Robot Teleoperation Design Requirements from End Users in Nuclear Facilities

Robot Teleoperation Design Requirements from End Users in Nuclear Facilities Alperen Kenan1, Paul Bremner1, and Manuel Giuliani2 Abstract— Despite the nuclear industry’s reliance on advanced robots being operated by humans, much of the existing research overlooks the operator’s perspective in the context of nuclear decommissioning. This study aims to address this gap by identifying the specific needs and requirements of robot operators in nuclear environments. Three focus groups of experienced robot operators from the UK Atomic Energy Authority and Sellafield Ltd. were conducted to explore key themes, including the operator’s role, tasks where robots are employed, and the risks associated with robot use. Findings reveal that: (1) robots in critical tasks are typically controlled by a team of operators; (2) for human-robot interfaces safety and reliability are the most important features, before effectiveness, intuitiveness and task focus; (3) due to high task variety operators see a need for various types of robots; and (4) operator error is regarded as the most significant and unpredictable risk. Based on these insights, a comprehensive set of 10 robotspecific requirements and 10 overall user requirements has been formulated. The paper provides recommendations for robot operators and designers, detailing how these identified requirements can inform the development of future teleoperated robots for nuclear decommissioning tasks. I. INTRODUCTION Robots deployed in nuclear environments offer the potential to minimize or eliminate human exposure to hazardous environments and prioritise safety, efficiency, and cost-effectiveness [1]. The use of robotics and computerized tools in nuclear facilities is identified as a highly recommended practice by the International Atomic Energy Agency (IAEA) [2]. The challenging and unpredictable environment of nuclear facilities necessitates that robots be teleoperated by human operators during tasks, raising the question of how to design an effective robot teleoperation system that ensures both reliability and safety. Various approaches have been explored to address this problem, including research on sustainable decommissioning strategies [3], [4], investigations into past nuclear accidents [5], [6] and times of crisis [7], consideration of human factors [8], [9], and the exploration of diverse robotic challenges [10], [11]. However, the operator’s perspective is often overlooked; what a designer believes a user needs may not always align with the user’s actual requirements, resulting *This work was supported by the European Commission’s Marie Skłodowska-Curie Actions (MSCA) Project RAICAM (GA 101072634). 1Alperen Kenan is a PhD student, and Paul Bremner is an Associate Professor in Human-Robot Interaction at the Bristol Robotics Laboratory, University of the West of England, Bristol, United Kingdom. Emails: [email protected], [email protected] 2Manuel Giuliani is a Professor for Geriatric Robotics at the Faculty of Electrical Engineering, Kempten University of Applied Sciences, Kempten, Germany. Email: [email protected] Fig. 1. Various types of robots utilised in nuclear facilities. Grounded robots include A: telemanipulators, B: industrial robots, and C: cranes. Mobile robots include D: UAVs, E: ground ROVs, and F: underwater ROVs. in usability issues. A more user-oriented design approach is necessary for achieving effective outcomes. To address this research gap we have conducted a design requirements analysis [12] as part of a user-centred design approach for producing a teleoperation system. Three focus groups [13] were conducted with a total of 15 experienced robot operators from the UK Atomic Energy Authority (UKAEA) and Sellafield Ltd. to explore key themes such as the operator’s role, preferences in human-robot interfaces, tasks involving robots, and the possible risks associated with their use. This paper makes three primary contributions: (1) we present an in-depth analysis of focus groups involving expert robot operators in the nuclear industry, offering insights into the end-user experience, the human-robot interface, operational tasks, and associated risks and hazards; (2) we derive both robot-specific and overall design requirements, proposing potential improvements for future applications; and (3) we discuss how these requirements can serve as a foundation for user-centred design in the development of future robot teleoperation systems. II. RESEARCH CONTEXT Background: To establish effective design requirements for teleoperated robots in the nuclear industry, it is crucial to understand the specific needs and preferences of the operators who remotely control these systems. Identifying these needs requires considering several key areas that we review here. Robot types: The practical use of robots in nuclear facilities dates back to the 1950s, and it remains a critical topic due to concerns about human and environmental safety [14]. Different tasks require different types of robots, which can be broadly categorized into mobile and grounded systems. Figure 1 illustrates the variety of robots used. Mobile robots consist of unmanned aerial vehicles (UAVs) [15], groundbased remotely operated vehicles (ROVs), which include wheeled and legged ROVs [16], and underwater ROVs [17]. Grounded robots consist of industrial robots [10] and telemanipulators [18] with haptic feedback, as well as electrohydraulic systems [10] and cranes [19]. These robots vary in form: mobile and industrial robots are often off-the-shelf; others are bespoke systems or custom-built from scratch. Tasks: Robots offer a safer alternative to human workers in various tasks within nuclear power plants, including material retrieval, swab sampling, inspection, surveillance, cleaning, decontamination, and waste handling. By employing robots for these functions, facilities