012)

Logistics infrastructure

- Accessibility of the infrastructure (Ballantyne et al., 2013; Harrington et al., 2016; Kiba-Janiak, 2016; Lindholm & Behrends, 2012; Macario et al., 2008, In Maggi & Vallino, 2016; Macharis et al., 2014)

- Land allocation for city logistics operations (Allen et al., 2018; Harrington et al., 2016; Kiba-Janiak, 2016; Lindholm & Behrends, 2012; Macharis et al., 2014)

- Urban space planning takes intensity of traffic into account (Harrington et al., 2016; Kiba-Janiak, 2016; Lindholm & Behrends, 2012)

Innovation and ideas

- Application of intelligent transportation tools in order to improve the system (Kiba-Janiak, 2016)

- Application of innovative ideas (Kiba-Janiak, 2016)

- Monitoring future trends (Kiba-Janiak, 2016)

- Technology and innovation transfer (Harrington et al., 2016)

Marketing

- Identification of city logistics stakeholders’ needs (Ballantyne et al., 2013; Harrington et al., 2016; Kiba-Janiak, 2016)

- Promotion of collective or ecological transport (Harrington et al., 2016; Kiba-Janiak, 2016)

People

- City logistics stakeholders’ experience in the implementation of ideas (Kiba-Janiak, 2016; Macharis et al., 2014)

- Security in road traffic (Kiba-Janiak, 2016)

Regulations

- Comply with legislation (Ballantyne et al., 2013; Kiba-Janiak, 2016; Macario et al., 2008, In Maggi & Vallino, 2016)

Environment

- Noise reduction (Ballantyne et al., 2013; Kiba-Janiak, 2016; Macario et al., 2008, In Maggi & Vallino, 2016; Macharis et al., 2014)

- Air pollution reduction (Ballantyne et al., 2013; Harrington et al., 2016; Kiba-Janiak, 2016; Lindholm & Behrends, 2012; Macario et al., 2008, In Maggi & Vallino, 2016; Macharis et al., 2014)

Finance

- Economic benefits for city logistics stakeholders accruing from implementation of projects

(Harrington et al., 2016; Kiba-Janiak, 2016; Macharis et al., 2014)

(12)

Consumers

Service quality

- Reliability (Ballantyne et al., 2013; Macario et al., 2008, In Maggi & Vallino, 2016; Macharis et al., 2014)

- Localisation of facilities (Harrington et al., 2016; Lindholm & Behrends, 2012)

- Lead-time (Harrington et al., 2016)

- Delivery window (Harrington et al., 2016; Macharis et al., 2014)

- Track and trace (Harrington et al., 2016; Macharis et al., 2014)

- Personal and friendly contact (Harrington et al., 2016)

- Flexibility ( Harrington et al., 2016)

Costs

- Lowest transportation costs (Ballantyne et al., 2013; Kiba-Janiak, 2016; Macharis et al., 2014)

- Compensation by low service performance ( Harrington et al., 2016)

- Service price (Harrington et al., 2016)

- Insurance options (Harrington et al., 2016)

Environment

- Attractive environment to life (Ballantyne et al., 2013; Macharis et al., 2014)

- Noise reduction (Ballantyne et al., 2013; Macario et al., 2008, In Maggi & Vallino, 2016; Macharis et al., 2014)

- Air pollution reduction (Ballantyne et al., 2013; Macario et al., 2008, In Maggi & Vallino, 2016; Macharis et al., 2014)

Industrial stakeholders

Maximize benefits (Ballantyne et al., 2013)

- Number of customers served (Awasthi et al., 2016)

- Proximity of customers (Allen et al., 2018; Awasthi et al., 2016)

- Service quality (Allen et al., 2018; Awasthi et al., 2016; Harrington et al., 2016; Macharis et al., 2014)

- Goods consolidation (Awasthi et al., 2016; Lindholm & Behrends, 2012)

- Trip reduction (Awasthi et al., 2016; Harrington et al., 2016; Lindholm & Behrends, 2012; Macharis et al., 2014)

- Green resource sharing (Awasthi et al., 2016)

Minimize costs (Ballantyne et al., 2013)

- Training (Awasthi et al., 2016)

- Location and relocation (Awasthi et al., 2016)

- Hiring costs (Awasthi et al., 2016)

- Resources (Awasthi et al., 2016; Macharis et al., 2014)

- Distribution (Awasthi et al., 2016; Kiba-Janiak, 2016)

- Traffic congestion (Harrington et al., 2016; Lindholm & Behrends, 2012; Macario et al., 2008, In Maggi & Vallino, 2016)

- Delivery failures (Allen et al., 2018)

Maximize opportunities

- Environmental impact reduction (Awasthi et al., 2016; Macharis et al., 2014)

- Energy conservation (Awasthi et al., 2016)

- Participation in green initiatives (Awasthi et al., 2016)

- Technological development (Awasthi et al., 2016; Harrington et al., 2016)

- Knowledge transfer (Awasthi et al., 2016)

- Organizational growth (Awasthi et al., 2016; Harrington et al., 2016)

(13)

Minimize risks

- Loss of clientele (Awasthi et al., 2016)

- Lack of resource sharing commitment (Awasthi et al., 2016)

