Identifying Suitable Environments for Autonomous Truck Development A Study on Controlled Outdoor Environments and Their Characteristics Master’s thesis in Management and Economics of Innovation, and Quality and Operations Management LORENZO BJÖRCK MOA GUNNARSSON DEPARTMENT OF TECHNOLOGY MANAGEMENT AND ECONOMICS DIVISION OF ENTREPRENEURSHIP AND STRATEGY CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2023 www.chalmers.se Report No. E2023:036 http://www.chalmers.se 1 REPORT NO. E2023:036 Identifying Suitable Environments for Autonomous Truck Development A Study on Controlled Outdoor Environments and Their Characteristics LORENZO BJÖRCK MOA GUNNARSSON Department of Technology Management and Economics Division of Entrepreneurship and Strategy CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2023 2 Identifying Suitable Environments for Autonomous Truck Development A Study on Controlled Outdoor Environments and Their Characteristics LORENZO BJÖRCK MOA GUNNARSSON © LORENZO BJÖRCK, 2023. © MOA GUNNARSSON, 2023. Report no. E2023:036 Department of Technology Management and Economics Chalmers University of Technology SE-412 96 Gothenburg Sweden Telephone + 46 (0) 31-772 1000 Gothenburg, Sweden 2023 3 Identifying Suitable Environments for Autonomous Truck Development A Study on Controlled Outdoor Environments and Their Characteristics LORENZO BJÖRCK MOA GUNNARSSON Department of Technology Management and Economics Chalmers University of Technology Abstract As the transportation industry is struggling with several structural problems, autonomous trucks have emerged as one potential solution to the issues such as driver shortage, shrinking margins and climate change. As of today, there is low technical maturity in autonomous solutions, and due to this the developers are struggling to find well-suited locations for the technology’s primary implementation. In light of this issue, several authors have turned attention to controlled outdoor environments as a potential arena for deployment of autonomous trucks, suggesting that these sites may be easier to operate than public environments. Despite this interest, there exists no comprehensive compilation of which types sites these potential controlled outdoor environments would consist of, when placing them in the context of autonomous trucks. Moreover, there is a lack of knowledge about which characteristics these controlled outdoor environments exhibit that make them suitable for autonomous trucks. In light of this, this report investigates both which characteristics of controlled outdoor environments that predisposes suitability, as well as identifying which of these controlled outdoor environments exhibit the characteristics that supports the operations of autonomous trucks. During the research, it was found that there were both site specific characteristics and general characteristics. The first category of characteristics were divided into the categories vehicle requirements, size of automation opportunity, traffic environment, operational conditions and site actor attitudes. Subsequently, the general characteristics were divided into weather conditions, regulatory conditions, economic conditions and autonomous loading and unloading. Moreover, the site types assessed by the study were dry ports, manufacturing plants, ports, intermodal terminals, transshipment centers, freight village/logistics clusters and airports. Both the impact of characteristics and the sites were discussed from several dimensions, namely use case, profitability, safety and regulatory conditions. Key characteristics from a deployment perspective was the internal transport volume on a site, its physical size, cargo type and form factor requirement, weather conditions as well as if the site was fenced-off. As for the evaluation of site types, ports stand out from the other sites due to their suitability on all dimensions, although they have specific requirements for vehicle capacity and performance. Next in line in terms of suitability are manufacturing plants, freight villages and airports. Finally, the report concluded that there is large internal variation between individual sites of the same type, and therefore it is important to study each site individually prior to a deployment decision. Key words: controlled outdoor environments, autonomous trucks, deployment factor, technology adoption, site characteristic. 4 Acknowledgements This master’s thesis has been conducted during the spring of 2023 at Chalmers University of Technology as the final project of our master’s program within Management and Economics of Innovation and Quality and Operations Management. The thesis was completed in collaboration with a case company that wishes to remain anonymous. Firstly, we would like to express great appreciation towards our supervisors at the case company. Their support throughout the process has been invaluable, with great input and ideas every week, as well as recommendations of great interview subjects that have been key to the process. Secondly, we would like to thank our university supervisor Kamilla Kohn Rådberg who works at the department of Entrepreneurship and Strategy at Chalmers University of Technology. Her advice on report structure and all-around guidance has enabled us to complete our thesis. Finally, we would like to thank all the interviewees who generously shared their valuable insights and experiences for the thesis. Their contribution has made this study possible. 5 Table of Content Abstract 4 Acknowledgements 5 Terminology 8 List of Tables 10 1. Introduction 11 1.1 Background 11 1.2 Scope, aim and research questions 12 2. Theoretical background 13 2.1 Transport automation 13 2.1.1 SAE - Levels of automation 13 2.1.3 Challenges related to autonomous trucks 14 2.1.4 Benefits of autonomous trucks 15 2.1.2 State-of-the-art and use cases of autonomous transport in indoor logistics operations 16 2.1.2 State-of-the-art and use cases of autonomous transport in outdoor logistics operations 16 2.2 Controlled outdoor environments 18 2.2.1 Logistics centers 19 2.2.2 Manufacturing plants 22 2.3 Deployment factors 22 2.3.1 Use cases 23 2.3.2 Profitability 23 2.3.3 Safety assurance 24 2.3.4 Human factors 24 2.3.5 Regulations 25 2.3.6 Impact on labor 25 2.3.7 Public acceptance and trust 26 3. Methodology 27 3.1 Research strategy 27 3.2 Data collection 27 3.3 Data analysis 30 3.4 Research quality 31 4. Empirical findings 32 4.1 Site specific characteristics 32 4.1.1 Vehicle requirements 32 4.1.2 Size of automation opportunity 34 4.1.3 Traffic environment 35 4.1.4 Operational conditions 36 4.1.5 Site actor attitudes 37 6 4.2 General characteristics 38 4.2.1 Weather conditions 38 4.2.2 Regulatory conditions 38 4.2.3 Economic conditions 39 4.2.4 Autonomous loading and unloading 40 4.3 Site descriptions 41 4.3.1 Dry ports 41 4.3.2 Manufacturing plants 41 4.3.3 Ports 43 4.3.4 Intermodal terminals 44 4.3.5 Transshipment center 45 4.3.6 Freight village/logistics cluster 45 4.3.7 Airports 46 4.4 Summary of characteristics exhibited by site types 47 4.4.1 Definition of the optimal state of site characteristics 47 4.4.2 State of site characteristics at site types 50 5. Discussion 51 5.1 Impact of site characteristics on the use case 51 5.1.1 Characteristics in the flow impacting the use case 52 5.1.2 Characteristics around the flow impacting the use case 53 5.1.3 Characteristics of the site actors ordering the truck operation impacting the use case 53 5.2 Impact of site characteristics on the profitability 54 5.3 Impact of site characteristics on the safety assurance 56 5.4 Impact of site characteristics on the regulatory preconditions 58 5.6 Impact of site characteristics on all key deployment factors 60 5.7 Suitability of controlled outdoor environments in autonomous truck operations 63 6. Conclusion 67 6.1 Future research 69 7. Managerial implications 70 Bibliography 72 A Appendix 77 A.1 Interview guide case company example interview 77 A.2 Interview guide site interview 77 7 Terminology AGV: Automated Guided Vehicle. A robot that is designed to perform transport operations without human intervention in a controlled setting. It follows a predetermined and guided path independently, to fulfill the purpose to which it is dedicated. Autonomous truck: An autonomous truck is a self-driving truck, operating without any human intervention, using advanced technologies such as Lidar, Radar and GNSS to navigate. Controlled outdoor environment: Outdoor private area with clearly defined boundaries. The site owner can sometimes decide who is granted access to the premises, therefore this environment tends to have a low amount of public traffic. Fenced-off: The denotation of a physical area which is enclosed from the surrounding environment by a barrier, more particularly a fence. Form factor: Referring to the physical specification of any type of hardware, including size, shape and arrangement of parts. In the context of autonomous trucks, this can refer to the chassis of the truck. Intermodal: This term refers to the transportation of cargo and goods by transferring it between several modes of transport, commonly used for long-distance transportation. Logistic center: A strategically located area or facility that is part of a supply chain, where all types of transport activities can be conducted, including storage, distribution and processing. ODD: Operational Design Domain. A description of the environment surrounding a vehicle, including but not limited to physical limits, traffic, road dimensions, environmental conditions, as well as functional constraints of the vehicle. Rigid truck: Or box truck, is a vehicle with a cargo compartment that is fixed to the chassis, without standardized dimensions. Mainly used for short-haul regional delivery. Site: A particular location or physical area for individuals or organizations that requires a space to conduct some type of activity. In this report, a site is considered a space where internal transportation occurs. Swap body: A cargo unit of specified dimensions that is designed specifically for transfer between shipping modes. The swap body is light and stackable, with a reinforced steel frame but collapsible sides. The design may differ between regions and transport networks. 8 Trailer: A load-carrier without a motor, that is designed to be towed behind another vehicle, carrying different types of cargo. Tractor truck: A powerful heavy-duty motorized vehicle, primarily used for towing of large load units over long distances. Trailer-agnostic: A term used to describe that different types of trailer units can be towed interchangeably by a vehicle. VAS: Value-Added-Services. Activities in a supply chain that to some extent enhances the state of the goods transported, e.g. labeling, maintenance, repair, and consolidation of goods. VRU: Vulnerable Road-User. Individuals in traffic that are at risk of being harmed by a motor vehicle, e.g. cyclists, pedestrians, or motorcyclists. 9 List of Tables 3.1: Overview of the conducted interviews during the thesis 4.1: Characteristics linked to vehicle requirements 4.2: Characteristics linked to size of automation opportunity 4.3: Characteristics linked to the traffic environment 4.4: Characteristics linked to the operational conditions 4.5: Characteristics linked to site actor attitudes 4.6: Descriptions of site characteristics of and site type 4.7: Outlining the optimal state of site characteristics 4.8: Site characteristics and their observed state at each site type 5.1: Key deployment factors and the characteristics deemed as relevant 5.2: The site type characteristics in the optimal state that are connected to key deployment factors 5.3: Optimal state characteristics divided by total characteristics 10 1. Introduction In this first chapter, the background to the following research study will be outlined, to contextualize the research in a purposeful way. Moreover, the research aim and questions will be introduced, together with the scope of the study. 