can significantly reduce risks and operational costs [20]. In the context of nuclear decommissioning, robots are being developed to perform tasks such as perceiving, grasping, cutting, and manipulating waste [1]. Challenges of Nuclear Environments: In the nuclear industry, we conducted focus groups in two key sectors: decommissioning [4] and maintenance [21] in both nuclear fusion and fission facilities. These sectors present significant operational challenges, making them highly relevant for understanding the requirements and limitations of humanrobot interactions in these contexts. According to the IAEA, decommissioning refers to the administrative and technical actions taken to remove regulatory controls from a facility [22]. Key challenges for robotics in nuclear decommissioning include high radiation levels, contamination, shielding, access limitations, communication issues, and the diverse characterisation of nuclear facilities [23]. The UK Atomic Energy Authority identifies robotics-based maintenance and remote techniques as essential challenges for fusion energy research [24], with specific challenges including architecture optimisation, operations in confined spaces, handling complex components, ensuring environmental compatibility, and enhancing rapid response capabilities. Related Work: Design approaches in previous work have sought to bridge the gap between robotic research and development in other sectors and its deployment in the nuclear industry [25]. These designs include snake-arm robots for conducting inspection and repair operations [26], underwater robots for inspecting piping systems [27], and ground robots for inspection tasks [28], [29]. However, a notable shortcoming of these studies is that they primarily concentrate on the performance of the designed robots within nuclear environments while giving insufficient attention to the challenges faced by operators in controlling these robots. A valuable source of knowledge on robot teleoperation can be found in robotics challenges. The DARPA Robotics Challenge (DRC) aimed to develop human-supervised ground robots capable of executing complex tasks in challenging conditions. By examining the official results of the DRC [30], lessons learned from the competition and the outcomes of individual teams [31], valuable insights can be gained. Design guidelines were formulated by analysing the humanrobot interaction (HRI) characteristics of the most successful teams within the specific task conditions presented by the DRC [32]. The ANA Avatar XPRIZE competition aimed to develop a robotic “avatar” system to enable telepresence for a human operator [33]. Valuable insights can be obtained from the performance of competing teams [34]. Analysing the outcomes of these competitions provides important insights into the challenges encountered during the teleoperation of robots in tasks that reflect real-life scenarios. A notable difference is that these implementations are often proofof-concept demonstrations, conducted over short timeframes and typically allowing multiple attempts for each task or subtask. In contrast, real-world applications must account for operator stress, work shift durations, control room dynamics in industrial settings, and the involvement of actual end-users from the nuclear industry. User-Centred Design (UCD) addresses these shortcomings by actively involving users in the design process. Studies on teleoperated robot development have shown that the use of UCD methods can lead to a product that meets the requirements of the end user [35]. Various studies have examined the usability of different remote control systems and their interfaces [36]. Notably, teleoperation interface design has also been studied in other sectors, such as medical examination [37] and autonomous vehicles [38], where user perspectives are taken into account. There are no studies that focus on user requirements in nuclear facilities. This novel research seeks to bridge the gap between robot teleoperation designers and the needs of the nuclear industry by prioritising user requirements from the operators’ perspective. By highlighting the significance of operator insights, this work underscores the necessity of a more useroriented approach in teleoperation interface design. III. METHODOLOGY To gather user requirements, we employed the focus group method, a qualitative approach used to explore people’s preferences and viewpoints [13]. By fostering discussions among end-users, this method enables participants to build on each other’s ideas, yielding deeper insights than interviews. The focus groups followed a semi-structured discussion format, allowing participants to freely share their ideas while keeping the conversation aligned with key research themes through focused thematic prompts. A. Participants In total, 15 robot control experts participated in our research. Of the 8 participants who agreed to share their demographic information, all were male, with ages ranging from 22 to 53 years (M = 35.13, SD = 11.92). Their experience in similar work roles ranged from 6 months to 28 years (M = 7.02, SD = 8.75). Given the range of robots described in Section II, we identified three key areas of expertise essential to nuclear facilities: operators of grounded robots, UAVs, and ROVs. For grounded robot expertise, 9 operators from UKAEA experienced in controlling grounded manipulators participated. For mobile robot expertise, 6 operators from Sellafield Ltd. participated, 3 UAV operators and 3 ROV operators. Participant job titles included Graduate and Senior Remote Handling Operations Engineers, as well as UAV and ROV Equipment Engineers. Participants knew their fellow focus group members. The disparity in team sizes between grounded and mobile robots reflects the relative prevalence of these robots in nuclear settings, as well as organisational structures and operational requirements. Grounded