Underground mine rescue robotic systems: insights into human-robot information exchange

Abstract

Mine emergencies demand rapid and informed decision-making under extreme conditions, often placing personnel in life-threatening situations. Robotic assistance offers the potential to reduce unnecessary human exposure during such operations. This study examines the specific informational needs and communication preferences of mine rescue personnel for designing robotic systems for underground emergency response. A semi-structured interview was developed and conducted with ten mine rescue personnel and subject matter experts (SMEs). Responses were analyzed using thematic analysis and compared with established cognitive models to derive key design recommendations. Drawing on both field experience and hypothetical rescue scenarios, participants provided insights into key functional aspects of robotic systems, including mapping and navigation, gas detection and environmental monitoring, communication capabilities, system reliability, control, and the robot’s specific roles during operations. The qualitative data was transcribed and analyzed to identify recurring themes and critical user guidelines. The findings revealed insights into the informational and interface recommendations of rescue teams, particularly the need for real-time situational data and customizable human–robot interfaces tailored to emergency scenarios. These results expose key deficiencies in the current human–robot interaction systems and offer actionable guidance for designing robotic technologies that better align with the operational needs of experienced responders. The outcomes of this study can serve as practical guidelines for developing effective interfaces to support underground mine rescue missions.

1 Introduction

Underground mine emergencies, although rare, are high-consequence events in which mine rescue personnel risk their lives by entering dangerous conditions. One of the most critical applications of robotics is in mine search and rescue operations,where it may be too dangerous, or entirely impossible, for humans to intervene (Reddy et al., 2015). However, environmental factors in underground mines (e.g., dust, gas, smoke, and darkness) create a range of challenges for the design and deployment of search and rescue robots. Such challenges include navigation, localization, communication in low-light conditions, power and energy efficiency, maintenance, human-robot interaction (HRI), and strict adherence to safety constraints (Chang et al., 2022).

To motivate this work, it is important to briefly outline how underground mine search and rescue (S&R) operations are currently conducted and where information gaps arise with and without robotic support. Following an explosion, fire, roof fall, or inundation, standard protocols require establishing a fresh-air base and deploying trained mine rescue teams equipped with breathing apparatus, gas detection instruments, and lifelines into areas where conditions are uncertain and potentially lethal (U. S.Department of Labor et al., 2013; DeJonckheere and Vaughn, 2019; Kilinc et al., 2014). In the absence of robotic assistance, any attempt to verify roof stability, gas conditions, fire spread, water accumulation, or survivable routes demands that personnel physically enter contaminated, low-visibility environments with limited time windows imposed by breathing apparatus duration, heat stress, and communication constraints. As a result, command centers must make critical decisions based on intermittent verbal reports, partial maps, and limited environmental data, which makes it difficult to maintain situational awareness beyond the team’s immediate surroundings and to project how conditions may evolve over time. Robotic systems could collect more data by extending sensing and communication deeper into the mine, scouting inaccessible or high-risk areas prior to human entry, and providing safer channels for video, mapping, and gas data transmission (Reddy et al., 2015; Li et al., 2022; Zhai et al., 2020; Wang et al., 2018). At the same time, integrating robots into mine rescue introduces its own challenges, including maintaining reliable communication through rock, maneuvering within constrained geometries, ensuring intrinsic safety, and designing control and visualization interfaces that fit established mine rescue procedures and cognitive demands under stress. These complementary benefits and constraints define the current state of the art and motivate a focused examination of what information mine rescuers need from robots and how they prefer that information to be delivered.

To address these challenges, various approaches have been pursued. Competitions such as DARPA’s Subterranean Challenge (Rouček et al., 2020) and RoboCup Rescue (RoboCup Federation, 2023) have fostered advancements in robotic exploration systems by promoting improvements in capability and reliability for subterranean environments. Additionally, numerous search and rescue robots have been designed and developed specifically for underground mining applications. Notable examples include RATLER (Krotkov et al., 1994), Numbat (Ralston et al., 2001), ANDROS Wolverine (Murphy et al., 2009), Groundhog (Li et al., 2022), Cave Crawler (Zhai et al., 2020), Souryu (Arai et al., 2008), Inuktun VGTV (Casper and Murphy, 2003), WA Water Service Company Robot (Wang et al., 2018), Subterranean Robot, Leader (Ray et al., 2010), Gemini Scout (Zhao et al., 2017), MINBOT II (Wang et al., 2014), CUMT V (Zhao et al., 2017), Mobile Inspection Platform (Novák et al., 2018), and Telerescuer (Moczulski et al., 2014). These examples, spanning development years from 1994 to 2020, demonstrate the broad effort to tackle underground mine rescue challenges.