1.1 Background As of today, the transportation industry is struggling with several structural problems. To start with, there is a substantial threat due to driver shortage, as the bulk of professional truck-drivers are retiring and recruitment is hampered by poor remuneration, lack of possibilities of advancement, and large problems connected to physical- and psychological health (Ji-Hyland & Allen, 2022). Moreover, the pressure on CO2 emission-intensive industries to reduce climate impact is extensive (Berggren & Magnusson, 2012). Lastly, the road transportation industry is struggling with profitability, for instance the road haulers in the UK had profit margins ranging from 1% to 4% from 2008 to 2018 (Statista, 2022). The low margins mean that the industry does not have much room for a higher cost base, since a small increase in costs can push the bottom line down in red territory. Lately, attention has been drawn to autonomous transport, which holds potential to address climate targets, driver shortages and profitability issues simultaneously. Ercan et al. (2022), state that autonomous transport is estimated to give a reduction of approximately 34% on overall transportation industry emissions globally, due to more efficiently planned transport, accelerated adoption of alternative fuels and intelligent charging decisions (Jones & Leibowicz, 2019). Additionally, Short & Murray (2016) proposes that autonomy creates an opportunity to increase the attractiveness of trucking by introducing remote driving stations, which entails improved flexibility and proximity to home. Naturally, this will lead to a decrease in the number of drivers needed, relieving long-term driver shortage (Short & Murray, 2016). Autonomous trucking is also expected to bring increased operational efficiency and reduce labor costs, hence providing relief to the shrinking margins of the industry (Slowik & Sharpe, 2018). Despite these promising prospects, there are several road-blocks for a wide implementation of autonomous trucks. First, Engström et al. (2019) describe six key deployment factors that need to be considered on the road towards large-scale adoption of autonomous trucks. These factors range to include use cases and business models, safety assurance, human factors, regulation, public acceptance and trust as well as the impact on labor. By assessing these key deployment factors in different applications for autonomous trucks, the challenge in finding a suitable implementation opportunity can be addressed in a structured way. Examples of applications provided by Engström et al. (2019) are both long-haul transport on motorways, as well as short-haul implementations off-road. Furthermore, van Meldert & De Boeck (2016) suggest that implementation should first take place in the short-haul, claiming operation in controlled outdoor environments to be especially interesting. So far, very little attention from the research community has been turned towards controlled outdoor environments in the context of autonomous trucks. 11 Up until now, actors in the autonomous truck industry struggle to find deployment opportunities that are available at the developing state of technical maturity, and due to economical constraints these should preferably already exist within current infrastructure (Metz, 2022). van Meldert & de Boeck (2016) as the stepping stone to large-scale deployment, without further elaborating on what characteristics outdoor controlled environments should exhibit in order for them to be suitable for autonomous truck implementations or on which controlled sites they can be found. 1.2 Scope, aim and research questions This research aims to extend the understanding of the opportunity to deploy autonomous trucks in the context of controlled outdoor environments. The scope of the study has been designed for this purpose, meaning that the study disregards all focus on long-haul public road transport of autonomous trucks, along with short-haul transport outside of controlled sites. Additionally, the study is delimited to research on freight transportation, meaning that the study does not address private or public transport. Within the broad concept of controlled outdoor environments, this study has focused on different logistics centers along with manufacturing plants, deemed as the most relevant and feasible technological starting points. Moreover, the controlled outdoor environments are studied in collaboration with a single case company, which has chosen to remain anonymous.With this context as background, the suitability of controlled outdoor environments is examined through the lens of the key deployment factors provided by Engström et al. (2019). In this study a focus is on a subset of these factors, namely use case and business models, safety assurance and regulations. These are chosen based on the interview material, as they are discovered to be of highest relevance to cases of controlled outdoor environments, throughout the interview study. The aim of this study is to fill the previously mentioned research gap on controlled outdoor environments in relation to autonomous trucks. For this thesis, this will include investigating what types of controlled outdoor environments that autonomous trucks could be implemented, at their current level of technological capability. In addition to this, this means to provide an understanding of which characteristics, attributed to a controlled outdoor environment, have an impact on a specific site's suitability for deployment of autonomous trucks. In order to fulfill this purpose, two questions have been formulated to guide the research: ● What are the characteristics of controlled outdoor environments that are connected to the suitability of autonomous trucks? ● Which outdoor controlled environments exhibit characteristics that can support operations of autonomous trucks? 12 2. Theoretical background This section will provide the theoretical background as well as the framework that the study is based upon. The first section describes different aspects of transport automation, starting out with the official definitions of automation levels, and concepts central to the technology. Consecutively, the challenges to development of autonomous trucks are described along with its projected benefits, and finally an overview of the current applications of autonomous transport in logistics is presented. The second part describes the different types of controlled outdoor environments investigated by this study, including a wide range of logistics centers as well as manufacturing plants. Finally, the key deployment factors are described in detail, to create a foundation for the following interview study and discussion. 2.1 Transport automation 2.1.1 SAE - Levels of automation The most foundational concept within autonomy, is the different levels of automation that SAE International (2021) has defined. In order to understand the levels of automation there are a few concepts defined by SAE International (2021) that need to be understood. The dynamic driving task (DDT), is the task of driving the vehicle, the difference between the levels are to which extent the autonomous driving system carries out the DDT. Moreover, the operational design domain (ODD) is of importance, since it describes the operating conditions where the automated driving system is supposed to work. This ranges to include restrictions on weather conditions, time of day, road- or other characteristics that must be present to ensure that the autonomous vehicle can function. The final important concept is the object and event detection and response (OEDR), that is the capability to monitor the environment and respond appropriately to the situations encountered by the vehicle. This can be performed both by a human driver or by an autonomous driving system. Moving into the levels, there are six levels of automation, starting from Level 0 and ending at Level 5. The Level 0 is no driving automation, which means that the entire DDT is handled by a human. At Level 1, driving assistance, the autonomous driving system performs an ODD-specific part of the DDT. This means either lateral or longitudinal, meaning either to turn or accelerate/decelerate, but not the two together simultaneously. Level 2, partial driving automation, is similar, but with the extension that the autonomous driving system can perform both lateral and longitudinal driving. Level 3, conditional driving automation, is a further extension, the autonomous driving system can now perform a wider ODD-specific part of the DDT with the expectation that the driver is ready to engage in driving when the system notifies it to do so. In Level 3, the autonomous driving system also performs the OEDR, meaning that the system recognizes the environment and acts accordingly to its requirements. At Level 4, high driving automation, the autonomous driving system now performs the entire DDT in a 13 specific ODD, and also performs DDT fallback. This means that the human in the driver seat is more of a passenger within the specific ODD. However, the human takes over the DDT if the autonomous vehicle exits the ODD. An example of this is an autonomous vehicle that has the capability to perform the DDT on a highway, but can not enter or exit the highway by itself, instead it needs to have a human driver able to perform those tasks. The last is Level 5, full automation, this means non-ODD-specific performance by an autonomous driving system of the entire DDT and DDT-fallback. This means that the autonomous vehicle can operate wherever a typically skilled human driver could drive the vehicle. 2.1.3 Challenges related to autonomous trucks In this section challenges of autonomous trucks will be described to understand the issue at hand. Firstly, there are technical challenges to autonomous trucks. The whole technical system needs to run smoothly to enable good operations of autonomous vehicles. Milford et al. (2019) mention sensors and interaction with vulnerable road users (VRUs) as major challenges that need to be solved. Sensors are used to identify data on the driving environment and are used by the vehicle to take appropriate action (Kocić et al. (2018). Milford et al. (2019) show that different sensors experience different challenges, an example of this is lidar technology that can provide information about the position of objects far from the vehicle, but whose effectiveness is limited when the weather conditions are adverse. Therefore, a combination of sensors is preferred, but even then challenges such as path planning persist (Kocić et al., 2018). The challenges with VRUs are different, and instead evolves around detecting, recognizing and predicting the actions of VRUs. Milford et al. (2019) describes that machine-learning could help in recognizing the VRUs. Moreover, the autonomous system has to communicate its intent with VRUs (Milford et al., 2019). The authors mention the prediction VRU action to be the most important of these challenges, as well as one of the most complex tasks (Amini et al, 2021). Humans base their prediction of VRUs actions on a guess of the goal of a specific VRUs in traffic, and autonomous systems should try to mimic this ability (Milford et al., 2019). Moreover, there are technical challenges at higher levels of autonomy. Anderson et al. (2018) write that Level 3- systems that require the human to be in-the-loop and be ready to handle the DDT-fallback are dangerous. This is due to the fact that the human needs to be awake and aware to take appropriate action at any time. Furthermore, Slowik and Sharpe (2018) write that a higher level of automation, Level 4 or more, is at risk of being hacked, meaning that there would be a safety concern rather than a safety benefit. Moreover, there are other challenges to autonomous vehicles, which are mainly social challenges and business model challenges. Slowik & Sharpe (2018) present two of the social challenges linked to the adoption of autonomous vehicles. That is that social acceptance may be a problem and that truckers may be losing their jobs. The social acceptance mainly regards safety and system reliability concerns, there is a need for a proof of concept of autonomous driving to make it widely accepted. The loss of jobs of truckers is quite substantial. According to Slowik & Sharpe (2018) there are 3.5 million truck drivers in the US only and the potential loss of these jobs are a high barrier to adoption of autonomous trucks. Talebian & Mishra (2022) further speak about this challenge. The potential job loss of truckers can incite labor unions to act against the legislation of autonomous trucks, putting up a potential regulatory barrier. 