Application of soft systems methodology to frame the challenges of integrating autonomous trains within a legacy rail operating environment

Increased demand on rail, due to climate initiatives and passenger numbers, places significant pressure on existing railway operations; specifically on capacity, operational flexibility, and network robustness. These pressures are exacerbated by constraints, which prevent the construction of new track and infrastructure. This results in the need to use existing infrastructure and operating processes. One proposed solution is digitalization, which results in autonomous rail, where automated and connected intelligent transport systems facilitate smart traffic management. However, this generates the challenge of integrating autonomous trains and their associated technologies with existing infrastructure and operations. To understand this issue from an enterprise level, this paper has applied Brian Wilson's Soft Systems Methodology (a variation of Checkland's methodology) to the problem situation. This methodology explores and investigates the existing rail system in Great Britain (UK less Northern Ireland) and its stakeholders. The paper aims to propose a solution into how to transform the legacy system into one which incorporates autonomous operations and ultimately becomes fully autonomous. It culminates in a series of models that are relevant to those with a legitimate interest in the system. The models identify the activities required to analyze whether autonomy is worthwhile and if it is, how to successfully integrate it with legacy operations. Additionally, the models provide the basis for which a formal stakeholder analysis can take place.

75% of inland freight from road to rail (or inland waterways).However, the report 4 also identified issues concerning management, such as the complexity of freight prioritization, currently (in Europe) passenger operations take precedence.Capacity is also an issue as demand (and thus congestion) 5 is predicted to increase: urban passenger rail by 2.7 times from 2019 to 2050 6 and freight operations by 140% from 2013 to 2030. 3 Therefore, the rail industry faces the major challenges of increasing network capacity, operational flexibility, and robustness, while satisfying (e.g., financial and environmental) constraints.One proposed solution is the use of emerging technologies and digitalization, 7 where automated and connected intelligent transport systems facilitate smart traffic management, using autonomous trains operating with legacy rail operations and infrastructure.
However, this proposal raises questions on the pathway to autonomous rail from the current rail environment, which has little or no autonomy.Is autonomy the way forward, is it worthwhile?Can the potential benefits for those with a legitimate interest be validated and verified to offset and justify initial financial investment and risk?(Throughout, legitimate interest will refer to those stakeholders who can have an impact upon, or are impacted by, the proposed changes to rail.)Can autonomy be shown to be capable of increasing capacity, operational flexibility, and network robustness (including redundancy)-while also recognizing its impacts upon people and their future employment?These questions result in a complex and messy problem, which is susceptible to poor or partial understanding that results in erroneous conclusions.This raises the question of how best to approach and frame the investigation into this problematic situation.
The paper is aimed at those who have a background in SSM and its other applications, rail industry experts who want to know about other methodologies or processes to explore the future of rail, or other engineering personnel who are interested in incorporating autonomy is other forms of transportation systems.Its aim is to explain the development of an enterprise model that is relevant to stakeholders with a legitimate interest in Great Britain (GB) rail.(Great Britain is the union of England, Scotland and Wales, excluding Northern Ireland, which has a different rail operating system).The model's purpose is to describe the necessary factors to allow the transformation to autonomous rail.
We build a consensus view, which resolves the conflict between different stakeholder opinions on how such a transformation could be achieved.The model then identifies the required activities to establish a pathway to autonomy.Our approach is underpinned by Peter Checkland's and Brian Wilson's Soft Systems Methodology (SSM). 8,9lson diverges from the more well-known Checkland methodology, by focusing on a more practical, business-oriented approach, where the system is modeled as an enterprise, which seems best suited to our problem situation.SSM follows a seven-step process to understand an existing system.This methodology is based on recognizing that most socio-technical systems have a purpose, which delivers benefits and/or detriments to people, and are thus described as Human Activity Systems (HAS) that include human relationships and the interdependencies within them.A HAS is defined to be a conceptual representation of a system where people carry out purposeful activities. 10SSM analyzes the soft and hard engineering aspects of the problem situation and provides a logicbased approach, which stands-up to academic rigor.Here "hard" refers to a physical system, which has well-structured functional components with definable quantifiable attributes, 10 while "soft" refers to HAS, involving people, culture, organizations, politics, subjectivity, and all the complexities that follow. 11SSM enables relevant questions to be asked about the problem, identifies who the key stakeholders are, considers multiple viewpoints, and bounds the scope of the investigation, through the expression and then curtailing of the problem situation.
Most existing literature on autonomy within Systems Engineering, and particularly within rail, addresses quite focused business cases, such as rail vehicle operation and maintenance optimization, 12 or improving operational staff skills, 13 rather than examining the wider enterprise, as we do.However one paper of note is Clegg et al. 14 that uses an adaptation of SSM called Process-Oriented Holonic modeling to analyze a train operating company's customer service system during disruptions.This methodology is similar to traditional SSM, as it is built on action research principles, but differs when it explores the systemic issues, by modeling process-oriented holons (a whole in itself, as well as a part of a larger whole) to understand the problem situation.However, similar to the other literature, Clegg et al. 14 does not consider the whole rail system.
Our approach challenges conventional thinking on autonomous trains and examines whether autonomy is the way forward, investigates how autonomy can be delivered, and whether its perceived benefits make the transition worthwhile.Furthermore, the papers' outputs, including the models, enable any person to undertake the remaining steps of SSM and examine this problem space.The models also are essential for a formal stakeholder analysis to take place, as they inform the first iteration of the design of the stakeholder engagement themes and questions.
The paper proceeds as follows: Section 2 explains the selection of the SSM approach.Then Section 3 explores the expression of the problem situation, while Section 4 develops conceptual models representing the problem situation.Section 5 discusses the findings and Section 6 concludes the paper with recommendations for future work.