robots, which perform precise physical interactions, typically require larger teams, whereas mobile robots, primarily used for visual inspection, require smaller teams. B. Data collection and analysis This study was conducted from December 2023 to September 2024 and comprised three focus group discussions and two on-site observational studies with field notes. Focus group sessions lasted about two hours on average. The session involving ground robot operation engineers was audiorecorded, whereas the other sessions could not be recorded due to facility regulations, so handwritten notes were taken during these sessions. Notes were subsequently documented and sent back to participants for verification to ensure accuracy and to address any potential misunderstandings. Followup surveys were used to gather participant demographic information. The observational studies provided real-world insights, complementing the focus group discussions. They highlighted the control room environment, operator interactions, task execution, and the researcher’s direct testing of the telemanipulator, thereby enhancing our understanding of the design requirements for nuclear settings. The frequency analysis method [39] was applied to the audio-recorded session. First, the audio recording was anonymised and transcribed. Parts spoken by the participants were extracted, and a frequency analysis was conducted to highlight keywords and categories used during the discussions. Computer-Assisted Qualitative Data Analysis Software (CAQDAS) was used to obtain word frequency. The frequency analysis results were used to quantify the occurrence of specific themes, and these findings were then merged with the framework analysis method [40] to provide a comprehensive understanding of the operators’ requirements, ensuring both the frequency and context of identified themes were considered. This method involves systematically applying a structured framework to identify key themes, patterns, and relationships across the data set. We first identified key themes and sub-themes from the data collected through notes and audio transcripts by implementing thematic analysis [41]. This information was then organised into categories to create an analytical framework. Using this framework, we indexed the data by summarising it into a matrix. C. Study environment and context UKAEA is a prominent UK government research organisation dedicated to advancing fusion energy. Our study involved a focus group with UKAEA’s robot manipulator experts and an observational study conducted at the Centre for Remote Applications in Challenging Environments (RACE) in Culham, Oxfordshire. RACE is a leader in the development of robotics and remote maintenance systems for the safe and efficient operation of fusion power plants. Sellafield Ltd., based in Cumbria, UK, is a prominent nuclear decommissioning site license company. Since its establishment in the 1940s, Sellafield Ltd. has become a global leader in nuclear waste management, site remediation, and decommissioning. Due to restricted access to Sellafield’s main site, we conducted a focus group with UAV operators remotely via Microsoft Teams. However, we were able to visit the Centre of Excellence, an off-site facility of Sellafield Ltd. in Whitehaven, West Cumbria, where we held a focus group with ROV equipment engineers. While both UKAEA RACE and Sellafield Ltd. operate within the nuclear robotics domain, their operational contexts differ. UKAEA RACE focuses on remote maintenance within well-defined fusion research environments, whereas Sellafield Ltd. addresses decommissioning in post-operational facilities, which present more variable and unstructured conditions. Despite these differences, both settings involve high-risk teleoperation and rely on experienced operators to manage complex robotic systems. The focus groups began with an introduction to our research and methodology, followed by questions on current remote technology use in nuclear operations. The discussions were split into two parts. The first focused on human-robot interaction in the control room, exploring operator roles, interface needs, and preferences. The second part addressed the remote environment, covering tasks and environmental challenges. The questions were developed by integrating user-centred design guidelines, focusing on understanding users, tasks, and the environment. They covered aspects such as user roles, preferences, long-term usage, expertise, goals, past designs, current methods, task structures, and environmental factors. Insights from the literature review ensured relevance to specific robot tasks. The questions were pilot-tested with external researchers for clarity and designed to prompt broader discussions rather than seeking individual responses. During the discussions, visuals were presented, explanations provided, and topic guides distributed, allowing participants to take notes. 1 D. Ethics Our study was approved by the Research Ethics Committee of the University of the West of England, College of Arts, Technology & Environment (CATE-2324-262). One week before the focus groups, we distributed participant information sheets and consent forms electronically, obtaining informed consent after addressing participants’ questions and before the start of discussions. Participants received a subset of simplified questions beforehand to inform them about the discussion topics and allow for preparation. Additionally, we implemented measures to ensure participant confidentiality and data security throughout the research process. 