Despite this extensive progress and the diversity of robotic systems developed, current rescue robots continue to face several significant limitations, including restricted operational range, dependence on tethered communication, maneuverability, uneven floors, loose debris, and narrow passages (Bakzadeh et al., 2023). While overcoming such limitations remains a central focus of ongoing research, these efforts alone are insufficient. The success of any robotic deployment in mine rescue operations hinges not only on functionality but also on alignment with the practical realities, workflows, and decision-making processes of rescue personnel. Thus, the human element must also be carefully considered. In particular, prior knowledge of the experiences, needs, and preferences of specialized mine rescue professionals is critical. These individuals operate under intense pressure and uncertainty, so their insights are essential to inform user-centered design and ensure safe, effective deployment in high-risk scenarios (Reddy et al., 2015). Additionally, design considerations related to data visualization and situational awareness have been underscored. Effective presentation of complex data and sustaining situational awareness for human operators are considered crucial in a user-centered design. In these high-stress, low-visibility settings, rescue teams must interpret vital information in real time. Poor data visualization can impair situational awareness and diminish the overall effectiveness of rescue operations.

This research aims to ensure that robot development aligns with real-world needs by directly involving end-users in the design process, thereby enhancing communication, usability, and the overall effectiveness of robotic systems in disaster response. This research aims to ensure that robot development aligns with real-world needs by directly involving end-users in the design process, thereby enhancing communication, usability, and the overall effectiveness of robotic systems in disaster response. In this paper, the term “human–robot information exchange” refers to both information transmitted from the robot to mine rescue teams (e.g., maps, gas and environmental data, system status, alerts) and information transmitted from humans to the robot (e.g., navigation commands, autonomy boundaries, overrides, and map edits). Accordingly, the study addresses the following research question:

“How do mine rescue personnel want to exchange information and control with a search-and-rescue robot—specifically, what data do they need to receive, what inputs or commands do they wish to provide, and how should these interactions be supported through the human–machine interface?”

To answer this question, we conducted comprehensive, semi-structured interviews with experienced mine rescue personnel to gain a deeper understanding of their expectations, needs, and preferences regarding preferred robotic systems used in underground rescue missions.

Throughout this paper, we assume an operational concept in which trained mine rescue personnel and command-center staff interact directly with the robot’s interface and retain command-and-control authority over the robot’s actions, whereas other underground miners are not expected to operate or configure the system.

2 Background and research gap

Prior research on search-and-rescue (S&R) robots has made great strides in core capabilities (mobility, sensing, autonomy), but often with comparatively limited emphasis on human factors and user-centered interface design. From early systems like Gemini Scout and MSRBOTS (Reddy et al., 2015), which introduced tethered communication and multi-sensor payloads, to the robust autonomous multi-robot teams in the DARPA Subterranean Challenge (Rouček et al., 2020; Kulkarni et al., 2022), researchers have increasingly tackled the mechanical and algorithmic challenges of subterranean robotics. Notable contributions include CUMT-V’s successful field trials in Chinese coal mines (Brown et al., 2025), and data-efficient coordination strategies such as topological mapping under extreme bandwidth constraints (Bayer and Faigl, 2021). Ten-year longitudinal studies (Tardioli et al., 2019) and survey reviews (Bakzadeh et al., 2023) confirm steady progress in hardware and autonomy, yet also reveal that many systems still fall short in addressing human–robot interaction (HRI) and interface usability in real rescue conditions.