Selection of methodology
The initial analysis of the problem situation identifies the rail system as a HAS, as in addition to physical components, it involves multiple stakeholders with various beliefs and uncertainties around the value of autonomy within rail.The system is also considered to be "complex" and "messy."Complex, because the system has interconnected parts with dynamic relationships and nonlinear interactions, 11 and messy, due to the system being poorly defined, unbounded, with no single right solution, and thus a judgment is required. 11Also, there is an ethical dimension as potential solutions may have winners and losers, hence the system can be considered as wicked. 11o tackle this challenge, a number of problem structuring methodologies were considered, in addition to SSM.These methodologies were identified by a literature review of works, which were undertaken in a similar spirit to that of our own.These methodologies were: Socio-Technical Systems Design (STSD), Viable Systems Model (VSM), Structured System Analysis and Design Methodology (SSADM), and System-Oriented Design (SOD) where: • STSD methodology: whose origins are within social science, 15 is an approach which considers social, organizational, and technical factors of the system to inform its design.The methodology tries to enhance productivity by creating meaningful work through sociotechnical thinking, by balancing capabilities within new technology and keeping pace with new organizational design. 16 VSM: based on Stafford Beer's Cybernetics, 10 which uses cybernetics to assesses "how systems are viable" and "capable of independent existence" (maintaining internal stability and adaptability to change). 17Hence, VSM analyzes the structure and functions of the system that are necessary for its long-term survival.
• SSADM 18 : a hard systems methodology developed by DeMarco and Yourdon, developed from computing and information technology fields, but later adapted for broader uses, for example, in health authorities and large-scale private sector organizations.It uses a step process to analyze the current system and identify its requirements for the system's life cycle.The methodology is based upon the waterfall model, which breaks down project activities into smaller manageable tasks, resulting in a logical design, which is converted into a physical design.
• SOD 19 : a skill-based approach for designer to learn how best to cope with complex dynamic problems.It uses systems thinking to capture the complexity of the system, where SOD enables the designer to build their own interpretation and implement systems thinking into the design of HAS, via design thinking and practice.
To ascertain the most appropriate methodology, a Strength, Weakness, Opportunity, and Threat analysis were carried out for each potential methodology, where their advantages and disadvantages were identified, compared against each other.The criteria used were 1. Considers soft systems issues (allowing for messy systems with various stakeholder goals and beliefs).
2. Considers hard system engineering (physical parts of the system and the exchange of data and ideas).
3. Ease of application (no need to be a modeling expert).

Outputs applicable to the real world.
SSM was identified as the most appropriate methodology for our investigation.The analysis showed that STSD is too open to interpretation and can be difficult to follow, hence not appropriate for this investigation.VSM and SSADM on closer inspection have too strong an emphasis on the harder aspects of the system, and SOD, whose origins are based on SSM, does not bound the problem spaces sufficiently, and The analysis shows that SSM is good for soft systems, ok for hard systems, and better than or equal to the other methodologies for the remaining three criteria.SSM considers both hard and soft aspects of the system, it identifies emergent properties and generates purposeful activities, and these outputs are relevant and beneficial to stakeholders.