1Further details on the focus group structure, topics discussed, and information provided are available in the supplementary Focus Group Topic Guide document. https://github.com/kenanalperen/FocusGroup-Topic-Guide.git IV. USER REQUIREMENT ANALYSIS We applied thematic analysis to the focus group discussions to identify key insights related to end users, humanrobot interfaces, robot tasks, and environmental factors. These four themes were selected to help us understand the specific requirements of users when performing tasks in challenging environments. A. End User In nuclear facilities, end users operate robotic systems in various roles. Key roles include Robot Handling Operators, who control the robots and oversee camera operations; Remote Handling Operation Engineers, responsible for equipment operation and documentation; Safety Personnel, who authorise specific robot functions; Junior Operation Engineers, who assist senior engineers with daily operations; and Responsible Officers, who ensure adherence to procedures and safety protocols. Here we refer to ‘operator’ as any end user controlling remote robotic systems. Operators must understand the equipment, meet customer requirements, and write documentation independently. They mainly differ in experience, defined by skills and knowledge, as mastering safe operation can take months. Previous experience may not apply due to cultural differences between organisations. Some robots can be operated by a single operator, while those engaged in critical missions often require a team, and team dynamics in the control room are crucial. Multiple operators share responsibilities per shift, with one typically more experienced in handling unexpected issues. Initially noisy with questions, the control room settles as the team familiarises itself with their roles, fostering effective dynamics and stability. The team often adapts to the preferences of the lead operator, e.g. when choosing camera settings. Human involvement in decision-making is essential in the nuclear industry, with operators having the final say. Transitioning to shared autonomy may take decades and requires rigorous testing. Automation faces significant challenges, especially under suboptimal conditions, where reliable sensor data is critical, which can be disrupted by radiation. Manual control is preferred in scenarios with limited knowledge. B. Human-Robot Interface The human-robot interface needs to clearly communicate the robot’s status and environmental data without distractions. While some settings are based on principles, others depend on the operator’s preferences, which can make customization challenging, especially with rotating teams. This often leads to compromises among operators. Preferences vary, such as whether operators prefer sitting or standing while controlling the robot, as well as their choices for camera views and feedback. An adaptive interface that adjusts based on the user could be a helpful solution. Operators must manage a lot of information. For grounded robots, using multiple camera feeds across separate monitors can feel overwhelming at first but becomes manageable with experience. To keep things clear, a condensed version of the information should always be available. Most mobile robots rely on standard equipment, often with just a fixed camera. The quality of these cameras varies: 4K models are used for general tasks, while lower-resolution, more durable cameras are used in radioactive environments. Live feeds are typically displayed on controller screens or larger monitors. Operators are open to testing new interface alternatives, as long as they are reliable, safe, and efficient. New technologies should ideally be tested in other industries before being introduced in the nuclear sector. Experts also point out that what works well for short periods may not be effective over long shifts. A design that’s fine for 30 minutes but uncomfortable after eight hours isn’t practical. Fig. 2. Alternative Interface Options. Figure 2 shows alternative interface options, with expert opinions for each. Virtual Reality (VR) is a valuable training tool for simulating errors in a controlled environment, though it cannot replace hands-on experience. While it is engaging for short periods, it is less suitable for extended use due to movement delays, motion sickness, and limited spatial awareness. As a result, VR is best seen as a supplement to real-world training rather than a substitute for real-life scenarios. Gesture-based control was found to be ineffective, with experts reporting inconsistent outcomes, increased effort, and concerns that operators may forget the gestures correspond to robot controls. Similarly, eye-tracking systems, which aim to reduce workload, present challenges when used with multiple screens, often resulting in inaccurate responses. Vibration is useful for alerts, but it should not be disruptive. Notifications should be sent to the entire team, rather than just the individual operator. Force feedback is critical for tasks involving rigid contact, but safeguards must be in place to prevent harm to the operator. Both force and visual feedback are essential for effective position control and for preventing damage during physical interactions. Tactile feedback can enhance tasks like swabbing. While operators tend to prefer physical interfaces such as buttons and joysticks, some controllers incorporate touchscreens to provide more versatile input options. Spatial audio feedback presents concerns due to environmental background noise, which can make it difficult to distinguish helpful sounds from distractions. While audio cues can reduce visual workload, repetitive cues can lead to desensitisation, causing operators to overlook critical warnings. All cues must be accessible to the entire control team. Voice commands are commonly used for human communication, but experts hesitate to automate them due to noise and accent variability, where the risks may outweigh the benefits. Human-to-robot voice commands must be consistent, and the system should adapt to evolving team dynamics as operators tend to use fewer words over time. C. Tasks The use of robots in tasks where human presence is impractical is crucial. Various tasks are performed by teleoperated robots, as