14 Engström et al. (2019) write that there are business model challenges to autonomous trucks. The authors note that in general the trucking industry has low margins, meaning that there is a need for fast economic return on investment for the industry as a whole. However, this could also mean that profitable and safe autonomous trucks would pave the way for a fast and wide adoption of the technology. 2.1.4 Benefits of autonomous trucks Autonomous trucks have several benefits, as described by Engström et al. (2018), and these benefits vary based on where autonomous trucks are adopted. The main categories of benefits seen in the literature are cost benefits, safety benefits and productivity benefits (Slowik & Sharpe, 2018; Andersson & Ivehammar, 2013; Hoque et al., 2021; Khan et al., 2022). The cost benefits in autonomous trucks are based on two main levers. The first being the decreased costs stemming from saving on eliminating driver salary. Slowik & Sharpe (2018) describe that 35% of the marginal cost per driven mile are driver costs and Andersson & Ivehammar (2013) also conclude that driver costs are a significant part of the total cost for trucks. This means that a decrease or an elimination of these costs would therefore be quite beneficial for the trucking companies. The second lever are the cost savings stemming from increased fuel efficiency (Slowik & Sharpe, 2018). The fuel costs are also a large part of the total costs of long-haul trucking, according to Sharpe (2017) the fuel costs can be between 25% and 40% of operational trucking costs. Thus reducing the fuel cost would be very beneficial. The fuel efficiency gains can either be realized through truck platooning (Hoque et al. (2021) or by being electric (Litman, 2017). The next category are the safety benefits. The benefit or risk of these systems vary with the level of automation. Slowik & Sharpe (2018) describe that already on low levels of automation, Level 0 or 1, the collision mitigation systems reduce accidents by 87%. Moreover, Anderson et al. (2018) write that systems up to Level 3 have many safety benefits such as lane assist and safety braking which help prevent fatal accidents. The final category is productivity benefits. Litman (2017) describes one form of productivity benefits, increased driver productivity, that is that the driver can perform other tasks while driving. This means that the driver will be able to work on other tasks relevant to the trucking company whilst driving. Moreover, Engström et al. (2019) tells us that there are productivity gains from increased hours of truck operation. Moreover, the authors mean that increased productivity by autonomous truck operations can stem from more flexible working hours for long-haul trucking. Furthermore, Slowik & Sharpe (2018) mean that autonomous trucks and specifically truck platooning give less road congestion. This means that less time will be spent stuck in traffic for the traffic system as a whole, which Harriet et al. (2013) view as a productivity gain. 15 2.1.2 State-of-the-art and use cases of autonomous transport in indoor logistics operations Having defined the levels of autonomy, the challenges and benefits of autonomous vehicles, the next step is to delve into what the state-of-the-art of autonomous vehicles in logistics operations is today. There are several potential use-cases for these. In logistics, there are two main applications of autonomous- or automated vehicles, namely indoor- and outdoor logistics operations (DHL, 2014). This section will cover the case of indoor logistics operations. According to van Meldert & De Boeck (2016), mainly split into production plants, cross-docking stations, warehouses and distribution centers. In this setting typically automated guided vehicles (AGVs) are used. AGVs are essentially automated robots that transport goods of some kind from one location to the other (De Ryck et al. (2020). van Meldert & De Boeck (2016) outline that AGVs can both be used for horizontal and vertical transportation of goods, and coupled with autonomous loading and unloading the whole logistics operation can be automated. The authors describe autonomous forklifts as a common use-case for vertical goods transportation. Ullrich (2015) further provides examples of horizontal transportation of goods where AGVs can be used as an alternative to fixed assembly lines, automating warehouse operations or handling specific equipment. 2.1.2 State-of-the-art and use cases of autonomous transport in outdoor logistics operations The next use-case for autonomous vehicles are in outdoor environment operations. The use-cases here are long-haul autonomous vehicle operation, last mile delivery and operation in outdoor controlled environments (DHL, 2014). The long-haul autonomous vehicles in logistics operations are divided into truck platooning and exit-to-exit automation. Truck platooning is a use case mentioned in the literature for adoption of autonomous trucks (Engström et al., 2019). The definition of this is according to Janssen et al. (2015) two or more trucks following each other at a close distance, since they are digitally connected to each other there is no need for a driver in the trucks that are following the first one. Autonomous truck platoons are thought to save energy due to the lower drag experienced by the following trucks (Tsugawa et al, 2011) and be more stable (Kim et al., 2022). However, truck platooning has been researched for several decades (Heikoop et al. (2017), and is still in testing phase (Yang et al., 2022). Slowik & Sharpe (2018) describe that more research needs to be done in order to understand which trucks to use for platooning, how much they should weigh and how many trucks should be in a platoon. Slowik & Sharpe (2018) further conclude that the subject needs to be further investigated in order to understand what impact platooning could have on truck fleets, in order for a business case to exist for the use of the technology. Another use case for autonomous trucks is operations in exit-to-exit highway automation (Engström et al., 2019). According to the authors, this is a use-case where autonomous trucks operate on highways, which 16 could be done with a range of different autonomous systems. Firstly, Slowik & Sharpe (2018) describe Level 1-systems in use and Level 2-systems approaching commercial readiness that have collision avoidance and driver warning systems. Furthermore, Engström et al. (2019) describe systems with Level 1 and Level 2 capabilities such as lane assisting capabilities and adaptive cruise control are in use today. Moreover, Slowik & Sharpe (2018) describe that Level 3 trucks by Uber ATG and Otto are being developed and tested in long-haul trucks, however they are not in a commercialization stage. Another use-case for autonomous vehicles in logistics is described by van Meldert & De Boeck (2016) as last-mile delivery systems. Last mile delivery is defined by Boysen et al. (2021) as encompassing all parts of the logistics system relating to urban private customer delivery. According to DHL (2014) this is the area that is most unpredictable for autonomous vehicles due to the vast amount of actors in the ODD, e.g. cyclists, pedestrians, trucks. Moreover, van Meldert & De Boeck (2016) point out that the automated driving system needs to be at Level 4, meaning more advanced than any system in current use. Examples of hypothetical use-cases described range from autonomous grocery shopping and autonomous parcels (van Meldert & De Boeck, 2016) to even autonomous drones (Brunner et al., 2019). The final use-case for autonomous trucks is described by Engström et al. (2019) as off-road trucking, and by van Meldert & De Boeck (2016) referred to as operations in controlled outdoor environments, which is also the main focus of this research report. According to Engström et al. (2019) and Meldert & De Boeck (2016) these enclosed areas are appropriate for the primary implementation of autonomous trucking, in advance to large-scale on-the-road adoption. Shah & Piragine (2018) describe that the first step in taking autonomous vehicles from “the security of controlled settings into the uncertain world of everyday traffic” is to implement them in controlled outdoor environments and then to implement them in long-haul transport and last-mile delivery. van Meldert & De Boeck, (2016) describes the reason for this being that these sites exhibit fewer regulations, less uncertainty and have a less complex liability issue. The authors also emphasize the improved security of the controlled environment as a strong argument for the fenced-off use case, where autonomous vehicles can be operated in a low-speed and simple route environment where disturbing elements are minimized. Engström et al. (2019) describe mine hauling, ports, yards and terminals as such potential areas for autonomous trucks. The area of operation is under investigation by several companies, both with and without driver cab. An area where this is already in use is the mine hauling (Brundrett, 2014). Moreover, Brundrett (2014) writes that the operation of autonomous trucks started back in 2008 when Rio Tinto and Komatsu collaborated to create a mining haulage truck operation, and that the commercialization of autonomous trucks is happening faster than autonomous public automobiles. In the segment of controlled outdoor environments there is also a variety of AGV:s (van Meldert & De Boeck, 2016; Ullrich, 2015). AGV in port operations are described in many cases, such as Kim & Bae (2004), Vis et al. (2001) and Ioannou et al. (2000). Ioannou et al. (2000) describes that AGV:s are used in different container operations, mostly in the terminal operations. Vis et al. (2001) further describe that usage could be that the AGV is implemented between quay crane and the straddle carrier. That is, the goods are unloaded from the boat, lifted onto an AGV and then transported to the straddle carrier that then 17 stacks the container. Moreover, Ullrich (2015) describes outdoor warehouse areas as a use-case for outdoor AGV:s, where they instead of handling containers handle palletized goods as the AGV:s are automated forklifts. The author further describes automated heavy load transporters in outdoor manufacturing plants as a use-case. As seen in this section there is a wide variety of potential applications for autonomous trucks and AGV:s in outdoor controlled environments. Engström et a. (2019), van Meldert & De Boeck, (2016) and Shah & Piragine (2018) describe examples of outdoor controlled environments, but the authors do not investigate the specific characteristics of different types of individual sites in any further depth. The next section will describe the general knowledge of the facility types that is the subject of this report. 2.2 Controlled outdoor environments As previously described by Engström et al. (2019), van Meldert & De Boeck (2016) and Shah & Piragine (2018), controlled outdoor environments constitute a promising use case for autonomous trucks. When talking about enclosed environments where the technology can be initially implemented, van Meldert & de Boeck (2016) exemplifies the concept with ports, manufacturing plants, airports and other different types of logistics centers and courtyards. However, the concept of logistics centers is broad, and classifications of different logistics centers are concerned with a high degree of confusion and conceptual ambiguity (Notteboom et al., 2016). Despite how a vast amount of typologies has been suggested to bring clarity to the field, no consensus has been reached (Meiduté, 2005). The two following sections aim to provide basic knowledge and contextualize different controlled outdoor environments in a structured way. Due to the broad ranging concept of logistics centers (Notteboom et al., 2016), most sites of interest in a transportation context can be accommodated under its roof, and in addition to this, manufacturing plants will be outlined as a separate section. 