This oversight is echoed in broader S&R literature. Shah and Choset (Shah and Choset, 2004) emphasized that rescue robots often become impractical in real-world scenarios unless their control systems reduce operator workload under stress. Liu et al. (Liu et al., 2007) similarly argued that “operability”—ease of use—is just as important as technical performance for rescue robots to be effective. More recently, Balta et al. (Balta et al., 2017) warned that robotic tools deployed without intuitive data management systems can overwhelm human responders, and they stressed the need for presenting mission-critical information in formats that are immediately understandable. These human-factor challenges are even more acute in mining contexts, where environmental stressors like toxic gases, collapse risks, and darkness amplify the importance of effective, trustworthy information exchange. Yet few systems have been designed with miners’ workflows or decision-making needs in mind. Previous researchers have effectively used interviews with end-users to guide the design and functionality of robotic systems across various domains. Previous research has used interviews and thematic analysis to guide the development of user-centered robotic systems across diverse domains. Beer et al. interviewed older adults to inform assistive robot design for aging in place, emphasizing independence and safety (Beer et al., 2012). In industrial settings, interviews with factory workers highlighted the role of collaborative robots in improving workplace safety and job satisfaction (Baumgartner et al., 2022). Studies have also emphasized user involvement in design processes. Interviews with robot designers showed that feedback from target users shaped emotionally effective social robots (Alves-Oliveira et al., 2022), while field interviews in pharmacies led to a prototype tailored for older adults (Chiu et al., 2025). Qualitative and empirical analyses have been widely applied to deepen insights into HRI research. Seyitoğlu and Ivanov (2024) analyzed 252 studies to develop a framework for emotionally intelligent robots. Sørensen et al. (2025) identified user acceptance themes such as utility and autonomy. Interviews with Kinova® Jaco® arm users informed design improvements through themes like training and independence (Styler et al., 2025). Other studies have examined social and healthcare robotics. Søraa et al. (2021) extracted themes on desired roles and interface preferences from interviews with older adults and caregivers. Another study with acute care nursing staff identified deployment challenges and contextual constraints for hospital robots (Ohneberg et al., 2025). These studies have offered valuable insights into the functionality, utilization, and human–robot interactions of robotic systems across diverse applications.

While research and development on mine rescue robot functionality and availability have gained momentum, no studies have explored the specific needs and expectations of mine rescuers regarding these robots, despite the extreme dangers inherent in such missions. This study attempts to fill this gap by conducting qualitative research using a semi-structured interview protocol with experienced mine rescue personnel who had real mine rescue experience and represented a diverse set of companies. Interviews were chosen as the research method because they serve as valuable tools for uncovering the narrative behind a participant’s experiences, gathering detailed insights on a subject, and obtaining unforeseen types of information (DeJonckheere and Vaughn, 2019). This paper addresses that gap by explicitly integrating user feedback and cognitive models into the development of a human–machine interface for mine rescue robotics, aiming to transform the robot from a passive tool into a collaborative cognitive partner.

3 Methodology and methods

3.1 Questionnaire design and development

The development of an effective interview tool for investigating human–robot interaction in underground mine rescue operations followed a structured, three-phase process informed by established field protocols, documented disaster histories, and insights from subject matter experts. In the first phase, the field protocols from official Mine Safety and Health Administration (MSHA) and the National Institute for Occupational Safety and Health (NIOSH) rescue handbooks and training guides were analyzed to understand team roles, data needs, and decision-making processes in high-risk scenarios. Next, all the reports on historic mine disasters were reviewed to identify information gaps where robotics could enhance safety (U. S.Department of Labor et al., 2013). Following the first two steps, initial interview questions were drafted. The questionnaire went through several iterations with expert feedback to improve relevance and technical accuracy. Collaboration with SMEs ensured alignment with research goals and real-world rescue needs. The final questionnaire included both specific functionality questions and general inquiries about the role of robotics in mine rescue missions.

3.2 Participant recruitment

Institutional Review Board (IRB) approval was obtained prior to conducting interviews. Following IRB approval, 10 experts with 10–50 years of experience were selected. The first participant was recruited through convenience sampling, with the rest selected via snowball sampling. In snowball sampling, existing participants recommend additional individuals who meet the inclusion criteria, which is useful for reaching small, specialized populations such as experienced mine rescuers.

Participants were eligible if they met at least one of the following criteria: (i) membership in or leadership of an underground mine rescue team; (ii) responsibility for mine rescue training, emergency-response planning, or mine safety management; or (iii) specialist involvement in the selection, provision, or evaluation of mine rescue equipment and technology. Participants included superintendents, instructors, directors, safety officers, rescue trainers, and engineers from 10 companies. Notably, 7 out of 10 had direct experience with one or more mine emergencies which was a great plus. The remaining participants, while not having entered the mine during an actual disaster, had extensive involvement in mine rescue training, mock drills, and preparedness exercises, and were therefore treated as subject matter experts on how rescue operations are organized, staffed, and equipped In the United States, only a handful of underground mine disasters have occurred in the past decade, with fewer than 300 individuals participating in actual emergency responses. Given the fact that only an estimated 15% of certified mine rescuers have ever been involved in a real-world mine disaster operation, this is considered a rare qualification (Calhoun, 2022; Kilinc et al., 2014). Access to such experienced personnel has become increasingly rare due to the declining number of underground disasters, industry retirements, and stricter safety regulations that have reduced emergency incidents. Within this context, interviewing these individuals with firsthand mine disaster response experience, regardless of when that experience occurred, is highly significant. Table 1 below details about the participants.