- Lack of information sharing commitment (Awasthi et al., 2016)

- Unfair profit allocation (Awasthi et al., 2016)

Table 2 - Overview stakeholders and their interests

2.3 - Autonomous vehicles

Freight delivery in the last-mile with autonomous vehicles is not much investigated yet (Dolati Neghabadi et al., 2018; Speranza, 2018). To the best of knowledge, only three articles did a significant suggestion. Boysen et al. (2018) describe a combination of small delivery robots deployed from a truck with promising result regarding efficiency. Alessandrini et al. (2015) at their turn, create a vision about the future role of autonomous vehicles within the domain of last-mile logistics. The same did Luettel et al. (2012) with a more specific focus on trends and available systems. This final article however, is already a couple of years old, which presumably is a drawback in such a rapidly developing environment.

Fagnant & Kockelman (2015) state that, for autonomous vehicles in general, there are multiple barriers that currently oppose implementation such as: Vehicle costs, certification, liability, security and privacy (Fagnant & Kockelman, 2015). Simultaneously, there are even bigger opportunities: drivers save time in order to do other activities, safety improves, traffic congestion reduces and both fuel consumption and pollution decreases (Fagnant & Kockelman, 2015; Luettel et al., 2012).

Dobrev et al. (2017) state moreover that, regarding positioning, it is still very difficult to have a secure and reliable navigation within urban areas, because GPS for instance is not accurate enough. Moreover, autonomous vehicles should deal with other road users, but are already capable to detect and predict the behavior of other road users (Millard-Ball, 2018). Furthermore, besides moving objects, autonomous vehicles should have also the possibility to interact with static objects as traffic signals even in situations with limited visibility (Perez et al., 2010).

Despite these technological criteria, Kyriakidis et al. (2015) state, based on Underwood (2014), that the most critical barrier would be the legislation. This is also supported by Fagnant & Kockelman (2015) who do not mention technology in their list of barriers for implementation. Millard-ball (2018) acknowledge this potential barrier as well and suggests on the one hand to reduce the legal position of pedestrians. On the other hand manufacturers of autonomous vehicles should not be liable in case of a collision with a pedestrian.

(14)

a wire, tape or laser markers. Free ranging AGVs navigate on vision, radars or GPS. There are even more sophisticated systems as WLPS (Wireless Local-Positioning System), but these systems required a complex infrastructure of anchor nodes (Dobrev et al., 2017). The same goes for path following techniques which makes this option also complex to implement.

Furthermore, AGVs offer significant advantages considering costs and efficiency (Bechtsis et al., 2017). Moreover, AGVs are both able to work independently and in cooperation with humans. This implies that all the more autonomous vehicles could learn from AGVs, because autonomous vehicles are in an early stage of development and aiming for the roughly the same capabilities as AGVs (Fagnant & Kockelman, 2015).

Considering fuel and pollution, another relevant topic is the electric drivetrain, as this is considered as more sustainable (Pelletier et al., 2016). Consequently to an electric drivetrain, the limited battery range is a relevant issue. The battery should be recharged occasionally, but electricity offers also significant benefits in comparison with conventional fuels, at least for AGVs (Bechtsis et al., 2017). Charging could be done by either existing infrastructure or mobile charge stations such as for instance a truck (Boysen et al., 2018; Rabta et al., 2018). James & Lam (2017) discussed already a vehicle routing problem that copes with this challenge. Especially considering the most efficient way to implement this within their operational activities: delivering freight. Routing in itself is a decent way to improve the efficiency for AGVs, just as scheduling (Fazlollahtabar & Saidi-Mehrabad 2015). The general objective functions for these issues are presented in Table 3 and for autonomous vehicles in the last-mile, these are presumably not significantly different.

No. Scheduling Routing

1 Makespan The global transportation cost

2 Total (Weighted) Completion Time Drivers and vehicles fixed costs

3 (Weighted) Mean Flow Time Number of vehicles and/or drivers

4 Mean Waiting Time Balancing of the routes

5 Lateness Penalties for not/partially served customers

6 Tardiness -

(15)

Chapter 3 - Methodology

This chapter discusses the methodology of this research on the basis of two sections: Overarching method and Stepwise description.

3.1 - Overarching method

This research will use a design science approach, which is defined as: “A natural approach that: analyzes the problem, designs a solution and develops it further in cycles of testing and redesign (Van Aken et al., 2016, p2). The first two subquestions from Chapter 1 have partially been discussed in Chapter 2 and this will be completed in Chapter 4. Secondly, this results in a list of design criteria in Chapter 5 to which the solution must comply, so a list of requirements which is among other based on the interests of Table 2. Based on these design criteria, the most feasible activity will be identified, as formulated in the third subquestion of Chapter 1. This solution will subsequently be designed in Chapter 6, based on these criteria, in order to develop a feasible application. In the end, the solution will be validated by a panel of experts and benchmarked with other options in Chapter 7. This structure aligns with Hevner (2004) and will be discussed in the action plan of Section 3.2.

There are multiple reasons for a design science approach. First of all, this study has an exploratory aim. As mentioned before in Section 2.3, there is not much academic literature available yet regarding last-mile delivery with autonomous vehicles, so a further exploration would be necessary. Combined with an envisioned concept as autonomous vehicles, this research could increase the general knowledge about both autonomous vehicles and last-mile logistics.