2.2.1 Logistics centers According to the UN connected logistics association Europlatforms EEIG (2004), a logistic center can be defined as a “ a clearly defined area within which all activities relating to transport, logistics and the distribution of goods - both for national and international transit - are carried out by various operators on a commercial basis”. As previously mentioned, this include a wide range of facilities, and this section will be organized according to a typology created by Higgins et al. (2012), where the authors have combined the considerations of several authors in the research field to form a novel typology based on size, influence, function in freight- and logistics processes as well as value adding activities. The main types of logistics centers have then been placed in a hierarchy accordingly, see Figure 2.1 (Higgins et al., 2012). The typology was selected since it distinguishes itself from other work by being based on a comprehensive theoretical framework of existing typologies, and also sheds light on the incoherences of the research field, for instance how different authors define facilities differently and views them as having different positions in the logistics hierarchy. 18 Figure 2.1: Standardized logistics center hierarchy (Higgins et al., 2012). Figure 2.2: Freight terminal types (Wiegmans et al., 1999) This typology is based on the five facility classes as presented by Wiegmans et al. (1999) in Figure 2.2, where the characteristics of geographical coverage, volume and capacity distinguishes them into five orders of magnitude. The focus on facility size is pervading the literature as a distinguishing feature when trying to structure the dispersed area of logistics facilities, which for instance is present in later work by Onstein et al. (2021). This paper is based on logistics facilities literature and a large dataset on logistics 19 facilities in the Netherlands, and it is in the same way investigating the relationship between size and site characteristics, such as activity type, product type, network structure, and market service area (Onstein, 2021). A similar approach and division criteria is used by Notteboom et al. (2016), who also classifies based on size, position in transport and commodity chains, and geographic market coverage etc. However, instead of concluding a hierarchy Notteboom et al. (2016) divides facilities into three primary functions; i) storage, deposit & warehousing, ii) cargo transloading & rapid transit, iii) value added services & light manufacturing. Comparing this typology is interesting since it sheds light on the functionalities of the modern logistics systems, while strengthening the criteria on which the Higgins et al. (2012) and Onstein et al. (2021) models are based. In the following section, an extended review of the Higgins et al. (2012) categorisation will be made, to outline the activities and attributes the most frequently occurring logistics facilities are referred to in a relatively unified way. The first level facilities, with the size classification S (Wiegmans et al., 1999), are completing the most simple tasks in the transport network. It serves as a logistic backbone due to fulfillment of basic logistic functions such as storage and serving as a general support in goods movement and transloading. The main facility types are 1) warehouses, generally being single facilities acting as a buffer in the supplier - customer relations, providing space for temporary storage and inventory. 2) distribution centers are one or several facilities with the main purpose of smoothening and bridging the flow of goods, combining warehousing, shipping, goods consolidation, cross-docking and transloading. 3) containers yards & inland container depots also counts as first level facilities, and they are dedicated to container storage and maintenance, and container movement and container goods modification respectively. The second level facilities, with size classifications ranging from M-L (Wiegman et al., 1999), encompasses a plethora of activities including extensive intermodal goods transfer, a large geographical market service area and a complete offering of value-added services. Higgins et al. (2012) elevates three types of second level facilities, firstly the 1) intermodal terminal, an important construct, which varies in size and range of activities. The general purpose is to handle extensive freight flows, and manage the transshipment of goods between transportation modes, e.g. rail, road and maritime. The intermodal terminal can act as a consolidating connection point between regions and continents, and also provide value-adding services to the passing goods. The 2) inland port is an extension of a port, often with a close infrastructural link to the mentioned that enables it to improve the mainport capacity in storing, storage and logistics management. Through this extension, the maritime freight flows can be consolidated and transshipped, and incoming goods can also be deconsolidated for local market distribution. Additionally, inland ports generally contain the extensive container handling of container yards and depots, and manage all customs-related activities. According to the typology by Higgins et al. (2012) the 3) freight village is the largest inland logistics facility in terms of both physical size and transport network impact. Typically, the freight village is characterized by its provision of shared logistics supporting functions, including administrative- and commercial support, maintenance areas, and worker amenities. Moreover, the freight 20 village has strong intermodal connections with road, rail, air and barge infrastructure. Overall, the site is dedicated to the support of efficient goods flows and facilitated supply change management. The third level facilities address the mainport terminals of the logistics system, generally classified as XL and XXL logistics centers. Mainport terminals connect continental inland transport networks, acting as an interface between national and international supply chains. The terminals are generally handling large maritime freight flows, hence creating a lot of economic and logistic activity in its surrounding hinterland, by inducing large good transport flows to the inland distribution chain. The mainport exchanges immense volumes of goods and passengers deconsolidate the flows onto intermodal transportation modes, and consolidates outgoing freight flows for shipping. In order to do this, the mainports contain all previously mentioned logistic activities, being transshipment, storage, maintenance, value-added services, customs, administration and workforce amenities. Due to their transport network impact, the mainport terminal has been described with many names with common examples being gateways (Notteboom & Rodrigue, 2009) and logistics nodes (Rimienè and Grundey, 2007), however Notteboom & Rodrigue (2009) include both airports and ports in this category. According to Weigmans et al. (1999), logistics facilities of this size are characterized by high volumes, high-capacity utilization, and large international companies on site. 2.2.2 Manufacturing plants While logistics facilities can be difficult to classify due to conceptual ambiguity and lack of consistency in nomenclature (Notteboom & Rodrigue, 2016), manufacturing plants are an even more varied group of controlled outdoor environments, with individual sets of manufacturing facilities and strategic goals (Vokurka & Davis, 2004). A general characteristic of manufacturing plants is their complex organization, where coordination and planning of all operations must be efficient to reach the plant overarching performance objectives (Sule, 2008). The author mentions several activities of the plants that are in need of coordination both during design and production, these ranges to include the utilized manufacturing processes, plant layout, material-handling, storage systems, and the unified operational cost estimate of the plant. According to Vokurka & Davis (2004), manufacturing plants are often classified based on the product they produce, which entail describing attributes such as volume, variety, complexity. They can also be classified based on production process, with attributes such as complexity and flow, or the market that they serve, which can differ regarding scope, need, diversity (Vokurka & Davis, 2004). Moreover, the authors emphasize the need for specificity in the configuration of manufacturing facilities and the assignment of facility functions, based on the production goals and the market service area. 2.3 Deployment factors As discussed in section 2.1, there exist several potential applications of automated trucking technology both on- and off-road, and they share challenges that in the first case may be easier to surmount in controlled environments. As stated by Engström et al. (2019), the target scenario for the development of autonomous trucks is to be deployed on large-scale on public roads while inducing revenue streams to the 21 solution developers and the customers. Engström et al. (2019) present six deployment factors that can be used to assess the potential for large-scale deployment of autonomous trucking technology in a specific arena, namely; use cases, business models, safety assurance, human factors, regulation, impact on labor, public acceptance and trust. In this paper, the different sites and their characteristics will be viewed from these deployment perspectives, when assessing the suitability of sites and the importance of certain characterizing elements in a controlled environment. The primary focus will be on the factors discussing use cases, business models, safety assurance and regulations, as the emphasis on human interaction as well as the attitude of the general public is less important in enclosed and private environments. Subsequently, the deployment factors of focus for the study will be reviewed. 2.3.1 Use cases Engström et al. (2019) write that “the potential benefits of automated trucking depend strongly on the specific use case considered”, meaning that the primary precondition for successfully implementing autonomous trucks is that the transport operation being automated are well-suited for the new technology. The authors place emphasis on the heterogeneity of trucking operations, and explain that the initial use cases of autonomous vehicles will be in environments of low complexity (Engström et al., 2019). van Meldert and de Boeck (2016) claim that the likely development is that autonomous vehicles will be allowed under extremely specific conditions; at one particular speed, specified weather conditions and certain routes. Fagnant & Kockelman (2015) states that the benefits in most use cases depend on the improvement of automated driving capabilities, and agrees with Canis (2019) who mention the potential of infrastructure adaptation as a way to construct a more favorable public landscape that could contribute to increase the number use cases for autonomous vehicles that are within reach in the near future. Engström et al. (2019) claims that the type of use cases that will be considered for the current autonomous technology could be operating large private fleets on fixed and predictable routes. Moreover, the same authors conclude that deployment of autonomous technology in its current state relies on developers to identify highly suitable venues with low-complexity use cases where autonomous solutions can be customized to solve individual carrier needs. 2.3.2 Profitability Automated trucking differs from automated private vehicles in terms of the motives for the investment, due to the fact that trucking companies are driven by the promise of expanded profit margins, in contrast to private vehicle owners who also prioritize comfort, status and safety (Engström et al., 2019). According to Fagnant & Kockelman (2015), widespread technology adoption is lagging behind to a large degree due to the extremely high investment barriers for both the trucking industry and private customers, even basic autonomous solutions being excessively expensive. The typical small margins of the trucking industry places a demand for rapid, and guaranteed, return on investment (Engström et al., 2019). Predictions are that when sufficient technological progress is accomplished, the operational efficiency improvement and cost reductions in freight will disrupt the trucking industry and adoption will accelerate (Sing Muddhar et 22 al., 2016). According to the authors, the potential economic benefits consist of efficiency improvements in around-the-clock operations, efficient fuel-consumption, as well as cost reductions when no driver is paid. Engström et al. (2019) state that it is essential to present a convincingly strong business case to customers, where sufficient value can be translated to the bottom-line, to incentivize commercial deployment of autonomous trucking in the long-term. In the proposed list of deployment factors proposed by Engström et al. (2019), this factor is denoted as “business models”, however the authors of this study har chosen to replace this with the word “profitability”, as this suits the scope of the research better. 