Eight interviews were conducted in person and two were conducted virtually. Each interview lasted approximately 1 h and was recorded. Subsequently, the audio files were transcribed. The main emphasis was on tackling the issues concerning the accessibility and functionality of search and rescue robots, which are essential for achieving successful operations. Thus, the question “what essential capabilities do search and rescue teams expect from a robot to guarantee successful missions?” encapsulated the primary issue we were exploring. To ensure a structured and comprehensive exploration of the key considerations for the use of robots in underground mine rescue operations, the interview questions were organized around a set of predefined categories. These categories were developed based on both existing literature and preliminary consultations with SMEs, aiming to cover the full range of technical and operational concerns relevant to mine rescue scenarios. Each category reflects a core functional area in which robotic systems may support or enhance rescue efforts, including mapping, navigation, gas data monitoring, communication, robots' roles, responsibilities, and reliability and control. Within each category, specific sub-categories were identified to focus the discussion on detailed aspects such as data visualization, information update frequency, communication priorities, and interface expectations. This structured approach helped to elicit targeted feedback from participants and facilitated a consistent analysis of the interview data across different thematic areas. The categories of interview questions are shown in Figure 1 below.

FIGURE 1

3.3 Empirical framework for data analysis-thematic analysis

Thematic analysis has its roots in the broader field of qualitative research, with early contributions from researchers like Barney Glaser and Anselm Strauss, who developed Grounded Theory in 1967 (Glaser and Strauss, 2017). Their work on identifying recurring themes in data helped lay the foundation for thematic analysis as a distinct method. In their work, themes were systematically coded and interpreted to generate theoretical concepts, a process that later influenced the formalization of thematic analysis. Over time, this method has evolved to become a flexible and widely used technique in qualitative research, enabling researchers to identify, analyze, and interpret patterns or themes within data (Glaser and Strauss, 2017).

Thematic analysis strengthened the research approach and provided a deeper understanding by allowing a focus on identifying key patterns relevant to answering specific research questions. It goes beyond mere data summarization, as a comprehensive thematic analysis interprets the data and extracts meaningful insights (Maguire and Delahunt, 2017). By using this analytical technique, the transcripts were actively, critically, and analytically engaged with, allowing for contemplation of the significance of the data (Maguire and Delahunt, 2017). This process entailed posing questions such as: how are these experiences interpreted by the individual? What underlying beliefs are held in their understanding of their experiences? What type of reality is uncovered through their narratives?

An inductive thematic analysis was conducted to analyze the qualitative interview data, following the six-phase guide proposed by Braun and Clarke (Braun and Clarke, 2006). This method was chosen for its flexibility and its suitability for exploring under-researched, applied contexts such as mine rescue robotics. The goal of the analysis was to identify patterns of meaning in experts’ expectations, needs, and preferences regarding the type, format, and delivery of information from underground rescue robots. This was conducted within a contextualist epistemological stance, recognizing that participants’ views are shaped both by individual experience and broader organizational or operational contexts. Figure 2 shows the detailed phases of this analysis.

FIGURE 2

3.4 Analytical framework for data analysis - Cognitive models

While the thematic findings offered empirical insights into user priorities, deriving evidence-based recommendations for interface design necessitated interpreting these results within the framework of established cognitive models. Cognitive models are theoretical frameworks that describe how people perceive, process, and respond to information, especially in complex, high-stress environments (Minz, 2024). In emergency response situations, mine rescuers must make rapid decisions based on partial, evolving data, often under pressure, with limited visibility, and high risk. Cognitive models allow us to predict how users will interpret information, where errors or overload may occur, and how interfaces can be designed to align with users’ mental expectations, attention limits, and decision-making processes. By applying these models, one can design robots and interfaces that are not only technically functional but also cognitively supportive, reducing confusion, building trust, and improving mission success.