Detailed account of SSM
SSM is a seven-step process that aims to frame and understand the problem situation, which results in solutions which accommodate various beliefs and needs.The product of SSM identifies actions that can be feasibly implemented and facilitates the desired changes needed to develop the solution. 20The seven-step methodology 9 consists of the following activities: 1. Define problematical situation.
3. Formulate root definitions relevant to the situation.
5. Compare models with the real world.
6. Define changes to the situation (desirable and feasible).
These steps are shown in Figure 1, which has been adapted from Checkland's SSM summary diagram. 8Note the continuous processes linking the different steps of SSM.
The initial steps of SSM develop an in-depth understanding of the problem situation through the examination of the current system and engagement with its stakeholders.This enables the discovery of what the system is perceived to be from the viewpoints of its various stakeholders, hence identifying the system's key relationships and interdependencies.This information is then translated into the form of conceptual models in steps 3 and 4, which identify the desirable transformation activities required by stakeholders to achieve a "desirable change" (where SSM uses this phrase to describe the outcome which brings about stakeholders' needs and wants).These conceptual models are then compared against the real world in steps 5-7, and then adjusted accordingly, thus allowing for further questioning of the problem situation.This comparison ascertains whether these desirable changes are actually feasible and what actions are required to make these changes become a reality. 8r balance, it is important to note the potential limitations of SSM; for example, there are no guarantees that the methodology will facilitate open and honest dialog. 21Also, participants may not take ownership of the problem or commit to the process, 22 and due to the dynamic nature of the real world, the recommended actions may not always be possible. 21

Exemplar applications of SSM
SSM has been applied to a variety of contexts, disciplines, and countries; where uses include organizational design, information systems, problem solving, performance evaluation, and education. 22Within the United Kingdom, SSM has been applied to numerous governmental problem situations, for example, the National Health Service 23  The sugar industry. 21The problem situation investigated the traditional techniques and issues associated with the production of sugar cane in South Africa.The research brought together multiple stakeholders and identified the various, and often competing, objectives, they were independently pursuing.The analysis focused not only on "hard" technical methods to increasing efficiency, but also "softer" aspects relating to individual stakeholders' perception of the industry.
SSM facilitated the awareness of interdependence among stakeholders and the need for a common approach to problem solving, while at the same time identifying key groups with the power and decision-making capabilities to effect meaningful change.
Emergency response planning. 25This analysis investigated Sweden's preparedness for an electrical power shortage and brought together stakeholders to identify the formal and informal communication routes adopted both internally within the power industry, and externally with affected customers, when such an outage occurs.It considered roles and responsibilities during an outage, the authorization controls required to access sensitive information, and the political and social motivations behind the decision-making process.SSM provided an improved understanding of this complex problem and the relation-ships between stakeholders, resulting in improved response planning, applicable to broader emergency scenarios.
Computer game design. 26The design of computer games is a complex process involving many different stakeholders such as analysts, programmers, animators, designers, artists, and musicians, and while SSM is not useful for detailed technical game specifications, it was used to identify and communicate the scope, environment, and purpose of a new computer game for all those involved in its development.The models generated from the analysis were used to illustrate the flow through the game's levels, via the use of storyboards, resulting in a set of guidelines that depict the high-level design of the computer game.
In these examples, the authors observed the importance of identifying the key stakeholders, who hold the power to make the desired changes.They need to be on-board with these changes or the transformation will not occur (sugar industry example).It is also recognized that the problem situation can be applicable to wider systems that were not initially within scope, where findings can be helpful or transferable to other systems (e.g., other autonomous transportation in our case) and vice-versa (emergency response example).Lastly, the concept of using a high level approach, where over-arching principles are developed that stakeholders can agree upon is paramount before any fine detailing can begin (game design example).Thus, in this paper, we have put particular emphasis on identifying key stakeholders who have the power to influence changes, to use knowledge learned from other more mature autonomous systems, and to produce high-level goals that stakeholders can agree on.