Secondly, the approach fits very well into the current trend towards more practical research (Holmstrom & Ketokivi, 2009; O’Keefe, 2014; Van Aken et al., 2016; Van Mieghem, 2013). This research uses a significant amount of non-academic literature, due to previously mentioned newness of autonomous vehicles. A design science approach offers the opportunity to use such non-academic literature (Van Aken et al., 2016). Moreover, the ultimate deliverable of this research will be a design that could be used directly and this opportunity is also given by a design science approach (Van Aken et al., 2016).

Thirdly, a design science approach offers the opportunity to implement the experience of AGVs into a new design for autonomous vehicles in last-mile delivery. Exploration and designing based on AGVs is possible, because this technology is in a mature stage of development (Bechtsis et al., 2017). This research aims to identify the useful design criteria from AGVs that are relevant for autonomous vehicles as well.

Useful design criteria for a last-mile logistics network design do also derive from Table 2. These interests from different stakeholders offer a decent introduction to design criteria that are necessary and desirable for a feasible design.

(16)

3.2 - Stepwise description

Figure 2 - Schematic overview of methodology

For this research, the five design evaluation methods of Hevner (2004) will be used: observe, analyse, experiment, test and describe. This guideline contains all relevant topics in a design science research and is used in a comparable research about “crowd logistics” (Frehe et al., 2017). A schematic overview of this approach, with the structure of Hevner (2004) included, is presented in Figure 2 and will be discussed in this section. Chapter 7 requires a remark, because this chapter includes two methods of Hevner (2004): Test and Describe. The structure displays simultaneously the stepwise action plan of this research.

Observe

With the theoretical background composed in Chapter 2, the next step is to observe the empirical environment during Chapter 4. Especially to expand the obtained knowledge and fulfill gaps that could not been answered by academic literature.

(17)

On the other hand, non-academic literature has been used in Section 4.1 such as news articles, because autonomous vehicles are relatively new and the development goes rapidly. This additional source of information adds valuable information due to the more recent publications.

Within this research, there is consciously chosen to ignore the customer perspective during the interviews. Especially based on the assumption that this perspective does not offer valuable information in this stage, because customers are not familiar with the concept of autonomous delivery vehicles yet. The customer perspective however, is included in the theoretical part in Section 2.2.

Nevertheless, Dolati Neghabadi et al. (2018) state in Chapter 2 that city logistics should be investigated from different stakeholder perspectives, so including customers. This implies that it would beneficial to investigate all three stakeholder perspectives (Harrington et al., 2016). However, I assume that this would extend the research significantly. Due to this extension, it would affect the quality and reliability of the results. Therefore, the focus will be on the industrial perspective and the other perspectives are included more laterally.

First of all, information regarding the current situation of autonomous vehicles is required. Therefore, the platform Autonoom Vervoer Noord has been interviewed as a general expert and in particular a project member who is working for the government as well. This stakeholder has not been mentioned by Harrington et al. (2016), but has strong connection with the institutional stakeholders. There is also a need for legal information since this is considered by Kyriakidis et al. (2015) as crucial barrier. Therefore, an account manager of the Dutch inspection institute (RDW) has been interviewed.

Secondly, the current environment of last-mile delivery is relevant for this research. Especially since the autonomous vehicle should operate in this environment, more information is required to identify potential barriers and opportunities. For this perspective, to persons from DHL has been interviewed: a depot manager for more operational information and an engineer for the tactical aspects. These experts are both industrial stakeholders (Harrington et al., 2016).

Thirdly, within last-mile delivery a party with relatively much power are the webshops, because these companies send the parcels and most often paying as well. This makes it a relevant industrial stakeholder to interview. Hence there has been spoken with a manager (Regiemanager) of Wehkamp.

Fourthly, academic literature does not provide much information considering the technological aspects of autonomous vehicles. This gap needs to be fulfilled in order to give a decent answer to the research question of this paper. For this information, a supplier of AGVs has been interviewed, because there is more experience with these vehicles than with autonomous vehicles on public roads as discussed in Section 2.3 (Dobrev et al., 2017). This supplier does not fit within the classification of Harrington et al. (2016), but important enough to use for this research.

(18)

explanation of the observing stage could be found in the interview protocol presented in Appendix I-V.

Furthermore, it is intended to do the interviews face-to-face. An additional advantage of face-to-face interviews, is the possibility to observe things yourself. For example how parcels are handled in distribution hubs. Nevertheless, due to practical reasons, the interview with the Dutch inspection institute was done by phone. This is explicitly stated in the transcript.

Figure 3 - Overview of empirical perspectives Analyse

(19)

Experiment

This step is about experimenting, which is building the logistical network design and presented in Chapter 6. The starting point are the necessary design criterion from Chapter 5, because these are the criteria that the design should meet at least. Subsequently, the desirable design criteria are used to improve the design further. The design will be structured by the same perspectives as the design criteria to increase the comprehensibility.