For this study, four cognitive models were deliberately selected based on their relevance to the cognitive demands of rescue operations. Once the interview transcripts had been coded, themes refined, and subthemes defined, we systematically revisited each theme and its representative participant excerpts to interpret them through the lens of the four selected cognitive models. This was done using concise operational definitions for each model. Situational Awareness (SA) was included to ensure that the perception, comprehension, and projection of mine critical environmental information (Endsley, 1995). Multiple Resource Theory (MRT) was incorporated to guide the distribution of information across sensory channels, thereby reducing cognitive interference during multitasking conditions (Wickens, 2002; Wickens et al., 2021). Cognitive Load Theory (CLT) was applied to limit unnecessary mental effort and improve the clarity of information presentation, enhancing decision-making in high-pressure contexts (Sweller, 1988; Sweller et al., 2011).

Mental Models (MM) were considered to ensure that system behaviors align with user expectations, enabling more intuitive interaction and reducing the risk of error (Norman, 2002). Although additional cognitive frameworks exist, these four were selected due to their established applicability in high-risk, time-sensitive, and interface-driven environments. Collectively, they provide a theoretical foundation for developing a cognitively compatible interface for mine rescue scenarios.

4 Results

4.1 Thematic analysis results

The transcripts were reviewed alongside notes taken during data collection to ensure a deep understanding of the content and context. Initial coding was performed manually and inductively, without applying any pre-existing framework. Initially, the data was independently coded by two researchers, with interpretations drawn based on individual perspectives. When discrepancies arose, discussions were held, and a consensus was ultimately reached on the final set of codes. Seventy-five codes were generated directly from the data, focusing on meaningful features related to robot functionality, situational awareness, data visualization, preferred communication formats, and decision-support needs. Both frequently mentioned items and less common but technically significant insights were coded.

Table 2 shows those emerging themes and their corresponding subthemes.

4.2 Cognitive model integration

The six themes derived from the interviews reveal not only what information mine rescue personnel need from robotic systems, but also how they prefer to receive and interact with that information. To translate these empirical findings into actionable, theoretically grounded design guidance, we interpret them through four established cognitive models: (SA) (Endsley, 1995), (MRT) (Wickens, 2008), (CLT) (Sweller, 1988), and (MM) (Norman, 2002). This analytical step bridges the gap between descriptive results and prescriptive design recommendations. In the following section, we map representative participant quotes to these cognitive frameworks, and in the subsequent discussion, each theme is revisited with its cognitive interpretation to derive targeted, human-centered interface design strategies.

Each excerpt was examined for alignment with these definitions and assigned to the most relevant model(s), with multi-tagging permitted when a statement clearly reflected more than one cognitive mechanism. This mapping did not influence the coding or theme generation stages; rather, it provided a structured, theory-driven interpretation of empirical findings. By situating participant priorities—such as threshold-based gas alerts, layered mapping with editable hazards, and event-triggered notifications—within established cognitive frameworks, we were able to explain why these preferences reduce cognitive load, enhance situational awareness, and improve trust and control in high-stakes rescue environments. This approach strengthened the link between field-derived insights and human factors theory, enabling the derivation of targeted, generalizable design recommendations. Table 3 below presents how the selected cognitive models align with quotes expressed by mine rescue personnel during interviews.

5 Design recommendations

The thematic findings have shown that participants’ perspectives are deeply shaped by operational realities, mental models built from field experience, and cognitive constraints under stress. Using SA, MM, CLT, and MRT as analytical frameworks (technical), the necessity of delivering information in specific formats, rhythms, and structures to support effective decision-making was discussed. We then achieved a carefully crafted synthesis of design considerations and recommendations that should shape the development process and production of human-robot systems for underground mine rescue operations.

From these themes, we derived six distinct design recommendations, organized into 40 individual design recommendations corresponding to the six themes. It is worth mentioning that, in multiple parts of this discussion, some quotes were intentionally repeated. These statements encapsulate key participant concerns that intersect multiple cognitive demands. By examining them through utilized theoretical lenses (i.e., MM, CLT, MRT, and SA), their layered significance becomes clearer. This approach reflects the richness of participants’ perspectives and strengthens the design recommendations for a cognitively sustainable interface.