Research method
Our research method was as follows: • A literature review and stakeholder analysis-to define the problem space and identify the key stakeholders, and their needs and wants.• The development of a rich picture-to illustrate these findings and express the problem space for the various stakeholder perspectives.
• The construction of the overarching root definition, and its breakdown into additional supplementary root definitions-to identify the transformation required to investigate the pathway to autonomous rail and the challenges identified by stakeholders.
• The development of an enterprise model-the expression and identification of the activities required to investigate the pathway to autonomous rail.
The research was carried out by members of the T-B Phase project, 27 who are experts in autonomy and collaborated with system experts from Thales (a major infrastructure company).
Note that this is the first iteration of the conceptual models and once a stakeholder engagement is completed (using the themes identified from these models), the models will be amended to incorporate these new findings and to form a validation process.

EXPRESSION OF THE PROBLEM SITUATION
An in-depth understanding of the issues surrounding our problem situation has been achieved through SSM's first two steps: (i) defining the problematic situation by performing a literature review and a stakeholder analysis, and step (ii) expressing the situation.This is achieved via the building of a rich picture (see Figure 3), which is a diagram that captures the essence of the issue, its conflicts, and understanding of it as perceived from the authors' and stakeholders' viewpoints. 23e literature review and stakeholder analysis identified that the main drivers for autonomy are: financial, political, social, and environmental, 28 due to the need to increase capacity, to improve reliability, to optimize the use of assets, and to make operations financially viable.Also, the analysis identified the current infrastructure used within existing rail operations and its various levels of automation and/or autonomy.Currently, the majority of autonomous trains operate in urban passenger environments (e.g., the Copenhagen Metro 29 ), whereas there are very few running in mainline freight (e.g., Rio Tinto AutoHaul, which is a closed system with dedicated tracks for freight 30,31 ) and even fewer in mainline passenger operations.Hence, we have identified a gap within the rail industry and the need to investigate autonomy within mainline operations.
A stakeholder analysis identified the key players within our problem situation and identified their perceptions of the system, autonomy, and how it may help meet future challenges.A quadrant graph (Figure 2), which maps stakeholders' power and interest in the system was produced, where key stakeholders are located in the top right quadrant and are Network Rail, regulatory authorities, trades unions, train operating companies (TOCs and FOCs).
A stakeholder needs and wants table (Table 1) for the proposed solutions has also been developed from this stakeholder analysis, to understand the drivers for change and where there are stakeholder conflicts.The needs will enable autonomy to be implemented successfully and the wants will enable the transformation to provide a desirable outcome.
Thus, using a holistic approach a rich picture (Figure 3 grate legacy and autonomous operations, who the key stakeholders are, the interactions and interdependencies within the system, and the desired outcome.The left-hand side of Figure 3 shows the legacy systems, while the right-hand side illustrates autonomous rail operations and its potential benefits.The middle portion of the picture shows the transitions process through the mixing of legacy and autonomous operations and explores the potential challenges in doing so. Also, Figure 3 shows the existing infrastructure within rail consists of three types: command, control, and communication (C3), track infrastructure, and on-board the train.Further, the picture highlights issues within the systems, such as it is not commercially viable for network owners and TOC franchisees to operate, thus the UK government subsidizes and provides grants. 45This may change in the future based on the newly published Williams-Shapps Plan for Rail Report. 46In addition, operations are greatly affected by unexpected incidents, which have a knock-on effects on the wider network, where failures cascade through operations resulting in significant delays and passenger dissatisfaction.Also, the current system does not take advantage of other transport systems with little integration, thus not maximizing on the benefits of a more effective and efficient wider transport system.
The perceived benefits of autonomy have been indicated in Figure 3, where operations could potentially become more reliable, effective, and efficient, through increasing capacity, operational flexibility, and network robustness.Integration with other transport systems could also make operations more cost effective and useful for passengers.
However, drivers' roles may change, which raises questions and conflicts with trade unions regarding their members' job security and conditions.There could also be issues around driverless trains and public perception of safety 47 , as well as who actually sees the benefits and profits from autonomy (i.e., which TOC operates an autonomous train first?).However, for any of this to take place, autonomy needs to be perceived to be worthwhile to those who are implementing the changes and incurring the initial costs; regulators, network rail, TOCs, and FOCs.Each of these stakeholders have different requirements, which will all need to be satisfied for autonomy to go ahead.

The expression of the problem space appears extremely useful and improves the understanding of the challenges the rail industry are facing and how potentially autonomy could help. Looking at the problem from an enter-
prise level furthers this understanding as rail is a system, which comprises of multiple other systems.Therefore, our models need to incorporate the various aspects/systems within rail and its stakeholder's differing perspectives.Hence, this need illustrates the usefulness of identifying stakeholders and ascertaining their interest and power within rail, through a quadrant graph, to identify those who have legitimate interest in the system.Furthermore, how the rich picture helps combine and simplify all these aspects and TA B L E 1 Stakeholder wants and needs.