Test and Describe

Once the design is created, it will be validated in Chapter 7. For this validation, a panel of experts will be composed. This panel consists out of both people that have been interviewed before and new people. Furthermore, the members of this panel have a diverse background. There are people from: a webshop, a carrier, a start-up operating in the last-mile and multiple people from an academic environment. The intention is to start an open discussion in which the developed design will be validated. This discussion is guided by myself. Especially in order clarify uncertainties and to adjust the discussion when required. In the end, the aim is to judge about the feasibility of the design and to identify the points for improvement.

The validation will be used, combined with literature and the input from the earlier interviews, to create a benchmark with other methods of last-mile delivery. On the one hand with the most popular method at this moment, as identified by the webshop in Chapter 4 and on the other hand a less complex alternative that will be explained in Chapter 7.

Describe

(20)

Chapter 4 - Empirical Background

This chapter discusses the non-academic environment, because autonomous vehicles are a relatively new phenomenon which results in a limited amount of academic literature. However, there is already a significant amount information on the internet available to complement this. The first section discusses this non-academic literature. Secondly, companies and other experts have been interviewed and these results are discussed in Section 4.2.

Furthermore, Section 4.1 makes a distinction between three perspectives: technology, legal and logistics. Technology is relevant, because there has not been written much yet, while it offers significant barriers (Fagnant & Kockelman, 2015; Dobrev et al., 2017). Legislation is also relevant, because some research even state this as the most crucial barrier (Fagnant & Kockelman, 2015; Kyriakidis et al., 2015). Finally the logistical perspective that is relevant, because this section includes concepts that already exist and therefore could provide inspiration for the future design. This is moreover a required addition to the limited academic literature (Dolati Neghabadi et al., 2018).

Section 4.2 is divided by interviewed perspective, as presented in Figure 3. Through this approach, it should be clear to which perspective the specific information belongs.

4.1 - Autonomous vehicles in non-academic literature

This section uses the following three perspectives: Technological, Legal and Logistical. These perspectives will be used in Chapter 5 and 6 as well.

Technological perspective

Autonomous vehicles will become a commonly used concept for last-mile logistics, as stated previously (Dijkhuizen, 2018; ZF, 2016). The New York Times (2018) states, according to an expert, that: “Fully unstructured driving by go-anywhere cars is a long time away.” This is primarily due to the current state of technology.

The technology behind autonomous vehicles consists of various component such that it becomes complicated to keep an overview. Therefore a taxonomy exists. The Society of Automotive Engineers (SAE) developed an international standard (SAE International, 2014). This standard (J3016) offers 6 levels, starting at level 0 that implies no automation, towards level 5 of full automation and is displayed in Table 4. An important segregation is observable between level 2 and level 3 (SAE International, 2014). Up to and including level 2, the human driver performs the tasks of ´monitoring of driving environment´, which makes it not a complete autonomous vehicle. From level 3, the automated driving system performs the monitoring as well. Within this research, the focus is especially on level 5 since that is the ultimate goal of autonomous vehicles. Moreover, in case a human remains required, the added value of an autonomous vehicle decreases. Kyriakidis et al. (2015) support this with claiming, based on Underwood (2014) who conducted a survey among experts, that level 3 is already achievable in 2018, but is not very practical yet. Level 4 will be deployed in 2025 and level 5 even in 2030. So there is a clear build-up.

(21)

partner at Trucks Venture Capital and a lecturer in transportation at Stanford University (The New York Times, 2018).

SAE

level Name Execution of steering and acceleration/ deceleration Monitoring of driving environment Fallback performance of Dynamic Driving Task System Capability (Driving Nodes

Human driver monitors the driving environment

0 No Automation Human driver Human driver Human driver n/a

1 Driver Assistance Human driver and system

Human driver Human driver Some driving nodes

2 Partial Automation System Human driver Human driver Some driving nodes Automated driving system (“system”)

monitors the driving environment

3 Conditional Automation System System Human driver Some driving nodes

4 High Automation System System System Some driving nodes

5 Full Automation System System System All driving nodes

Table 4 - Levels of automation for on-road vehicles (SAE International, 2014) Legal perspective

As mentioned in Chapter 2, Kyriakidis et al. (2015) argue that legislation would be the most significant barrier for the implementation of autonomous vehicles. For the Netherlands however, this does not seem to be the case. The Netherlands are, according to themselves, precursors in the field of legislation and experimenting with autonomous vehicles (Toezine, 2018). This is supported by the recent ´Experimenteerwet´ that allows remote supervision (Rijksoverheid, 2017). Due to this law, it is no longer mandatory that a human is physically inside the vehicle. Moreover, the Dutch inspection institute (RDW) is very prestigious due to its independency and grants, resulting from this, the most certificates in the world (Toezine, 2018). So legislation forms a barrier to a lesser extent than previously thought and the Dutch RDW seems to be a decent institution to use as guideline for the design.

The RDW (2017) requires some conditions for pilots with autonomous vehicles and these conditions are presented in an assessment framework. This framework is relevant, because the future design needs to pass this framework as well. The sequential process steps of this framework are presented in Figure 4. This implies that the next step can only be started when the previous step has been completed.

(22)

should be presented in a detailed action plan, provided by the applicant, that needs to be assessed.