illustrated in Figure 3. Fig. 3. Various Tasks performed by teleoperated robots Tasks involving physical interaction with the environment are critical in nuclear operations and often require both visual and force feedback. In delicate operations, reducing force feedback can enhance performance, particularly when visual feedback is more crucial for precise position control. Robots lacking force feedback are dependent on the operator’s expertise for contact-based tasks. In lowradiation environments, robotic systems are used for material handling, including transporting barrels and components, with manipulators capable of handling loads of 5 to 10 kg, while cranes are utilised for heavier items. Decommissioning activities involve mechanical search operations, where tiles may be removed in known environments or debris cleared in unknown settings, requiring operator proficiency. Swabbing, essential for contamination analysis, demands precise contact force control. Destructive inspection methods, such as puncturing and drilling, are less common but can be adapted for tasks requiring the extraction of small material samples. For inspection, mobile robots equipped with multiple sensors assess measurements, valve positions, and room temperatures. While certain tasks require physical contact, most inspections are non-contact, encompassing dimensional, visual, radiation, temperature, and surface contamination assessments. Many mobile robots utilise LiDAR for real-time 3D mapping, which is crucial during signal loss, enabling autonomous return to known locations. Grounded robots typically rely on CAD models and static as-built data for mapping. Although research has explored integrating static and dynamic maps, practical use remains limited, with visualisation mainly relying on camera-based systems. D. Environment 1) Obstacle Considerations: Access ports in vessels are key constraints due to their fixed size, while obstacle shapes and sizes vary by environment. Water and debris can interfere with sensors, necessitating robot adjustments, as sensors may struggle to function in aquatic conditions. Structured environments are predictable, whereas unstructured ones can change unexpectedly, posing challenges for operators who must work within known areas to avoid hazards. Implementing dynamic mapping could enhance change detection and real-time data integration in such environments. For UAVs, collision avoidance is effective outdoors but lacking indoors; thus, protective cages are essential. 2) Risks and Dangers: Figure 4 illustrates the common risks and dangers associated with teleoperating robots in nuclear environments. Fig. 4. Risks and Dangers Operator error is the most significant and unpredictable danger in robotic operations. Mistakes by operators can lead to unforeseen damage to robots, making such errors difficult to anticipate and mitigate. While environmental hazards can often be planned for, operator actions remain unpredictable. Robots must be designed to withstand the radiation levels of their operational environments, with their sensors and electronics similarly resilient. Environmental conditions can vary widely; knowing the temperature limits is crucial for selecting appropriate robots. Outdoor UAVs also need to function effectively in cold temperatures. Wireless communication faces challenges in nuclear facilities due to heavy shielding from radiation protection. This often necessitates the use of cables or repeaters for connectivity. For mobile robots, the ability to return to the last known connection point is vital in case of communication issues, while outdoor UAVs can autonomously return to base using GPS. Operators should be notified of connection strength and alerted to any weaknesses. Task duration is limited by battery life, requiring mobile robots to manage their power effectively to avoid becoming unusable or classified as nuclear waste. Humidity and water presence pose additional challenges, as robots must function reliably without malfunctioning due to debris or rain. Dust and contamination can damage both mechanical components and sensors, with UAVs potentially lifting dust due to downwash. Noise generated by robots can also interfere with operations, although newer UAV models are becoming quieter. Certain environments may also feature high magnetic fields or vacuum conditions that can disrupt robot control, though these situations are not universally applicable. V. DISCUSSION A. Key Findings The findings from this study reveal several critical aspects of human-robot interaction in nuclear facilities. Our main finding is that robots engaged in critical tasks are generally controlled by a team of operators, each assuming different roles in the control room. Second, safety and reliability are the most crucial aspects of the human-robot interface, taking precedence over other features like effectiveness, intuitiveness, and task focus. Third, given the diversity of tasks performed in nuclear facilities, there is a need for multiple types of robots to minimise risks to human workers. This calls for flexible robotic systems capable of addressing a range of operational challenges. Finally, operator error was identified as the most significant and unpredictable risk, suggesting that further research is needed on how to mitigate human errors. These findings collectively emphasise that while technology continues to evolve, the human role remains central to ensuring the safe and effective operation of robots in critical environments. B. User Requirements To derive user requirements, we continued with the framework analysis method. Following data familiarisation and thematic framework identification from Section IV, we performed indexing using the frequency analysis. Based on the themes and indexing, we summarised the data into a charting matrix. Finally, we identified and mapped user requirements for different categories, i.e., telemanipulator operators, UAV operators, ROV operators, and overall requirements. 