Stakeholder Wants Needs
Catapult Advances in rail, transport, and technology sector. 32ecommendations for autonomy that can be applied to other sectors.

DEFRA and Environmentalists
No negative environmental impacts & operations more environmentally friendly. 33lution considers the environment.

DfT and UK Government
Reliable and sustainable service that is safe, secure, and integrated, with successful franchises. 46lution increases capacity & reduces operational cost.Also, solution integrates rail with other transport systems.

FOCs
Increased capacity, flexibility, and prioritization. 34 solution that increases network capacity and redundancy.

Investors
Product that is commercially viable.Autonomy integration is successful.
Local authorities Safe, reliable, and sustainable service, that provides local employment and is passenger-orientated. 35 Increased network capacity and rail system integrates with other local transport systems.

Manufacturers
Their products are commercially viable.
Guidance on new autonomous technology.
perspectives into one diagram gives one an overview of the problem space.
Hence, the rich picture can be used in a stakeholder engagement to start the discussion around this problem space.

CONCEPTUAL MODEL
Now that the perceived problematic situation has been expressed, a purposeful activity model is developed (step 3), which defines the activities that are needed to address the problem situation.Consequently, a statement, known as a root definition (RD) is used to describe the purposeful activity or transformation process from a singular viewpoint. 9 RD is based upon the idea that there is an input, which undergoes a transformation process which produces an output.To further enrich the RD and test the goodness of it, a checklist known as CAT-WOE is used. 48CATWOE checks that at a minimum, the RD consists of a transformation process (T) and a belief (W).In addition, the remaining CATWOE elements add further 'richness' to the analysis.These elements are: the customers (C), the beneficiaries or victims of the transformation, the actors (A), the people who undertake the transformation activities, the owners (O), the person or persons who could stop or change the proposed activity, and environmental constraints (E), the factors which affect the degrees of freedom within the system, such as laws, budget, and so forth.The RD can also include the criteria the activity needs to satisfy, for instance the performance measures to determine whether the purposeful activity has been a success or not.
For this investigation an overall RD for the problematic situation has been developed (RD1) and an additional 15 supplementary RDs have been established to further the understanding of the activities required to investigate the pathway to autonomous rail.The following statement has been created for the problematic situation (RD1) and was tested against CATWOE (see As mentioned, 15 supplementary RDs have been developed to increase the richness of the investigation.Figure 4 shows all 16 RDs and how they are structured and fit within each other.For example, RD3 and RD4, fit within RD2, which in turn fits within RD1.These RDs have then been sub-divided into 37 activity groups, which when combined form a CPTM (step 4), which is described by Wilson as an enterprise model. 23gure 5 illustrates the enterprise model for the investigation into the pathway to autonomous rail.It shows the required activities for the investigation and provides a logical method for checking whether there are any gaps within the proposed investigation process.Due to the complexity and depth of the model, Figure 5 is only a high-level representation of the model.However, the left-hand corner box, shows a lower level view of one of these high-level activities: "exploration of the current GB rail environment and wider transport system" (the red ellipse highlighted in the main diagram).The box details the various activities required to perform this activity, such as; defining the rail environment, its role, and identifying its challenges, while also listing the monitoring and control actions to ensure that these activities are carried out correctly.
The model consists of a set of activities comprising of inputs, transformations, and outputs, which are supervised through the use of monitor and control actions.Within Figure 5, there are four inputs, where external information feeds into the system for investigating the pathway to autonomous rail.Additionally, there are two outputs, (i) the pathway to autonomous rail and (ii) management of stakeholder requirements.These provide the next steps to the SSM process and future work where the conceptual models are compared to the real world (steps 5-7) through the use of a stakeholder engagement.This engagement validates and verifies the models and provides a series of purposeful activities 9 to investigate the pathway to autonomous rail.
The development of conceptual models has been imperative to understand the transformation process required to transition the legacy GB rail system into a fully autonomous one and how autonomy can be investigated.The application of SSM to this problem space has identified the usefulness of SSM.The methodology has helped us understand and express this complex system and problem space, which has various stakeholders with differing perspectives.Furthermore, SSM, with its clear steps, allowed us to go into an in-depth analysis of the GB rail environment and its potential transition to autonomy, while maintaining a high level, enterprise viewpoint of the situation.This culminated in a model, which laid out the activities required to investigate the pathway to autonomous rail and whether it is worthwhile.