2.3.3 Safety assurance In their paper, Engström et al. (2019) explains this deployment factor as the issue of how one can ever be able to completely ensure that an autonomous system can respond appropriately to all edge cases it will come across in a public traffic environment. According to Wang et al. (2020), the safety of autonomous vehicles is fragile as it depends on technical and social parameters, including automation level, traffic conditions, weather conditions, regulations, and vehicle capabilities. Sing Muddhar et al., (2016) adds to this by lifting the problem of vehicle-human and vehicle-vehicle interaction when the driver is removed from the system, which creates additional complexities when trying to make autonomous vehicles become a natural part of the public traffic environment. Similarly to the case of profitability, the expectations in terms of traffic safety when considering a fully functionable autonomous system are high-leveled, with examples such as reduction of car-crashes, speeding, in-attention, and decreased risk-taking (Sing Muddhar et al., 2016; van Meldert & De Boeck, 2016). Fagnant & Kockelman (2015) brings forward the complexity related to the mix of public traffic environments, where the requirements on object recognition and adequate responses in unpredictable traffic situations or “edge cases”, comprise a major safety challenge. As of today, safety is among the most critical issues, and for the future of autonomous trucking, and vehicle manufacturers, tech companies and suppliers need to collaborate in this matter to accumulate a sufficient amount of safety evidence through simulation, test tracks and eventually on-road field tests (Engström et al., 2019). Sing Muddhar et al. (2016) agrees with this, by stating that 100 millions of miles need to be driven in order to demonstrate rigorous road safety. 2.3.4 Human factors This deployment factor is mainly concerned with how the interaction between humans and autonomous vehicles can be handled, together with the changes the technology brings about in the organizations and lives of humans (Engström et al., 2019). At the core of this, Sing Muddhar (2016) points out the problem that machines are inherently different from the human brain, and therefore this creates problems in the interaction between humans and autonomous vehicles (van Meldert & de Boeck, 2016). Both Engström et al. (2019) and van Meldert & de Boeck (2016) describe research on the operator-vehicle interaction in pre-Level 5 automation, where current challenges are related to negative behaviors when humans become 23 acclimated to the autonomous system. Examples brought forward by the authors range to include overreliance on the autonomous system, technological misunderstandings, and excessive distraction from the driving tasks, and also problems in the vehicle-operator driving handover. Education and training in autonomous technology can to some extent be used to mitigate these issues (Engström et al., 2019). From a more positive point of view, Sing Muddhar et al., (2016) state that autonomous technology at its mature state can actually be used to remove and mitigate human “error” from public traffic. 2.3.5 Regulations In order to deploy autonomous vehicles over time on both a short and a long perspective, the legislative structure must be compliant with technological evolution (Carp, 2018). At the core of the current inertia in legislative change lies the fact that regulations presuppose a driver in the vehicle, which in itself constitute a large legal barrier when creating new vehicle designs and solutions (Engström et al., 2019). In the meantime, Carp (2018) points out, autonomous vehicles must comply with existing traffic- and vehicle regulations for manual vehicles. According to Canis (2019), autonomous vehicles will at some point require completely new legal standards, and Carp (2018) states that the development is handled by offering specific exemptions, meaning that current R&D is run on a case-by-case basis. Collingwood (2017) raises two other regulatory issues that have incurred attention in this field of innovation, namely, the liability and the privacy questions. In short, the liability is concerned with responsibility assignment in autonomous incidents, and privacy revolves around the use and control over all private data processed in the vehicle (Collingwood, 2017). Engström et al. (2019) places a demand for a data-driven approach among policy makers, Carp (2018) requests legislative efficiency for the autonomous vehicle question, and Collingwood (2017) concludes that legal issues may hamper technology development and delay commercialization. 2.3.6 Impact on labor According to Engström et al. (2019), the trucking industry is aware of the risks associated with autonomous disruption. A majority of long-haul transports might be carried out without a driver, and loading- and unloading carried out by operators in the origin- and destination points (Fagnant & Kockelman, 2015). There are also speculations of further spill-over effects in connected branches, such as driver licensing, traffic policing and insurance sales (Faisal et al., 2019). Due to this potential for immense unemployment throughout the industry, the same authors claim that this could infer conflicts between labor groups in the freight railroad industry and the autonomous car manufacturers. However, in light of the present problems of escalating driver shortage, and remaining technical barriers to on-road public automation, the worries of large-scale unemployment have been dampened (Engström et al., 2019). There may also be positive effects of autonomous trucks in the industry, as these have potential to improve the driver working conditions, improve attrition rates, and create new driving tasks (Engström et al., 2019). 24 2.3.7 Public acceptance and trust A central challenge to large-scale deployment of autonomous vehicles, which is particularly critical for large trucks operating in a conventional traffic environment, is public acceptance (Engström et al., 2019). According to Hőgye-Nagy et al. (2023), the attitudes towards autonomous vehicles plays a greater role in technology acceptance than concrete experiences. Therefore, the introduction of autonomous vehicles in the society should be accompanied by education and public communication in the benefit and safety question around the phenomena (Engström et al., 2019). Perceived value and trust lies at the heart of bringing about technology acceptance on a broader scale, and adoption is fully reliant upon the value creation (Yuen et al., 2020). Finally, Collingwood (2017) specifically points out the issue of the user privacy issue, which as of today is central to establishing trust. 25 3. Methodology In this chapter, the methodology of the research study is described. First, the research approach will be discussed, followed by the data collection, data analysis and finally, the research quality. 3.1 Research strategy The research strategy can, according to Bell et al. (2019) follow two very different paths, either being qualitative or quantitative. In this study the chosen method is the qualitative one. The reason for conducting the study in a qualitative way, is due to how qualitative study is able to convey vivid descriptions of complex issues (Sofaer, 1999). There are a multitude of ways of doing qualitative research, examples include but are not limited to longitudinal design, cross-sectional design, comparative design and case study design (Bell et al. (2019). In this research the study will be done by performing a case study. Performing research designed as a case study, means to extensively analyze a single object of analysis in a particular case. In this research the case study will be of a single organization. The chosen organization is well-suited to the purpose of the study, due to the high availability of data in the company and the fact that the company is a pioneer in the segment of study. However, this case company has chosen to remain anonymous, and for this reason the interview data is anonymized. The case study will provide insight into the industry of autonomous trucks as a whole, but through the lens of this single object of study. The research will be conducted in an inductive manner, that is letting theory emerge from the research. The rationale behind doing the research inductively, is simply due to the fact that the specific business area of autonomous trucks in controlled outdoor environments is quite complex. 3.2 Data collection In the following section the collected data will be described. The data collected will be of both primary and secondary nature, where the primary data collected will be interview data from relevant employees at the case company. The interview data, that is primary in its nature, was sampled using three concepts. First, the data was sampled in a purposive way using purposive sampling, that is, according to Bell et. al (2019) that the sample is selected carefully and that the research questions guide which participants that are chosen for the study. Second, a subgenre of purposive sampling was used, that is theoretical sampling. According (Glaser & Strauss, 2017) this entails gathering data to form categories and concepts that together form a theory. The study follows this inductive theory-emerging approach as mentioned above in order to create a wider understanding of this new market and the specific requirements that the new technology has to adhere to in the current state-of-the-art. Third, participants were sampled through snowball sampling 26 which according to Bell et al (2019) is when a small group of people are initially approached and interviewed and the next round of interviewees are sampled by asking the first group. The interviewed subjects were from several categories, both from the case company and other interesting parties. The non-case company specific sampled participants were academics from different backgrounds, including research institutes, university researchers and research consultants, as well as freight transport experts and logistics facility leads. Moreover, the case company interviewees had different roles within the company. The interviewed participants had product responsibilities, e.g. product strategy and product managers, team leader or director roles, e.g. team leader of technical operations, director of regulatory affairs, senior technical director or more business roles, e.g. business developer, and finally a company executive, senior vice president autonomous freight. Initially, a pre-interview round was done internally at the company to understand the problem at hand and what information was in need of gathering. After this round, the interviews were conducted both internally at the company to get a wider view of the case, as well as externally to get a better understanding of the industry as a whole. The sampled participants were interviewed in a qualitative manner in semi-structured interviews. The semi-structured interviews are according to Bell et al. (2019) interviews where the interviewer has prepared questions but has some freedom to ask follow-up questions or ask them in a different order. These were conducted using an interview guide, recorded whenever possible and the recording was supported by taking notes during the interview. In total, 21 interviews were conducted during the study. The interviewed participants are presented in the table below. Interviewee type Interview ID Professional title Interview date Length of interview Research Institute Research Institute RI1 Research Director, PhD in Logistics 6/3 - 2023 60m Research Institute RI2 Senior Researcher R&D Policy 9/3 - 2023 60m Research Institute RI3 Senior Researcher Humanized Autonomy 9/3 - 2023 60m University Researcher University Researcher UR1 Associate Professor in Service Management and Logistics 15/3 - 2023 45m 27 University Researcher UR2 Associate Professor in Service Management