Secondly, there is the ´preparation´ step in which the RDW demands a complete document with preconditions about: the vehicle (i.e. the vehicle is sound, safe and responds adequately to malfunctions), behaviour (i.e. able to respond to other road users and observes traffic rules) and others (protection against unintended use and no safety risks due to adjustments during the test). Furthermore, all the involved parties will be approached. These parties are at least: The RDW, The Institute for Road Safety Research, Authorities and Advisory parties.

The third step is the ´assessment´ in which the vehicle should pass both a ´happy flow test´ and a ´stress test´. So the vehicle needs to prove in a closed test environment that it acts appropriate in both ordinary and extraordinary situations.

The fourth step is executing the real pilot. During this pilot, safety has the highest priority. In case of an accident with material damage or personal injuries, ´driving with the exemption is suspended with immediate effect´ (RDW, 2017, p.10).

In the end, the final step is to evaluate the project with all involved parties in order to expand the general knowledge.

Figure 4 - Process steps of the assessment framework provided by the RDW (RDW, 2017, p.3)

Logistical perspective

There are already some concepts for autonomous vehicles in the last-mile currently, each with their own logistical characteristics. These concepts give an impression regarding the potential possibilities and they can serve as an inspiration for the design.

For the transportation of substantial amounts of freight, the concept of a large self-driving pod has been introduced by for instance Mercedes-Benz (Bestelauto, 2018). These pods are suitable for routes with many parcels to the same destination, for instance servicepoints. The vehicles are basically a car without a chauffeur, driving on ordinary roads while transporting a ´bulk´ of parcels. Since this vehicle drives to only one destination, the flexibility is not very high. For every customer, the vehicle should return to its starting location to be reloaded, assumed that it is not desirable that one customer has access to the parcels of other customers.

The concept of Alessandrini et al. (2015) offers more flexibility from this perspective. They also mention self-driving pods, but suggest simultaneously a concept with a train. This results in train with seperate, independent pods such that customers do not have access to other people’s parcels. Furthermore, each pod has the ability to drive solitary, without the other trailers of the train, to deliver to a specific customer. Moreover, these trains are narrow enough to drive on sidewalks, which increases the applicability of the concept.

(23)

carriages could have different purposes. One carriage could be used as parcel locker while another is a ´bulk trailer´ for parcels. Moreover, the train of NEXT offers the option to be driven by a human. This increases the feasibility, because it turn the vehicle into a level 3 automation rather than level 4 or 5.

Finally, there are presumably also situations with no large customers included. For these situations there has been developed a solution as well. The Estonian company Starship offers small delivery robots to deliver products to only one customer (Starship, 2018). So for very small volumes. These vehicles drive on sidewalks with a limited speed, but due to the limited capacity, Starship has started a joint-venture with Mercedes-Benz to combine these robots with vans. As a consequence, the robots could be reloaded quicker such that the robots become more efficient. Boysen et al. (2018) have been investigated the concept already and found promising results regarding a significant reduction of vans while the same number of parcels could be delivered. 4.2 - Interviews

The previous chapter described the four different perspectives for empirical research: Carriers, Webshops, Suppliers of autonomous vehicles and General experts. These perspectives will be discussed one by one to complement and verify the information in Chapter 2 and Chapter 4.1 such that a list of design criteria could be composed in Chapter 5.

Carrier perspective

First of all the carrier perspective for which there has been spoken with both an engineer and a depot manager. The main purpose of these interviews is to shape the environment of the last-mile and to understand the current processes, because an autonomous vehicle should replace one of these processes.

Regarding this environment, the carrier agrees that volumes are rapidly growing and that the challenges to cope with these volumes increase. The engineer thinks even that the current logistical network is untenable in the future, because the flows become too big and too diffuse.

Consolidation of parcels is therefore crucial, as the engineer told: “We would like to consolidate these parcels, because that is the most efficient.” The higher the consolidation of parcels, the higher the efficiency. Looking at the specific supply chain of the carrier, this aim for consolidation is observable as well and displayed in Figure 5. Parcels are consolidated as long as possible such that the volume is sufficient for larger trucks. As a consequence, the final leg of the whole supply chain is relatively short and this results subsequently into a minimal loss of efficiency due to not fully loaded vehicles. These vehicles carry “approximately 100 parcels” per route, according to depot manager.

However, regarding the scope of this research, this consolidation has a consequence. The flow between the terminal and depot is not suitable, because this is done with large trucks. There are only a few terminals in The Netherlands, but number of depots is much larger. These depots serve a range of approximately 25 kilometers and are owned by the carrier itself. Since these depots are owned by the carrier, there are dedicated people available for the sorting process. This is beneficial for autonomous vehicles, because these people could take care of these vehicles.

(24)

range of 2 kilometers, are these points owned by a retailer. According to the engineer: “The retailer has its store as well. He prefers not to unload such a thing (autonomous vehicles).” The retailer receives a financial compensation for the actions it performs for the carrier, but the compensations are not very high. This results in a situation that the retailer wants to put as less as possible effort in the parcels, because it is not their core-business.

Furthermore, servicepoints are not attractive, because these points are not relevant for every parcel. Depending on the choice of the consumer, the third step of Figure 5 could be skipped and the parcel could be delivered directly from the depot to the consumer. Therefore, this step has been colored differently. Deployment from a servicepoint would result in an additional transport from the depot to the servicepoint, before the parcel goes to the consumer.