Whether this was through; the use of CATWOE to construct the
From the analysis of the current system, expressing the problem situation and developing conceptual models, the following five key areas within GB mainline operations have been identified for further investigation (see Figure 3).
1.The first concern, which needs to be addressed, is the need for increased capacity.Creating additional new track sections, which run along existing routes across a network is extremely expensive, damaging to the environment, and politically challenging.Hence, there is a preference to introduce autonomous trains within an updated current infrastructure footprint.The use of autonomous trains is one possible way in which to increase capacity, due to their potential to operate with reduced headway.This can already be seen within future rail plans of increasing automation levels in ETCS (European Train Control System), 28 which allows for moving and virtual blocks to be implemented to ensure train separation.
However, there are significant issues surrounding the effectiveness of autonomous trains and their ability to increase capacity, especially with regards to the transition from legacy to autonomous operations.These issues include: (i) what proportion of autonomous trains are required to see a benefit to capacity, (ii) does the ordering of legacy and autonomous trains have a significant impact (e.g., platooning train types)?(iii) Would platooning need to be considered as well as prioritization in scheduling?(iv) Is it possible to prove that freight scheduling occurs at the very last minute, meaning there is reduced flexibility (especially at pinch points along rail corridors).However, autonomy could help solve this issue, providing additional flexibility through minimizing reliance on staffing and creating a more agile operating system, through better timetabling and the ability to change train services and operations at short notice.However, whether these benefits can be achieved via the integration of autonomous and legacy operations needs further investigation.
3. The third issue is increasing operational robustness including redundancy.Current networks have limited operational robustness, where failures can cascade through networks resulting in significant delays which are costly to TOCs, FOCs, network owners, and create dissatisfaction in the traveling public.However, utilizing autonomy may help solve this: with trains being connected and transferring additional information between: each other, different stations, and control centers (e.g., position, speeds, routes, or failures).This might enable autonomous trains to react to incidents quicker than currently possible, thus creating more efficient and effective operations, which are more responsive to failures.However, further investigation is required into whether this decisionmaking is made from a control center, by the train or a mixture of the two.Also, there are significant technical questions around how this control will work, especially with mixed legacy and autonomous fleets, as the benefit might only be achieved on fully autonomous lines.
4. Fourth, command, control, and communication (C3) is essential for successful operations, however autonomy and its operations require considerable thought into how exactly mixed and autonomous running will coexist.One question in particular concerns command and control (C2): where does the balance lie with decision making?This could be with the trains, the C2 centers or a hybrid solution, and if it is the latter how does the control get handed over?To ensure C2 works effectively there also needs to be good communication between trains and rail operating infrastructure.With autonomous trains having greater situational awareness, it may provide the opportunity for a train itself to reroute or adjust its initial path to deal with local incidents by having the ability to make that decision.However, to make wider tactical and operational decisions autonomous trains need to know the wider operational picture and one might question whether these decisions should be approved by a human operative.Therefore, a decision on what level of control is given to an autonomous train before human intervention is required needs to be investigated further.
5. Finally, we need to consider the impact of autonomy on policies, regulations, and standards.These will need to be carefully analyzed to incorporate autonomous trains and mixed fleet operations.
It is likely that a new set of safety and cyber regulations, which incorporate autonomous operations, will be required.These regulations will need to be validated and verified before autonomous trains can be adopted, to ensure the safe operation and acceptance by the public and operational staff.Autonomy also raises issues around infrastructure updates, staff role changes, and operational procedures.
All five of these areas will need to be further investigated, hence these topics have been explicitly incorporated within the enterprise model and will be analyzed in future work.Additional future work includes: the remaining three steps of SSM need to be carried out, therefore a further stakeholder engagement must be undertaken.This engagement involves stakeholders expressing their perception of the problem situation and building their own conceptual models.These views and models are then compared against our conceptual models (step 5), which in turn are adjusted to better represent the real world.
The models are then used to develop recommendations on how best to proceed with the pathway to autonomy (steps 6 and 7).