5.1 Mapping recommendation - Familiarity and functionality

Participants emphasized the need for detailed, editable, and familiar maps that accurately mirror the underground environment in real time. One participant stressed, “I need to know water, gases are a huge thing, low O₂, high CO, and heat sources” (subthemes of ventilation and airflow). Another reinforced, “It should show roof falls, ventilation doors, stoppings, obstacles” (subthemes of hazards and stability; passability and geometry). These concerns appear interlinked: the environmental hazards (gases, water, heat) and physical infrastructure (roof conditions, doors, barricades) together create a complex risk landscape that responders should simultaneously monitor. This holistic situational awareness is critical, as responders rely on MM that seem to be built through personal field experiences to anticipate hazards and navigate safely. The demand for infrastructure information such as “… door [is] open or closed … ” “roof condition,” “broken, sagging, collapsed areas,” and “obstructions on both sides” (subtheme of subthemes of hazards and stability), confirms the mapping system should be cognitively dense yet visually manageable. This emphasis on needing detailed yet manageable representations of diverse elements, ranging from environmental hazards to structural conditions, strongly imply these recommendations. In interface and spatial visualization design, when multiple types of information should be presented simultaneously without overwhelming the user, it is common to organize such data into overlayers. These layers, often toggleable or stacked, allow users to selectively view relevant categories (e.g., gas levels, heat sources, structural integrity) and reduce visual clutter. Thus, although the term “layer” was not used by participants, this interpretation is grounded in their stated needs and aligns with established visualization strategies for maintaining clarity and cognitive efficiency under pressure. This interpretation aligns with the SA framework, which emphasizes that “situation awareness is enhanced when displays provide integrated information that supports comprehension and projection, rather than requiring the operator to integrate large amounts of raw data mentally. “In addition, participants emphasized including familiar elements on maps, such as “safe entry [points]”, “barricades, refuge chambers, and show water lines” (subthemes of people and assets; water and terrain) because these reflect their established mental models formed during their previous mine safety training and operational experience. Thus, when maps mirror the formats, zones, and indicators that participants are trained to recognize, then, they can quickly interpret and act on the information, reducing the need for additional cognitive effort to translate unfamiliar symbols or layouts. This alignment improves the accuracy and speed of decision-making in urgent situations. These recommendations are consistent with human-factors research on display integration and spatial information design. Wickens and Carswell’s Proximity Compatibility Principle (Wickens and Carswell, 1995) suggests that integrating data layers is only beneficial when the user should mentally combine them; otherwise, separating or toggling layers can reduce interference and improve task performance. Endsley and Jones (Endsley, 2011) further emphasize that layered displays can enhance situational awareness rather than overwhelming the operator. Recent cartographic and HCI work also shows that progressive disclosure and uncertainty-aware overlays reduce cognitive load and improve interpretability in mobile or high-stakes map use (Griffin et al., 2024) (Warden et al., 2023).

5.2 Gas and environmental data - Action-oriented, not just informational

Gas data, specifically CH₄, CO, and O₂ (subtheme of gas measurement and priority), were identified by nearly all participants as the highest priority. One put it bluntly: “if you don’t have air quality numbers, the rest of it is irrelevant.” This underscores that air quality is the immediate, non-negotiable safety concern underground, and without it, other information holds little operational value. Another explained the connection between gas and projection: “if I knew I had a fire, I knew I had a high CO … fire gases don’t mix … then I can come sit down with a map in my team and say, here’s what you got.” (subtheme of thermal/fire cues). This reveals how participants use gas readings not only as current data but as a basis for mental simulation and forecasting, enabling informed strategic planning. These previously mentioned statements again demonstrate SA, where gas patterns are not passively consumed but mentally simulated into likely scenarios.

To support this, participants requested that systems show “trigger levels,” “explosive limits on monitor,” and “color codes … red for danger, green for safe.” One participant elaborated, “If it hits certain levels, just send the data back—don’t push useless data.” Thus, clear and threshold-based indicators are essential to enable rapid comprehension and action in high-stress environments, where detailed data parsing is impractical. This aligns with CLT: in high-pressure contexts, unfiltered streams of data increase workload and risk. Interfaces should prioritize gas types based on context, filter background information, and generate alerts only when safety thresholds are crossed. Participants also wanted combinatorial awareness: “gas is one, next would be water issues,” and “the thing we do is turn off all the power on the ground. So, water is built everywhere, travel ways, ventilation [airways]…” This shows the importance of designing multi-variable interpretation, where interfaces should support intersections of gas, water, and ventilation conditions. Auditory alerts also came up frequently: “it could have just a beep to get our attention, but it doesn’t have to be a continuous alarm,” and “use audio beat, color-coded screen, and thermal scan.” These preferences are consistent with MRT, as different sensory channels are used for different urgency levels. Taken together, these preferences point to a need for robot systems interfaces that are threshold-driven, context-aware, and multisensory to enable quicker decisions under pressure.