Finally, autonomous vehicles have potential, but should offer more than just replacing a driver. This is not sufficient argumentation for the carrier to implement autonomous vehicles. An example given, was regarding parcel lockers. This is a promising concept for the carrier, but local authorities are very preserved with permissions for placing these lockers. When these lockers have wheels and become a vehicle, this might bypass this issue.

Figure 5 - Flowchart processes carrier Webshop perspective

Service quality is very important for the webshop and delivery is a very important aspect within the complete customer journey, because “it is the only stage in which you make the customer a promise.” Namely, when and where their parcels will be delivered. Webshops share most of the concerns of carriers: the volumes are growing and also the challenge to deliver everything on time. As a consequence, the service quality is an even bigger issue.

The service level is therefore monitored by the webshop and there are agreements between the webshop and the carrier considering the service level. The person who delivers the parcel is part of the customer journey and influences the perceived service level. It is notable that the deliverer could especially influence the service level negatively. When this person does its job properly, it is perceived as ´normal´ by the customer, but when this person does not, it is experienced negatively.

There is however no explicit need for a physical person who delivers the parcels, which is beneficial for autonomous vehicles. In case the physical person will be removed from the process, it is still possible to achieve a sufficient service level. “Personal contact is not necessarily important.” This offers opportunities for autonomous vehicles, because it means that the disappearance of a physical person is not a crucial barrier.

(25)

General Experts

Besides the general side of last-mile logistics, there is also the more technical aspect of autonomous vehicles. As described in Chapter 3, a platform and the Dutch Inspection Institutie (RDW) has been interviewed as general experts.

Legislation is an important point of attention, because an autonomous vehicle is not allowed to drive on the public road without a permission. Such a permission is provided by the local authorities and inspection institute.

Legislation however, is not the crucial barrier for implementation which is supported by the RDW that claimed that: “The law is ahead with respect to the technology.” Traditionally, the law contains plenty of references to a physical driver. However, the ´Experimenteerwet´ has recently been created, which allows the RDW to grant additional exemptions in order to test with autonomous vehicles. The RDW has moreover a proactive attitude towards these tests and is closely involved, because their goal is to expand the available knowledge.

Nevertheless, the most important thing for both experts is safety. They prefer an unnecessary emergency stop above an accident. Without a guaranteed safety, it is unacceptable that an autonomous vehicle drives and the RDW does not provide an exemption for the law as well for a specific pilot with autonomous vehicles. Consequently, although autonomous vehicles could drive faster, it is currently not allowed to drive faster than 15 km/h according to The Platform.

Safety is especially important, because autonomous vehicles are still in an early stage of development currently. The capabilities are limited and further development is required. Autonomous vehicles are currently only capable to drive a pre-programmed route. The vehicles are not able to leave this route, not even to get around an obstacle. When an obstacle occurs, the vehicles stops for safety reasons, regardless what obstacle it is. According to The Platform: “Technology is not able yet to distinct an object and to judge whether the vehicles can continue its journey or not.”

Supplier of autonomous vehicles

Limitations regarding technology are also highlighted by the expert of MotracLinde, a supplier of autonomous vehicles. This supplier offers primarily vehicles for indoor use, ´often it is even a prerequisite´ and for a reason: it seems relatively difficult to create a working solution outside. Especially since there are many factors that could influence the performance of the autonomous vehicle.

Positioning is one these factors and according to the interviewee: “Visual sensors are the most opportune.” Even though the most commonly used technique are sensors, because these sensors are the most easily applicable. The expert told that: “You need to consider which technology performs the best for this application.” In this case, vision would perform presumably better than sensors or GPS.

(26)

Ideally, magnets are placed underneath the street. However, this requires a drastic adjustment to the existing infrastructure while there is not even an international standard yet.

A technology with visual is presumably easier to create and requires first of all clear references along the route such as clearly visible lines. These references need to be kept in a proper condition, otherwise the performance of the vehicles deteriorates and especially in case of bad weather. Furthermore, the surface needs to be as flat as possible since this influences the positioning and subsequently the performance as well.

Secondly, the technology is not able yet to avoid obstacles and the supplier perspective supports on this point the expert of Autonoom Vervoer Noord. An autonomous vehicles will make an emergency break when there is an obstacle on its route. The vehicle is not able yet to identify the obstacle in order to decide whether it can continue or not. Moreover, an autonomous vehicle is currently not capable to leave its route. This suggest that there should be a person available who can intervene the vehicle and helps it to continue the route.

(27)

Chapter 5 - Design Criteria

This chapter analyzes both the theoretical framework of Chapter 2 and the non-academic literature, combined with the interviews, from Chapter 4. The analysis results in a list of ´necessary design criteria´, criteria that the design must meet as a minimum and ´desirable design criteria´ that would improve the design. These criteria are summarized in Table 5 and separated into a technological, legal and logistical perspective.

5.1 - Technological perspective

The first section discusses both the necessary and desirable design criteria from a technological perspective.