1) Telemanipulator Operator Requirements: Both position and force control are important for telemanipulator operators. Based on our analysis, the requirements for telemanipulator operators are as follows: T1 The system should deliver both force feedback and visual feedback to the user upon any physical contact with the environment, enabling the operator to maintain effective control over force and position. T2 Force feedback should be scalable to suit task delicacy, operator physical strength, and preferences, allowing users to optimise feedback for control and comfort. T3 The system should include adjustable limits to prevent excessive forces on the operator’s side and the remote environment, ensuring user safety and minimising object damage. T4 The gripper of the manipulator should incorporate a locking function to secure objects during movement, allowing the operator to handle them confidently without the risk of dropping an object. 2) UAV Operator Requirements: Controlling UAVs in enclosed environments brings specific challenges. U1 Operating UAVs in enclosed spaces necessitates the use of protective cages to minimise collision damage and enhance operator confidence by mitigating navigation risks. Most indoor UAVs lack collision avoidance systems and rely on operator skills for safe navigation. U2 UAVs should be equipped with camera and LiDAR systems to enhance operators’ situational awareness and facilitate informed decision-making. U3 The system should provide real-time mapping and ensure that the UAV autonomously returns to the last connection point in the event of disconnect, increasing the operator’s confidence by mitigating the risk of loss. 3) ROV Operator Requirements: R1 An interface should feature expanded displays that provide the operator with comprehensive information on the robot’s status and its environment, rather than relying solely on a single camera feed. R2 For underwater ROVs, job-specific information, including depth from the surface, offset from the bottom, and distance from the base location, should be clearly displayed to the operator. R3 Designs should include backup plans for robot retrieval in case of failure. Tethered robots are preferred for easy retraction but must address challenges like tangling and limited range. Operators should have the flexibility to choose whether to use the robot in tethered or untethered mode, depending on the task. 4) Overall Requirements: O1 The human-machine interface must prioritise safety and reliability, followed by efficiency, clarity, and intuitiveness. O2 The number of operators controlling the robot should be taken into account during the design process. If multiple operators are involved, display outputs should be accessible to all team members. O3 The interface should provide essential job-specific information, ensuring operators are aware of the robot’s status. To minimize distractions, data filtering can keep key information accessible when needed, avoiding constant display on the main screen. O4 Teleoperated robots are preferred over automation. Even when shared autonomy functions are used, the system must provide a manual control option to address unforeseen outcomes and enhance operator confidence in the event of unexpected automation errors. O5 The mechanical design of the robot should provide customisation and flexibility for the operator, including connection points for easy attachment of sensors and task-specific parts, as well as modular equipment for quick part replacement. O6 Collision avoidance limitations should be integrated into both the interface and the robot to protect humans, the robot, and the environment. Although precautions can reduce the risk, collisions cannot be fully eliminated; therefore, the interface should be designed considering potential physical crashes, ensuring the operators can operate with confidence. O7 The system should provide warnings for unexpected behaviour or hazardous conditions, using auditory signals, vibrations, or visual displays, and ensure they reach the entire control team without disruption. O8 The remote environment can be quite noisy, making constant sound streaming potentially distracting for operators. Audio should still be accessible when needed. O9 Operator error is considered the most significant and unpredictable risk, making it essential for designers to prioritise human-robot interaction in the design process. O10 Data recording should be enhanced to enable operators to replay tasks for analysis and serve as a black box in the event of failures. It is crucial to log comprehensive data on the robot’s status and the environment. C. Implications of the Requirements Requirements T1, T2, T3, and T4 have been implemented in the RACE Mascot system [42]. However, these requirements are not met by commercial off-the-shelf robots. MultiDoF bilateral teleoperation with force feedback, including position/force scalability with limits and operational functions, remains an area of active research. The robots used were all developed in-house with bespoke modified systems, which are not directly available for use by other facilities. Requirements U1, U2, and U3 were met by the specific UAV operated by the Sellafield operators, the Elios 3 [43]. However, some facilities may opt for other UAVs due to high-cost considerations, and most commercial indoor UAVs do not meet these requirements out of the box and require modifications. We suggest that the UAVs implemented should meet these requirements, either by using specific commercial UAVs or by modifying off-the-shelf models. For requirements R1 and R2, most ROVs display information on the robot’s status and environment, such as battery levels and temperature. However, the information varies significantly between robots, and end-users are dissatisfied with the lack of accessible detail, particularly the