The proposed use of threshold-based, color-coded hazard indicators aligns with research showing that urgency-based color schemes improve the speed and accuracy of hazard detection under time pressure. Studies in aviation and process control have found that well-designed color coding allows operators to make rapid go/no-go decisions without parsing raw data. Safety standards such as ANSI Z535 also specify consistent use of red, amber, and green to communicate danger, caution, and safe states, supporting the recommended palette for gas monitoring interfaces (Murphy et al., 2009; Friedrich and Vollrath, 2022; Blundell et al., 2020).

5.3 Trust and reliability-transparency, experience, and control

Trust in robotic systems was conditional and shaped by accumulated field experiences. “… once it proves itself, I want to see the trials and make sure it was actually accurate” (subtheme of evidence and credibility), one participant said. Another was more skeptical: “… no, I don’t trust anything electronic. If it [system failure] only happens one percent of the time, that’s not acceptable.” These statements reflect mental model violations, where the system’s behavior did not align with expectations in past missions. These concerns about trust are grounded in miners’ direct or indirect experiences with real-world failures of rescue robots during past disasters. For instance, during the 2006 Sago Mine disaster, the Wolverine V2 robot from CRASAR became inoperable after its tether became entangled and power was lost within the first 40 feet and also, regarding the Wolverine robot, a single operator was unable to track the robot’s location and operate it simultaneously (Murphy et al., 2009). In the 2007 Crandall Canyon collapse (Teaster and Pavlovich, 2008), a robot provided by the DoD’s Robotics Systems Joint Project Office failed due to dust-related sensor interference and unstable terrain (Teaster and Pavlovich, 2008). In Australia’s Beaconsfield Mine collapse (2006), a robot from the Australian Centre for Field Robotics could not be used at all due to extreme spatial constraints and tunnel irregularities (Reddy et al., 2015). Also operating the Numbat was demanding due to poor lighting conditions and unfamiliar terrain, significantly increasing the operator’s workload and led to mission failure in the testing step. These failures violated users' expectations of robot systems’ reliability and resilience in high-stakes environments and reinforced the mental model violations highlighted by participants. As such, the demand for extensive validation before deployment, as one participant put “once it has a history”, reflects not only theoretical concerns but practical, field-based distrust rooted in prior breakdowns. Participants demanded validation and calibration: “you don’t need to verify it - once it has a history, … once it has a successful mission, then the idea is to use it.” This points to a trust trajectory that can be earned through consistent performance and explainable behavior. CLT supports this: systems that obscure their decision-making increase mental workload and reduce trust. Others emphasized human control: “I want to control the robot’s actions unless I’m desperate,” and “data—yes, decision-making—no” (subtheme of control and overrides). Trust was not about full autonomy, but about shared control, where humans always have final authority. An Interface should display robot confidence levels, error logs, and manual override functions to reinforce confidence without removing oversight.

These findings reflect broader HRI literature showing that trust is improved when systems are transparent, align with user MM, and offer clear avenues for human override. Shared MM has been shown to support faster coordination and error recovery, while transparency and explanations help calibrate trust and avoid overreliance. Reviews in HRI highlight that providing evidence of past performance and allowing operator control are key to sustaining trust in autonomous systems (Tabrez et al., 2020; Ezenyilimba et al., 2023; Almasi et al., 2023).