Necessary Design Criteria

The first necessary design criterion is that an autonomous vehicle is not suitable for home delivery, because it is not capable to deal with: gates, stairs or other obstacles. These are obstacles that are previously mentioned by the supplier in Chapter 4 as point of attention. Home delivery however is the most popular delivery method nowadays according to the webshop in Chapter 4, which makes this a relevant criterion.

A second necessary design criterion is that there should be a fixed route between fixed locations, because the route should be programmed in advance, according to the supplier of autonomous vehicles. Moreover, an autonomous vehicle has not the intelligence to leave this route, because a level 5 of automation is required while only level 3 is currently available as mentioned in Section 4.1 (Kyriakidis et al., 2015; SAE International, 2014). This criterion also supports the previous criterion regarding home delivery, since home delivery does not have fixed locations. Depending on the available parcels, the delivery locations change.

Thirdly, the vehicle requires clear reference points to follow its route, such as clearly visible road markings, according to the supplier in Chapter 4. These markings should be well visible, since they need to be visible at night and with bad weather as well. This requires relative small alterations to the current infrastructure. This design criterion becomes more challenging in case there is no fixed route, since more roads must contain these reference points.

Fourthly, the route should not be too long, because an autonomous vehicles does not drive that fast according to The Platform in Chapter 4. The maximum speed will be presumably 15 kilometers per hour such that a 30 kilometers journey for instance will cost already 2 hours in an ideal situation. Furthermore regarding the previous criterion, a longer route would result in more reference points as well.

Desirable Design Criteria

(28)

Secondly, it would be desirable to implement the objective functions of Fazlollahtabar & Saidi-Mehrabad (2015) from Section 2.3, because this would optimize the routing and scheduling and therefore increases the efficiency. Such an optimization would namely improve the business case and matches the interests from both industrial stakeholders and consumers for a minimization of costs. The costs are also mentioned by Fagnant & Kockelman (2015) as an important barrier for autonomous vehicles, as described in Section 2.3.

A third desirable design criterion is a level 5 of automation, because this would result in a vehicle that does not require any supervision from a human at all, as appears from Table 4. Kyriakidis et al. (2015) stated moreover in Chapter 4 that the current level 3 of autonomation is not very practical.

5.2 - Legal perspective

The second section considers the legal perspective with the same distinction as the previous section: necessary and desirable design criteria.

Necessary Design Criteria

Regarding the legal perspective, safety is first of all the most important necessary design criterion. Especially since this has the highest priority for both the government and the RDW, as appeared from Section 4.2. Kyriakidis et al. (2015) claimed even in Section 2.3 that legislation would be the most important barrier. This claim however, is questionable in The Netherlands since especially the RDW has a positive attitude towards autonomous vehicles as described in Chapter 4. As long as the safety is guaranteed, exceptions to the law are possible. Although the law is currently focussing on pilots only, legislation will not be the critical barrier in The Netherlands.

A second necessary design criterion from a legal perspective is nevertheless that the vehicle should be sound. This is an important criterion from the assessment framework as discussed in Section 4.1. A vehicle that is sound, supports namely the previous criterion regarding safety.

Thirdly, an autonomous vehicles should behave appropriately towards other road users and has to comply with traffic laws. This criterion derives from the assessment framework from Section 4.1 as well and is closely related to safety. This supports the findings of Millard-Ball (2018) in Chapter 2, because this author highlights also the importance of this design criterion. Desirable Design Criteria

(29)

5.3 - Logistical perspective

The third and final section of this chapter discusses the design criteria from a logistical perspective.

Necessary Design Criteria

A first necessary design criterion is the capability to carry as many parcels as possible in an ideal situation, because this increases the efficiency of the vehicle. However there is presumably a maximum for this, since the vehicle is not able to serve every neighbourhood when the vehicle is too big. An additional drawback could be that it is more difficult to achieve permissions for a bigger vehicle, because it is considered as less safe, which is a necessary design criterion from a legal perspective. Still, the need for efficiency is consistent with the interests from different stakeholder perspectives, as described in Table 2. It results subsequently in a trip reduction for the industrial stakeholders (e.g. Lindholm & Behrends, 2012; Macharis et al., 2014). The institutional stakeholders benefit from this trip reduction as well, because it leads to less traffic and therefore an improved accessibility (e.g. Ballantyne et al., 2013; Harrington et al., 2016; Kiba-Janiak, 2016). An additional benefit of the reduced traffic, is the positive consequences for the environment which is an interest for all perspectives (e.g. Awasthi et al., 2016; Ballantyne et al., 2013).

Secondly, the solution with an autonomous vehicle should have added value compared to current methods of delivery, because otherwise there is no need to implement the proposed solution. This added value could be present in terms of efficiency as mentioned with the previous criterion, but a relief of the shortage of employees discussed in Chapter 1, could be a reason as well (Berenschot, 2017). Chapter 7 will discuss a benchmark to verify this criterion.

Thirdly, the parcels should be loaded into the vehicle and unloaded from the vehicle as well. In the current supply chain, the driver performs these processes, but an alteration of the logistical network is required for this aspect when this driver disappears. Without this criterion, parcels will not move into the vehicle or move out of the vehicle.