reliance on a single camera feed. For R3, most commercial off-the-shelf robots use battery power and are not designed for tethered operation. Some robots feature a ‘return to last connection location’ function, but operators find it unreliable and prefer tethered robots for better battery management and the ability to retrieve them in case of issues. For O1, end-users agreed that safety and reliability are priorities for designers, but efficiency, clarity, and intuitiveness are sometimes overlooked. For O2, most designers assume a single operator controls the robot, with the display targeting a single user. However, in critical missions, a team is often involved, and team dynamics must be considered in the design. For O3, O7, and O8, operators mentioned that information should be accessible without distracting the team. They also noted that warnings and audio feedback are sometimes insufficient and do not reach all team members, while at other times they are overly distracting. Regarding O4, human-in-the-loop is a requirement for nuclear industry missions, with automation still being debated. Operators feel that the technology readiness level is insufficient for critical missions. O5 is available in some robots but not all, with operators noting that the ability to attach taskspecific parts is important. For O6, while some robots have collision avoidance functions, operators stated that these are not always reliable, and robots should always be designed with collisions in mind. O9 is widely accepted by the community, as operator error remains the most significant and unpredictable risk. O10 is often not fully implemented, with operators reporting issues related to large data sizes, storage limitations, and data security regulations, which can prevent easy task recording and playback. D. Future Directions for Development In addition to the requirements, our analysis of the focus group data identified the following potential system development pathways: F1 A log-in function could be beneficial for shared control interfaces, allowing the system to recognise the operator in control and allowing customisation of the priority, location, style, and size of displayed information. F2 Having a digital twin interface that displays real-time information, highlights active actuators and thrusters, shows rotational speed values, and indicates the robot’s orientation and position would enhance understanding and assist operators in debugging when components behave unexpectedly. F3 In known environments, the system can integrate static maps, robot link positions, and dynamic mapping to enhance real-time accuracy for the operator. F4 Operators often need to focus on the display without looking at the control interface. Consistent layouts across interfaces are essential to prevent mistakes. F5 Some buttons provide only constant velocity input, limiting finer control over the system’s movements. Operators prefer proportional control over on-off control. F6 Multi-robot collaboration is currently managed at the operator level. Enabling intelligent communication between robots could reduce the risk of collisions and accelerate the completion of the task. F7 Operators prefer to feel part of the machine, directly interacting with the remote environment through telepresence. Interface designs should expand the range of stimuli channels to enhance this sensation. F8 Current joysticks lack force feedback, but experts acknowledge its potential for collision avoidance. E. Limitations The participant sample size, composed of experienced operators, is limited to 15 participants from two organisations within the same country and does not include operators from commercial reactors, which may impact the generalisability of the findings. Future work should build on the insights of this work by exploring the experiences of operators from a broader range of nuclear facilities. VI. CONCLUSIONS In this study, we investigated the specific needs and requirements of robot operators in nuclear environments, addressing a gap in current research by focusing on the operator’s perspective to determine how to design effective robot teleoperation systems. Through the analysis of three focus groups, comprising 15 experienced operators from the UK Atomic Energy Authority and Sellafield Ltd., we identified three key findings. Firstly, no single robot can address all challenges, necessitating a diverse range of teleoperated robots that may need to collaborate to perform various tasks, but having multiple different control interfaces increases the learning and experience requirements for operators. Secondly, in critical missions, robots are typically operated by teams rather than individuals, so designers must account for this teambased dynamic rather than designing interfaces for a single operator. Finally, since operator errors pose the greatest and most unpredictable risk in robotic operations, safety and reliability must be prioritised as the foremost features of the human-robot interface. We present a comprehensive list of robot-specific and overall user requirements and discuss how these insights can be translated into technical specifications for teleoperated systems. The contributions of this research are valuable for both the robotics and nuclear communities, offering a concrete framework to guide the design of future teleoperated robots and improve both user satisfaction and operational effectiveness in hazardous environments. ACKNOWLEDGEMENT This work was supported by the European Commission’s Marie Skłodowska-Curie Actions (MSCA) Project RAICAM (GA 101072634), and by UK Research and Innovation (UKRI) under grant number EP/X025977/1. Additional support was provided by the UKAEA/EPSRC Fusion Grant 2022/27 (EP/W006839/1), which enabled the utilisation of related work for the decommissioning of fusion devices. The authors would like to thank the anonymous operators from UKAEA and Sellafield