and Logistics 15/3 - 2023 45m University Researcher UR3 Professor of Maritime Transport Economics and Logistics 16/3 - 2023 45m Research Consultant Research Consultant RC1 Research Manager at Consultancy Firm 22/3 - 2023 45m Freight Transport Expert Transport Analysis FTE1 Qualified Investigator 1/3 - 2023 30m Transport Analysis FTE2 Statistician in Traffic & Statistician in Goods Transportation 13/3 - 2023 30m Swedish Transport Administration FTE3 Investigator of Freight Transportation 15/3 - 2023 45m Logistics Facility Lead Freight Village Management LFL1 Managing Director 14/3 - 2023 45m Port Management LFL2 Port Development Strategist 27/3 - 2023 45m Port Management LFL3 Strategic Development and Innovation 4/4 - 2023 45m Case Company Case company CC1 Product Strategy 30/1 - 2023 30m Case company CC2 Senior Product Manager Autonomous Vehicles 31/1 -2023 30m Case company CC3 Business Developer 7/2 - 2023 30m Case company CC4 Product Manager in Autonomous Freight 7/2 - 2023 30m Case company CC5 Team Leader of Technical Operations 9/2 - 2023 30m 28 Case company CC6 Director of Regulatory Affairs 26/2 - 20223 + 17/3 - 2023 30m + 60m Case company CC7 Senior Technical Director 8/3 - 2023 30m Case company CC8 Senior Vice President Autonomous Freight 21/3 - 2023 45m Table 3.1: Overview of the conducted interviews during the thesis 29 3.3 Data analysis The case study is analyzed through a qualitative data analysis method. The chosen method for the study is thematic analysis (Bell et al., 2019). Thematic analysis is according to Maguire & Delahunt (2017) the “process of identifying patterns or themes in the qualitative data” and is described more thoroughly by Braun & Clarke (2006). They describe a six-step process delineating all the important steps in doing this type of data analysis. The six-step process is described in this section and in each step of the process there will be a description of how the process is implemented in this research. Braun & Clarke (2006) describe the first step in the process is the familiarization of oneself with the data. Thesis means that the interviews that have been conducted are read through in several iterations to understand the data in a deeper way. In this study, the interviews were recorded whenever possible, therefore in the first step the recordings were watched and the notes from the interview were further developed. Braun & Clarke (2006) further describe the second step in the process, that is to generate the initial codes, this is done through noting interesting snippets of data and collecting them in a systematic fashion. In this report, this was done by making comments in the google docs document that was used for each interview. In the third step Braun & Clarke (2006) describe that codes are being gathered to form themes together. This was done by collecting all comments in the google docs document in an excel sheet and were sorted according to the initial theme they were thought to belong to. A weakness of thematic analysis is pointed out by Bryman & Burgess (1994), this is that it is difficult to assess the exact reason for themes being created. Moreover, the authors suggest using frequency of mentions as a criteria for forming the categories to assess this weakness. This is why in this study, frequency tables were used to create the themes. Braun & Clarke (2006) describe that the fourth step is to review the themes, meaning to refine the generated themes. In this report, this was done by iterating all the themes once more. The fifth step is according to the author to define and name the themes. In this report, themes are the overarching characteristics of outdoor controlled environments, that is vehicle requirements, size of automation opportunity, operational conditions and site actors. The sixth step is to produce the report, and the result of this will be found in the empirical findings of this report. 3.4 Research quality In this section, the considerations regarding the research quality will be described. Bell et al. (2019) suggest that reliability and validity are important quality criteria for the qualitative researcher to analyze. Since the study is of a qualitative nature, these two criteria will be the ones that are described. Firstly, the validity concerns according to Bell et al. (2019) the integrity of the conclusions from research. Golafshani (2003) explains validity as both the accuracy of measurement and whether the measurement does evaluate what is intended. Bell et al (2019) further describe that there are four types of validity; measurement validity, internal validity, external validity and ecological validity. Furthermore, the authors 30 describe that measurement validity is not relevant for qualitative research and that internal and external validity are the most relevant ones in this research context. Therefore this paper will only discuss internal and external validity. Bell et al (2019 writes that internal validity is concerned with the match between the observations made in the study and the results from it. This means, if the study has a causal relationship between the observations and the following results. The interviews in this study have been conducted independently and many different interviews have been conducted, making sure to have a wide variety of information to base the results on. The external validity, according to Bell et al. (2019), are concerned with the generalizability of the study, in practice this means if it is possible to apply the results in other contexts than the specific research one. In order to ensure external validity in this report, the sampled participants are not only company specific ones, but are researchers on the subject and external experts, meaning that the perspective should be one from the general industry and not only from the company. Secondly, reliability is according to Bell et al. (2019) concerned with the repeatability of the research study, however it is hard to meet this quality criteria in qualitative research due to the nature of the research. It is impossible to perform another experiment on a specific social setting in the exact same circumstances due to time passing. This is consistent with the view by Golafshani (2003). However, Bell et al. (2019 provide a suggestion internal reliability in qualitative research that is agreement between the researchers in a multi-researcher setting. Golafshani (2003) argues that a good method to establish reliability is to use triangulation, that is to confirm the research findings using several independent sources. In this study, this has been handled by conducting multiple interviews, and also by doing a theoretical framework on the subject matter. 31 4. Empirical findings The following chapter is based on the interview study that included 21 interviews with both case company employees and external parties. The interview material was coded and consecutively divided into three categories. The thesis findings are synthesized into the sections Characterics of sites and Site descriptions. Hence, the first section presents the characteristics of controlled outdoor environments, deemed important from a deployment perspective. These characteristics are divided in site specific characteristics and general characteristics, depending if they are connected to a specific geographical location or not. In the second part, the data gathered on different sites are presented for each type, and the section is concluded with a table summarizing the characteristics of the site types. 4.1 Site specific characteristics All characteristics of importance to site selection will be discussed in this section, and in the review, they have been divided into five areas with regard to their degree of interrelation. The five areas of characteristics are vehicle requirements, size of automation opportunity, traffic environment, operational conditions, and site actors. Throughout the interviews, all interviewees of relevant competence were asked about their perception of suitable preconditions for implementation of autonomous vehicles, given the current state of technological capability. In the 21 interviews, several recurring themes were discovered. 4.1.1 Vehicle requirements Vehicle requirements Site characteristic Times mentioned Form factor 9 Cargo type 5 Payload 3 Table 4.1: Characteristics linked to vehicle requirements In the first category of attributes, namely vehicle requirements, the form factor of the autonomous vehicle was brought forward as a key consideration when assessing a site and developing a truck platform (FTE3, LFL1, LFL2, LFL3, CC2, CC3, CC7, CC8, RC1, UR3), and the discussion stems from the trend in autonomous freight to focus on rigid body trucks. The reason for the importance of the form factor, that is the hardware design of the autonomous truck, depended on the perspective of the consulted interviewee. 32 Inside the case company, the Product Manager at Autonomous Freight (CC4), with a bridging role between market and technology development, explains the trade-off related to developing a form factor that works in all industrial settings, while reaching a sufficient robustness in the autonomous operations. On the commercial side, technological capability to handle containers and swap body containers is heavily requested, as the need to transport these load units is much more prevalent than the need for rigid body trucks (CC3 & CC8). The Vice President at Autonomous Freight (CC8), further emphasizes the importance of developing robust trailer-agnostic vehicles to unlock the majority of business opportunities in the industrial site segment and access sites of high transport density, and CC1 & CC4 described the technical difficulties connected to developing a trailer agnostic solution. CC2, Senior Product Manager Autonomous Vehicles, comments the form factor development by stating that “the product must be developed to target the industrial site segment without invalidating it for other applications”. Hence, CC2 concludes that a versatile form factor is required to reduce the friction with customer operations when implementing autonomous technology at new customer sites. The Technical Director of Autonomous Freight (CC7), explains the importance of the form factor when going into an industrial site by stating that “my first question is; how are the goods of this site packaged?”, to see if the site is out of scope or not. According to FTE3, CC1 and CC7, it is also important to be aware of the payload requirements on a site, to ensure that it is technically feasible to handle. FTE3 explains that this can vary greatly depending on the industry, and exemplifies with the timber industry, where goods are both heavy and bulky. Transport professionals hold a general perception that targeting the large transport volumes of the transport system will require container handling, as the sites with the highest transport density are transporting standardized units that are incompatible with rigid trucks (UR3, FTE3). Also, transport industry actors clearly state that the bulk of their transport flows will be inaccessible with a rigid truck (LFL1, LFL2, LFL3). Another factor directly connected to the form factor, is the type of cargo on a site and the payload of the cargo (CC3, CC5, CC7). Different sorts of controlled outdoor environments are to a varying degree challenging for the autonomous technology in terms of cargo packing and cargo characteristics (LFL1, CC7, UR1, RC1, UR3). For instance, the goods of the port segment are to a large degree containerized (RC1, LFL2, LFL3) and in the chemical industry goods are too sensitive to be handled autonomously (CC7). Moreover, cargo types can determine how efficiently you can operate a vehicle on site, as certain sites have operational set-ups that require detachable load units, for them to act as “warehouses-on-wheels” (CC2, CC3). 