For the unloading, enough space around the vehicle is required. A customer would otherwise not be able to open the door of the locker and moreover to pull the parcel out of the locker. Furthermore, in case this criterion is not fulfilled, it might harm the customer-journey, which is important for the webshop as appeared from Chapter 4. Moreover, this could even result in a loss of clientele which should be minimized according to Awasthi et al. (2016) in Table 2.

Another necessary design criterion is the population density. For cost efficient reasons, there should be enough people living in the area of delivery. Maximizing the number of customers served by an autonomous vehicle is also an interest for industrial stakeholders, as appeared from Table 2. Driving to multiple locations is a favourable option neither, because both trip reduction was an interest from Table 2 and a longer routes offers multiple drawbacks as discussed in Section 5.1.

(30)

choose for a delivery method with autonomous vehicles or, in case there is no choice, to choose for a webshop that uses autonomous vehicles.

Desirable Design Criteria

On top of the final necessary design criterion, it is even desirable that the vehicle stops within a range of 500 meters from the customer. Not only because this results in an improvement of the current situation, but also because this is the targeted range (Harrington et al., 2016).

Another desirable design criterion from a logistical perspective, is the aim for a solution that offers more advantages than just replacing a driver, according to the carrier in Chapter 4. This goes a step further than the previously discussed necessary design criterion to offer added value compared to current delivery methods. Although there is a shortage on the labour market, according to Berenschot (2017) in Chapter 1, this advantage will change as soon as things get worse economically. The carrier suggested therefore autonomous parcel lockers in Chapter 4, because these would be easier to place than conventional parcel lockers. Especially due to permissions for parcel lockers that are difficult to obtain.

Furthermore, a location with not too much traffic is desirable due to safety reasons. Especially since the expert of the RDW indicated safety as the highest priority in Chapter 4. A busy street could results in an unsafe situation when a customer empties its locker and this could also harm the customer-journey as discussed in Section 4.2.

Necessary Design Criteria Desirable Design Criteria

Technological Perspective

● Not suitable for home delivery ● A fixed route between fixed

locations

● Clear reference points along the route

● The route should not be too long

● As less (potential) obstacles along the route as possible ● Implement optimization

objectives regarding routing and scheduling

● The vehicle should be a level 5 of automation

Legal Perspective

● Safety needs to be guaranteed ● The vehicle should be sound ● The vehicle should behave

appropriately

● No alterations required regarding existing infrastructure

Logistical Perspective

● Carry as many parcels as possible

● Added value compared to current methods of delivery

● Parcels should be loaded into and unloaded from the vehicle ● Enough space around the vehicle ● The population density should be

sufficient

● Location within a range of 2000 meters from the customer

● Location within a range of 500 meters from the customer ● More advantages than just

replacing a driver

● A location with not too much passing traffic

(31)

Chapter 6 - The design

This chapter describes the design of the autonomous parcel locker, which is basically a self-driving parcel locker. The design complies with the criteria from Table 5, which indicates that the design is feasible. Nevertheless, there will be a validation and benchmark with other methods of delivery in Chapter 7 to evaluate the design.

The design of the autonomous parcel locker will be discussed with the perspectives from Chapter 5: Technological, Legal and Logistical. Within the sections of this chapter, the necessary design criteria will be the basis to prove that the design complies with these criteria. The desirable design criteria will be used to clarify the design even more

Within the supply chain of a carrier, the autonomous parcel locker operates as a movable pickup point, or collection point in terms of Aized & Srai (2014). The vehicle is deployed from a fixed location. This location receives large amounts of parcels from different regions and serves as a hub for multiple neighbourhoods and cities. At this location, there are humans available to load the parcels into the autonomous parcel locker. Subsequently, the vehicle drives autonomously to a fixed location in the vicinity of the customers. This location is a conventional parking spot where the vehicle will park for the rest of the day. As a consequence, the autonomous parcel locker serves as a hub for a specific neighbourhood or small city. Customers can go to the vehicle when they desire, such that the flexibility increases. Furthermore, the customer the consumer unloads his parcel from the vehicle himself. So the physical deliverer is completely replaced by both other people and autonomous technology. The vehicle performs best in areas with a relatively high population density, due to the larger number of parcels that it carries. Therefore, the concept focuses on urban areas rather than the countryside.

Referenties

GERELATEERDE DOCUMENTEN

Table 2.1 presents features of studies discussed above and provides a comparison between these studies and the current study based on the following factors: whether

From this research can be concluded that there are six factors that influencing last mile delivery, these factors are: customer requirements, omni-channel customers

In all tested scenarios, a similar pattern is visible which indicates that the longer the cluster interval is, the shorter the routing distance becomes, apart from the five and ten

The system AR- system can track the locations of parcels, present the location to the driver and reduce the delivery time and thus increasing the efficiency of

Factors found relevant to social micro hubs but are not found in the used literature of other last mile solutions as described in Chapter 2 are: ambiguity, return time,

To answer the research question of this thesis, the benefits of last mile collaboration in a White Label Last Mile delivery setting with use of Urban Consolidation Centres are a

Involving customers in the last-mile may have implications for the level of perceived control that providers like e-retailers and 3PL retain over the last-mile process.

No significant relationship between a customer’s residential area and last-mile delivery option (home delivery versus picking up from an alternative location) has been found for