CONCLUSION
This paper explores the application of SSM to the practical domain of GB railway system, where it is analyzed from an enterprise level.
In this paper, we have described the use of the first four steps of SSM to understand the challenges of integrating autonomous trains with the GB mainline legacy rail operating environment.Through the examination of the current system and the expression of the problem situation to illustrate the various stakeholder perspectives, multiple conceptual models have been developed.This analysis has identified five key areas that stakeholders have deemed to be important and require further investigation for autonomous rail to become a reality.These being capacity, operational flexibility, C3, and policy, regulations and standards.Additionally, the development of an enterprise model has identified the activities required to investigate the pathway to autonomous rail and provides the themes for a formal stakeholder analysis.
Future work should include the verification and validation of the enterprise model through a stakeholder engagement and completing the remaining steps of SSM.This will deliver a series of recommendations regarding an appropriate approach to the pathway to autonomy.Once this is achieved, it is planned to develop an architecting framework for the pathway.

F I G U R E 1
SSM seven-step process.It is split into conceptual (red ellipse) and real-world (black ellipse) models.Steps 2-4 are iterated to improve the model.relies too much on the designer's own interpretation-and hence, are not ideal for this investigation.
and the defense sector (British Aircraft Carriers 24 ).In addition, the following three examples, chosen from academic literature, focus on problem situations, which consider organizational and technology-enabled change (similar to ours).All three used rich pictures, root definitions, CATWOE, and conceptual models to analyze their problem situation and develop purposeful actions.
) has been built to clearly express the problem situation: namely, how to inte-F I G U R E 2 Stakeholder quadrant.DEFRA, Department for Environment, Food, and Rural Affairs; DfT, Department for Transport; Catapult, Connected Places, and Transport Systems; TOCs and FOCs: Train and Freight Operating Companies.

Figure 3
Figure 3 explores the transition from legacy to fully autonomous operations using mixed fleet.It identifies key questions around how best to integrate autonomous trains in order to increase network capacity, flexibility, and robustness-while considering what potential regulatory updates are required for safe and secure operations and how the system will handle unexpected occurrences, single and multiple point failure modes ("what ifs"), and their cascading effects.

5 DISCUSSION
RDs, to ensure all aspects of the transition process has been included (e.g., actors, viewpoints, transformation, etc.), to use supplementary RDs to add richness and depth to the overall RD, or the final development of the enterprise model to show all the activities required to investigate this transition.These conceptual models have enabled us to outline the pathway to autonomous rail, at a sufficient level of detail to allow us to act and investigate this transition.In other words, without CATWOE and these RDs, the analysis and investigation of the pathway of autonomous rail could not happen, as they are required to identify the various branches and viewpoints in this investigation.So far, this analysis of the current GB rail environment and the development of an enterprise model comprises of the first four F I U R E 4 definition structuring and breakdown.steps of SSM and the work towards developing a pathway to autonomy.
the initial financial risk to those stakeholders who have to underwrite the transition (UK Government, network owners, TOCs, and FOCs) is worthwhile? 2. Second, operational flexibility: due to the modal shift from road to rail, one key question is whether rail operations can cater for the increased demand and provide the same level of flexibility as road transportation for local hub to hub journeys.Last mile delivery is difficult for rail, but other novel uses of technology, for example, drones, might provide an alternative solution to road as part of an integrated transport solution.Existing operations within the GB prioritize passenger operations over freight, meaning the majority of freight transportation occurs at night.This is greatly impacted by maintenance, which is also usually scheduled at night, hence reducing operational flexibility.Issues also lie with the fact that passenger operations can be timetabled months in advance, whereas I G U R E 5 Enterprise model for the pathway to autonomy (high level)-each ellipse represents an activity group and arrows show information and activity group links.Double headed arrows indicate bi-directional information flow.The insert in the bottom left-hand corner shows a lower level, more detailed set of activities that are required to perform the "exploration of the current GB rail environment and wider transport system" activity (the red ellipse highlighted in the main diagram).The red ellipses in this exploded view denote individual activities, C.A. are control actions to ensure activities are conducted correctly, and C are constraints that need to be satisfy.The rectangles indicates inputs (yellow) and outputs (gray) of the model.

Table 2
).A government owned system to facilitate a safe, secure, effective, and efficient delivery of future GB rail service operations within a defined time frame, by identifying and assessing the benefits and costs of varying degrees of automation within train operations and the transition to an appropriate TA B L E 2 CATWOE test for RD1.