33 4.1.2 Size of automation opportunity Size of automation opportunity Site characteristic Times mentioned Size 6 Transport volume 6 Multiple facilities 2 Distance 6 VAS 2 Table 4.2: Characteristics linked to size of automation opportunity As touched upon in the previous section, the size of automation opportunity on a site is critical when looking from the commercial perspective, the characteristics linked to this can be seen in table 4.2. Essentially, the size criteria is summarized by a transport economics expert (UR3), who claims that “the amount of internal transportation is proportional to the size of the plant”. Throughout the interviews, there has been a compelling consensus regarding the need for sufficient on-site transportation flows in order to reap the benefits of autonomous trucks (RI1, UR2, FTE3, CC2, CC7, CC8). According to UR2 and LFL1, having sites with multiple facilities will likely drive the need for more internal transport flows, increasing the possibility of finding suitable flows. Moreover, if there are value-added-services on the site, it may also lead to suitable flows (UR2). UR1 exemplifies value-added service with the case of consolidation, meaning to package different products on a site that are going to the same destination together. Sizeable opportunities for autonomous transportation on a site is an absolute necessity from a business case perspective. This stems from the need to spread the initial investment on as many vehicles as possible (CC3, CC8), moreover, the profitability is currently directly connected to the number of vehicles on an individual site (CC1). Larger site sizes means a higher amount of internal transportation volumes targetable by the capabilities of the current autonomous trucks, which would create an opportunity to operate a larger fleet while minimizing the number of sites where implementation investment, development of site specific solutions and customer conversion is necessary (CC4). The perception that site size and voluminous transport flows are necessary to motivate investments in expensive autonomous technology is not specific to the case company but also shared by logistics experts (UR1, UR2, UR3, RI1, FTE3). For instance, a logistics researcher (UR2) claims that “the autonomous solution must drive down the operational costs to motivate large investments”, and an experienced investigator of freight transportation (FTE3) states that “the internal goods volume must be sufficiently large to motivate the investment cost of the autonomous solution”.Moreover, the distance of the flow is also important according to (CC7, LFL3, UR3, RI1, FTE1, FTE3). The distance should be long according 34 to researchers of human and machine interaction in order to lengthen the fully autonomous operation. However, according to UR3 the short-haul truck operations are the most costly, in terms of driver costs. 4.1.3 Traffic environment Traffic environment Site characteristic Times mentioned Fenced-off 6 Mixed traffic 5 VRUs 6 Table 4.3: Characteristics linked to the traffic environment Moving on to the section regarding Traffic environment, the case company is at the current level of autonomous capabilities strictly focusing on controlled traffic environments as the primary use case and business opportunity of autonomous vehicles (CC4). The research and case company perspective agrees on the need for a fenced-off or semi-fenced off environment in driving contemporary autonomous trucks (RI1, RI2, RI3, LFL2, LFL3, CC4, CC7). This logically connects to the consistently mentioned issues of combining autonomous trucks with mixed traffic (RI1, RI3, LFL1, LFL2, CC1, CC4, CC6. CC7) and operating them in environments with vulnerable road users (RI1, RI3, LFL1, LFL2, CC1, CC4, CC6. CC7). The fenced-off characteristic can facilitate the definition of the circumstances in the surrounding environment of the vehicle, which constitutes the documentation necessary to build evidence of vehicle safety in a specific ODD (CC7). Moreover, the speed in these mixed traffic environments could be a problem, if there is a high difference in speed between the autonomous truck and the surrounding traffic it could be a safety issue (CC6). Specialists on humans in the transport system (RI1), the current capacity of autonomous vehicles requires them to be further developed in enclosed environments, which are protected or semi-protected. According to the Director of Regulatory Affairs (CC6), the regulations take the perspective of the vulnerable road user, hence the permit process becomes facilitated in fenced-off environments to which no unauthorized personnel have access. Current autonomous vehicles can not dynamically act as a part of their traffic environment, as many traffic situations and traffic signals are outside their capability scope, which turn them into a safety risk (CC6, CC7). Both RI1 and RI3 highlights how a potential mix of vehicles, both autonomous and non-autonomous, jeopardizes the security of the traffic system, and RI3 states that “it is in the fenced-off areas I believe business opportunities currently exist”. The perception of the case company is also clear, for instance CC1 claims the presence of vulnerable road users in autonomous operations to be highly challenging from a technical perspective, and the Senior Technical Director concludes that “public road operations come with a lot of risk”. 35 4.1.4 Operational conditions Operational conditions Site characteristic Times mentioned Predictable operations 5 Repetitive route 7 SLA/uptime 6 Table 4.4: Characteristics linked to the operational conditions The next category of characteristics is concerned with the Operational conditions of different sites, meaning the operational set-up and the site performance requirements, as well as physical attributes. The Team Leader of Technical Operations (CC5) elaborated on a concept called functional delta related to this, which is essential when assessing new sites and refers to the difference between the actual capability of the autonomous truck and the capability required on a site. Within the scope of this delta, one important parameter was the predictability of operations, that is to what extent transports are planned in advance or scheduled (LFL1, LFL3, CC1, CC2, CC7, UR2), which is in turn connected to the need for repetitive and simple routes (UR1, LFL1, LFL2, LFL3, CC2, CC5, CC7). Moreover, to meet the requirements of the customer, it is essential that the uptime criteria is met and reflected in SLA agreements (RC1, CC1, CC5, CC7, CC8, LFL1, FTE3). Appropriate preconditions for creation of customer satisfaction should exist, since the customer expects to buy an autonomous solution that works (CC4, CC8). The Product Strategy Manager in Autonomous Freight (CC1) states that “the uptime, predictability and precision of the autonomous vehicle is extremely important to ensure customer satisfaction”. Currently, a site with low precision requirement and flexible delivery windows is better suited to internal autonomous transport since it allows for uptime problems and vehicle breakdowns without causing interruption to site processes (CC1). Moreover, according to the leader of the solution implementation team (CC5), “the operational environment and the complexity of the route will determine the ratio of vehicles/driver”, which is crucial to the business case and perceived value of the autonomous solution (CC4, CC8). The amount of trucks managed by the same operator, or, on the flipside, the percentage share of a route that a vehicle can drive independently, will determine the total operational cost on site (CC4). Hence, route complexity is proportional to site staffing costs, and CC5 describes the optimal operational conditions of a site as “stable, predictable, open, large, and without traffic or pedestrians” to minimize the functional delta. Additionally, a presumption related to autonomous technology is that they are more efficient than ordinary vehicles, since they operate without driver constraints (CC3). 36 4.1.5 Site actor attitudes Site actor attitudes Site characteristic Times mentioned Collaboration between on-site actors 2 Table 4.5: Characteristics linked to site actor attitudes An important site characteristic discovered was also the state of on-site collaboration between the actors in a multiple-actor site (UR3, RC1, LFL1). A part of this is the attitude towards autonomy of the site owner (RI3, CC3, CC6, CC8). The commercial part of the case company is that having engaged and active partners in autonomous transformation is extremely facilitative and beneficial (CC3, CC6, CC8). The Senior Vice President of Autonomous Freight (CC8) points out how “finding more forward-looking customers is an extremely important parameter when deciding which sites to approach, in the same way as when replacing combustion engine trucks with electric trucks”. This alludes to the form factor discussion as well, since for instance the willingness to convert site transport into a new form factor requires a collaborative attitude and engagement for the autonomous solution (CC3, CC8). Two interviewed humanized autonomy researchers (RI1, RI3), emphasizes the importance of the site owner engaging in creating technology acceptance among the involved parties, using education and organized change management. Also external stakeholders, such as road-owners in favor of autonomous trucks are beneficial to implementation, as stated by the Director of Regulatory Affairs (CC6) , saying that “it is extremely important with proactive partners, for example to have a mature road owner saying that a beneficial traffic environment can be created in collaboration”. 37 4.2 General characteristics In the interviews, four types of general characteristics that are not specific to the sites, have been brought up, namely weather, regulatory, and economical conditions. These are outlined in the following paragraphs. 4.2.1 Weather conditions Weather conditions are critical according to several actors in the case company (CC1,CC2, CC5, CC7, CC8) and an external expert (RI1). According to CC1, “autonomous systems are sensitive and can be severely affected by water puddles, snow, fog, unpredictable environments with VRUs and obscured angles, these are difficulties that are shared over the entire industry”, announcing those as being the major challenges of the industry. Rain is a factor that is hard to handle for the autonomous system according to the team leader of technical operations (CC5). Interviewee CC1 concurs, and states that “operating in the dry climate of Madrid would be optimal”, referring to the limitations of sensor technology. Moreover, snow and leaves are a problem for autonomous vehicles, creating the need for extra maintenance on the site with operations such as snow and leaves removal according to a research director and a PhD in logistics (RI1). This drives down vehicle uptime (CC8) and creates the need of having backup vehicles to meet SLA requirements (CC5). Hence, weather conditions are a core component in the ODD definition, since they determine and limit the capabilities and the performance of vehicles in the local environment (CC4, CC7). Sites with a high degree of precipitation can be strategically problematic, for instance CC1 claim that coast-near operations and ports will require a as well as a sensor set-up makeover. According to CC5, the site selection is key in order to ensure a favorable climate, a precondition for high-uptime operations. 4.2.2 Regulatory conditions The weather conditions are directly connected to the regulatory aspects of autonomous vehicles, through the ODD. According to the Director of Regulatory Affairs at Einride (CC6) and a Senior Researcher of R&D Policy (RI2), the ODD definition is the basis of the current permit process, where regulatory exceptions are made for vehicles to drive autonomously under a set of specified conditions. EU law predisposes that the autonomous vehicle must be capable of handling all situations that it is faced with, hence the stability in fenced-off environments have risen as a potential facilitator in the search for a market where it is applicable in its current capacity (CC6, RI2). Meanwhile an autonomous vehicle can legally be granted technological exceptions and special permits from the law, general traffic regulations and VRU safety can in no way be compromised from the perspective of the law (CC6). According to RI2, some types of sites could consistently be more regulatory suitable for autonomous vehicles, due to their requirement of being strictly fenced-off for safety reasons, for instance manufacturing plants or ports with potentially dangerous machinery. However, CC6 claims that as technology develops, the need to search for environments with certain characteristics will be eradicated, as regulations would adapt when the safety requirement is fulfilled. RI2 concurs, claiming that the problem lies within the technology- and monetization rather than regulatory limitations. LFL3 is of the perception that regulatory change is 38 slowing the development down, but does in a similar way point to the need of looking into fenced-off opportunities to start out with. CC6 points out that while it is easier, the amount of sites that actually fulfill the criteria for being completely fenced-off is extremely low. While the technological capabilities are lagging behind, CC6 emphasizes that it is “extremely important to find proactive and collaborative regions and partners, who are willing to make traffic adaptations that facilitates autonomous vehicle implementation”, and CC6 presses how future physical- and digital infrastructure must be adapted to support disruptive forms of traffic. To summarize, the joint perception of the legal experts in autonomous transportation is that semi fenced-off and fenced-off opportunities should be pursued in the first instance. 