Ltd. who volunteered to participate in the experiments. REFERENCES [1] I. Vitanov et al., “A suite of robotic solutions for nuclear waste decommissioning,” Robotics, 2021. [2] H. Varjonen, Nuclear Power Plant Outage Optimization Strategy 2016 Edition, 2016. [3] K. Park et al., “Sustainable decommissioning strategies for nuclear power plants: A systematic literature review,” Sustainability, 2022. [4] M. Laraia, “Nuclear decommissioning,” Planning, Execution and International, 2012. [5] I. G. McKinley, H. A. Grogan, and L. E. McKinley, “Fukushima: overview of relevant international experience,” Genshiryoku Backend Kenkyu, 2011. [6] K. Miyahara et al., “Use of knowledge and experience gained from the fukushima daiichi nuclear power station accident to establish the technical basis for strategic off-site response,” JAEA-review, 2015. [7] G. Chapman et al., “Nuclear security in times of crisis,” Centre for Science & Security Studies, 2021. [8] N. P. Moray and B. M. Huey, “Human factors research and nuclear safety.” 1988. [9] A.-M. Teperi et al., “Developing and implementing a human factors (hf) tool to improve safety management in the nuclear industry,” 2018. [10] R. Bogue, “Robots in the nuclear industry: a review of technologies and applications,” Industrial Robot: An International Journal, 2011. [11] R. Smith, E. Cucco, and C. Fairbairn, “Robotic development for the nuclear environment: challenges and strategy,” Robotics, 2020. [12] C. Abras et al., “User-centered design,” Encyclopedia of HumanComputer Interaction, 2004. [13] R. A. Krueger, Focus groups: A practical guide for applied research. Sage publications, 2014. [14] J. White et al., “Evaluation of robotic inspection systems at nuclear power plants,” Remote Technology Corp., Oak Ridge, TN (USA), Tech. Rep., 1984. [15] A. Vale, R. Ventura, and P. Carvalho, “Application of unmanned aerial vehicles for radiological inspection,” Fusion Engineering and Design, 2017. [16] I. Tsitsimpelis et al., “A review of ground-based robotic systems for the characterization of nuclear environments,” Progress in nuclear energy, 2019. [17] M. Nancekievill et al., “Development of a radiological characterization submersible rov for use at fukushima daiichi,” IEEE Transactions on Nuclear Science, 2018. [18] G. S. Sundar, R. Sivaramakrishnan, and S. Venugopal, “Design and developments of inspection robots in nuclear environment: A review,” Int. J. Mech. Eng. Rob. Res, 2012. [19] J. C. Castro et al., “Analysis of the structural girders of a crane for the license renewal of a bwr nuclear power plant,” Procedia Structural Integrity, 2019. [20] H. Roman, “Robots cut risks and costs in nuclear power plants,” IEEE Computer Applications in Power, 1991. [21] T. Ikegami, T. Shimura, and M. Koike, “Plant life management and maintenance technologies for nuclear power plants,” Hitachi Review, 2001. [22] “Iaea safety standards: Decommissioning of facilities. general safety requirements part 6,” 2014. [Online]. Available: https://www-pub. iaea.org/MTCD/Publications/PDF/Pub1652web-83896570.pdf [23] T. Robinson, “The challenges for robotics in nuclear decommissioning,” Engineering & Technology Reference, 2016. [24] U. A. E. Authority, “Robotics challenges for fusion energy,” 2023. [25] R. Smith, E. Cucco, and C. Fairbairn, “Robotic development for the nuclear environment: Challenges and strategy,” Robotics, 2020. [26] R. Buckingham and A. Graham, “Nuclear snake-arm robots,” Industrial Robot: An International Journal, 2012. [27] A. Mazumdar et al., “A compact, maneuverable, underwater robot for direct inspection of nuclear power piping systems,” in ICRA, 2012. [28] Z. Zhonglin et al., “Design and function realization of nuclear power inspection robot system,” Robotica, 2021. [29] B. Bird et al., “Vega—a small, low cost, ground robot for nuclear decommissioning,” Journal of Field Robotics, 2022. [30] E. Krotkov et al., “The darpa robotics challenge finals: Results and perspectives,” The DARPA robotics challenge finals: Humanoid robots to the rescue, 2018. [31] M. Tranzatto et al., “Team cerberus wins the darpa subterranean challenge: Technical overview and lessons learned,” arXiv preprint arXiv:2207.04914, 2022. [32] A. Norton et al., “Analysis of human–robot interaction at the darpa robotics challenge finals,” The International Journal of Robotics Research, 2017. [33] K. Hauser et al., “Analysis and perspectives on the ana avatar xprize competition,” International Journal of Social Robotics, 2024. [34] C. Lenz et al., “Nimbro wins ana avatar xprize immersive telepresence competition: human-centric evaluation and lessons learned,” International Journal of Social Robotics, 2023. [35] M. Giuliani et al., “User-centred design and evaluation of a teleoperated echocardiography robot,” Health and Technology, 2020. [36] G. Adamides et al., “Usability guidelines for the design of robot teleoperation: A taxonomy,” IEEE Transactions on human-machine systems, 2014. [37] C. Chirapornchai et al., “User requirements for a robot teleoperation system for general medical examination,” in RO-MAN, 2022. [38] G. Graf and H. Hussmann, “User requirements for remote teleoperation-based interfaces,” in International Conference on Automotive User Interfaces and Interactive Vehicular Applications, 2020. [39] R. H. Baayen, Word frequency distributions. Springer Science & Business Media, 2012. [40] J. Ritchie and L. Spencer, “Qualitative data analysis for applied policy research,” in Analyzing qualitative data, 2002. [41] V. Braun and V. Clarke, Thematic analysis. American Psychological Association, 2012. [42] R. Skilton et al., “Mascot 6: Achieving high dexterity telemanipulation with a modern architectural design for fusion remote maintenance,” Fusion Engineering and Design, 2018. [43] K. Young et al., “Mission results: Creating a 3d map of a very high radiation confined space using the lidar-equipped elios 3 drone,” Nuclear Science and Engineering, 2024.