4.2.3 Economic conditions Economic conditions have been mentioned several times by subjects in the study. As a freight transport expert (FTE1) and a researcher (RI1) puts it, the most important factor to have successful autonomous transport is to have a business opportunity that generates cash flow. Economic conditions can mainly be divided into profitability and operational efficiency. The profitability considerations have been mentioned both by researchers (UR2, RI1, RI2, RI3), a freight transport expert (FTE1) and internally at the company (CC5, CC8). The most mentioned consideration is that there is a hope that autonomous trucks will give cost savings to the company that operates the trucks (UR2, RI1, RI2, RI3, FTE1, CC8). Higher profit margins will, according to both a university researcher (UR2) and two internal interviews (CC5, CC8) be achieved through an increased number of vehicles on the same site. Moreover, profitability will be impacted by the number of FTE:s that are needed in order for the autonomous driving system to work (CC8), where a lower amount of FTE:s per operating truck will mean higher profitability (CC8). However, this could take some time to occur as a researcher (RI3) suggests that initially the implementation of autonomous trucks will mean creation of new roles and removal of others, making it potentially a zero-sum game. The operational efficiency considerations are mainly the improved efficiency that autonomous trucks are thought to achieve, this is a view that is supported both inside the case company (CC8) and by external parties in the form of a researcher (UR3), a research consultant (RC1), and freight transport experts (FTE1, FTE3). These efficiencies stem from not needing to adhere to planning restrictions due to driver rest times (FTE1, UR3), improved utilization rates of trucks (CC8) or the ability to drive on night time (RC1). Moreover, there are inefficiencies in short-haul transportation with low staff utilization that autonomous trucks can remove (UR3). Another operational efficiency consideration is the quality of transports, according to the case company (CC8) high quality of delivery, e.g. precision of deliveries, is what the customer demands. Moreover, another employee at the case company (CC1) describes that reliability of service is also very important to ensure high quality of service to the customer. This is also supported by a researcher (RI3) that describes that if the system works in the morning but not in the afternoon there is no need for the service. Overall, business case viability is evidently one of the most critical aspects in order to succeed in the scaling and adoption of autonomous freight, as stated by RI2, RI3 and CC8. According to CC8, the 39 developers of autonomous technology are in need of demonstrating a solution that breaks even in the daily operations, stating that “..it must lift beyond being pilots and R&D”. Moreover, an underpinning assumption of this thesis is that certain factors in the surroundings, or ODD, of an autonomous truck can improve its general level of function (CC4, CC7). In order to have a good business case and a good truck, the development efforts must be based on market needs (RI2, CC1, CC2) but a simpler environment could be a suitable starting point to support the development. 4.2.4 Autonomous loading and unloading The final consideration regards autonomous loading and unloading of goods. According to a researcher (UR2) there is no gain from autonomous technology if there is a need for two persons who are loading and unloading the vehicle. However, the existing autonomous solutions are costly (UR2). Moreover, the autonomous loading and unloading is context dependent (UR2, UR3 RC1) meaning it is not generalizable for every type of good. It seems that unit goods such as container or trailer goods are more feasible to handle autonomously (UR2, UR3, LFL1). Moreover, there is also an uncertainty regarding who has the responsibility for loading and unloading on site, if the responsibility lies on the autonomous truck provider there is a need to solve this autonomously (CC2), or else it will hurt the provider’s profitability. To conclude, according to two persons at the case company (CC2, CC8) and a university researcher (UR2) reaping the benefits of automation will ultimately require automated loading and unloading. 40 4.3 Site descriptions The aim of this section is to summarize the data gathered on different types of industrial sites that can be considered when looking for controlled environments suitable for autonomous truck implementation. It describes the characteristics at different types of industrial sites in the context of the current state of technology in autonomous transport. 4.3.1 Dry ports The findings on dry ports are mainly based on the inputs from UR1, an Associate Professor in Service Management and Logistics, also the specialist who invented the dry port concept. Dry ports are located anywhere geographically, but as a rule they will have a strong rail-port connection. The flow between the port and the dry port is commonly operated by train, although truck transport also occurs. It is characterized by high frequency transport, which is predictable with a high volume density. However, dry ports are purposefully positioned to serve as an access point for a hinterland that is large in volume and in need of the goods delivered by the port that the dry port is in turn connected to. As a rule, dry ports are strictly fenced-off since they handle non-tolled goods and provide full customs services and in some cases simpler value-adding services. Apart from this, the amount and character of internal flows varies a lot in between the dry port sites, although most dry ports are focused on simple transshipment operations. Some dry ports are located or directly incorporated in a larger logistics area, but not necessarily. UR1 claims that it is mostly in the flow between port-dry port, dry port-hinterland or dry port-logistics parks that exhibit the characteristics suitable for autonomous vehicles. However the majority of these will require open-road capabilities in the vehicles. Dry port operations would require a truck with a form factor capable of handling container transportation, but according to UR1 most dry ports will be too small to constitute a feasible opportunity for autonomous transport internally. The prospects to identify suitable internal flows to take over are small due to the focus on transshipment, and UR1 concludes by stating that “only extremely large transport volumes could create a viable case for internal autonomous transport inside dry port, freight volumes must be nothing but gigantic”. 4.3.2 Manufacturing plants Manufacturing plants has become a focus segment due to them generally offering simplicity in ODDs and regulations, although they otherwise exhibit significant internal variation based on industry affiliation and geographical location (CC2, CC7). A general observation from the interview study, is that a larger size of plant indicates a higher amount of potentially targettable flows, higher good volumes overall and also larger roads inside the plant (UR3, CC8, FTE3). 41 The variation in activity between different production facilities makes it hard to create an overview of the different transport needs in the segment. There are some joint functions of flows that exist on a generalizable level in many manufacturing plants, according to the Service Management and Logistics professor UR2. There are the ingoing flows, that usually goes from an on-site distribution center or an assembly plant, to production. Then, there is a category of outgoing flows, where goods and products are moved in a more finalized form from the production unit to further distribution or assembly (UR2, LFL3). The third variant of flows are going between the facilities, for instance storages or different production steps, and it can be viewed as if the different activities performed in the facilities are generating internal transports (UR2, FTE3). According to CC7, UR2 and FTE3, manufacturing plants do in many cases have multiple facilities and activities on the same or adjacent sites. For instance, CC7 mentions that for the manufacturers in the automotive- or home appliances segment it “makes sense to have all production steps on the same site, in close connection to a tightly clustered supply chain”, increasing the amount of targettable routes within a similar operational environment. CC2 and UR2 believes that the internal transports between facilities are suitable for automation, however the outgoing transports may be even easier to operate since goods processing is finalized and the need for precision in delivery and uptime is decreased. UR2 states that ingoing transports may be the hardest to operate autonomously, since the process input relies on extremely tight schedules, and the same thing goes for the transport from the surrounding supplier network. Moreover, FTE3 turns the attention to how many industries perform a majority of the production activities indoors, inside factories or warehouses, which may remove the need for transport between the facilities of the plant. In general, manufacturing plants with internal transport that is predictable and flexible in terms of SLA requirements are ideal for the current capabilities of autonomous vehicles (LFL3, CC1, CC7). Apart from the transportation set up of a manufacturing plant, CC7 emphasizes the importance of industry affiliation since it determines “several factors; form factor requirement, payload capacity, speed, and special equipment usage”. The Senior Technical Director (CC7) continues by saying that “the first question I ask when we assess a site is; what are the goods, and how are they packaged?” and points to the immense need of careful consideration in site selection. An illustrative example of this is provided by FTE3, who claims that while heavy industry, such as mining or lumber, is well-suited in terms of freight volumes, transport precision and uptime requirement, constitutes an technically immense challenge in terms of payload and equipment specialization. On the other hand, sites that handle fast-moving consumer goods have simple and frequent routes, as well as easily manageable cargo, but much unpredictability within operations timing and freight volumes, in combination with tight SLAs (CC7). CC2 also points out how it is common that manufacturing plants use containers as mobile warehouses, which releases the tractor to other activities in the meantime, and improves the efficiency of the transport vehicle. Hence, the fact that a rigid truck lacks this ability hurts this form factor's performance and operational flexibility that is required by certain environments. As far as the traffic conditions are concerned, this differs greatly considering that some plants will have public roads inside the site (FTE3, CC2), while others are strictly fenced-off for legal reasons, for example due to the presence of dangerous machinery (RI2). 42 4.3.3 Ports Throughout both external- and internal case company interviews, a recurring view is that ports would constitute an attractive arena for implementation and development of autonomous trucks (RI2, RI3, UR1, UR2, UR3, FTE3, CC6, CC7, CC8 ), while the two interviewed Port Development Specialists (LFL2, LFL3)