5.4 Update frequency and autonomy-pacing that matches cognitive workflow

When asked about update timing, participants offered rhythms aligned with decision-making cycles: “Every ten to 15 min gives me enough time to think, strategize what I want to do, and then write it up and send it around the table.” Others preferred “every 5 min” for high-risk stages or “every 1000 ft of progress” (subtheme of update cadence). This illustrates that information pacing supports cognitive functions at multiple levels, including situational awareness and the internal coordination of thought within individual rescuers. Thus, this shows the need for user-defined update intervals that match cognitive and operational rhythms which are supported by MRT and SA. Some participants expressed a desire for real-time anomaly detection: “if the robot finds fire, return. If it finds an explosive, return. If it finds a person, return immediately.” This demand for event-based interruption indicates the need for systems that push data based on conditions, not just intervals. Participants also wanted distance-based update options, noting, “tell us what’s there every 10 ft.” Therefore, interfaces should allow users to configure update pacing based on distance, time, or condition triggers, and clearly indicate “time since last update” and “next expected update.” Taken together, these insights point to a need for interfaces that support configurable update logic that balances scheduled intervals with event-driven alerts to match the way rescue teams think, operate and respond underground. The recommendation to combine user-defined update intervals with event-triggered alerts is supported by alarm-management standards (ANSI/ISA-18.2, IEC 62682) and cross-industry human-factors research (International Society of Automation, 2016; Pruitt et al., 2023). These guidelines emphasize prioritizing critical alerts, suppressing non-actionable ones, and tailoring update pacing to operational tempo to reduce alarm fatigue. Applying these principles in mine rescue robotics can ensure that operators receive timely, relevant data without cognitive overload. d5.5 Interface Appearance and Usability - Modular, Hierarchical, and Customizable.

Interface layout was another area of strong consensus. Participants rejected cluttered, flat displays: “all the data in one screen? No—every screen just one data,” and “just show me section by section, every 20 breaks or 1000 ft” (subtheme of layout patterns). Others requested “pop-ups,” “scrollable panels,” or “a home screen that gives it all to you if needed.” Several participants expressed the importance of contextual grouping: “Combine all the data—water, gas, temperature—in one section and flag it red if it’s dangerous” (subthemes of alert cues; data layers). Others requested flexibility: “Some people may like everything at once. I’d prefer to scroll through.” These responses underscore the need for role-based or customizable dashboards, which allow responders to choose between granular or summary views based on their task and stress level.

Research on multimodal alerts demonstrates that combining auditory and visual cues enhances detection rates and decision speed in high-stakes environments. Human-factors studies in driving, aviation, and clinical monitoring show that well-coordinated multisensory warnings reduce missed alerts and speed operator response. Additionally, dashboard usability reviews highlight the importance of customizable, role-based displays to support diverse operator preferences and maintain situational awareness (Almasi et al., 2023; Mehrotra et al., 2022; Burdick et al., 2022; Edworthy et al., 2021).

5.5 Human input and map editing-empowerment and shared cognition

Participants are expected to be able to actively shape the information space. “Let me add it … I think there should be something here … compare [the] new map and [the] real map,” one said. Another emphasized, “if they spotted anything like roof fall, we want to add” (subtheme of editing and annotation). These statements reflect SA and MM adaptation, where users interpret new inputs and update their internal and shared representations accordingly. Another participant reinforced, “yes, we want to update [the] mine map based on [the] development map” (subtheme of reconciliation and versioning). These quotes show the importance of interfaces that allow manual annotations, map corrections, and visual synchronization with new hazards, ensuring that evolving situational knowledge is embedded into shared cognition. Systems that auto-update without user review risk violating mental alignment and may erode trust. Interfaces should allow intuitive map annotation, flagging of hazards, and undo/history features to support reflection and collaboration. Overcomplicated editing systems would reduce use during high-stress operations and risk outdated information persisting. The goal should be to create a shared cognitive environment between user and robot, a digital mirror of the underground space that evolves through joint interaction (subtheme of role and control preference). This analysis reveals that mine rescue personnel are not merely data recipients but strategic interpreters, seeking interfaces that fit their cognitive workflows, prior experiences, and operational goals. From real-time maps to gas visualization, from update pacing to manual override, every interface function should align with cognitive theory and field reality (subtheme of command authority).

Allowing operators to annotate and edit maps is consistent with cognitive systems engineering principles, which emphasize externalizing state information to support shared situational awareness. Editable overlays and change-history features enable teams to reconcile evolving field observations with system data, improving trust and accuracy. Recent HCI studies in cartography also show that user-driven, layer-by-layer construction supports interpretability without increasing workload (Endsley, 2011; Griffin et al., 2024).

A summary of interface design recommendations was also presented in Table 4 below.

These violations lead to mistrust, increased error risk, and loss of situational control—especially in emergencies where every second matters. Designing for search and rescue robotic systems is not just about sensors and autonomy—it is about supporting the human mind in one of the world’s most dangerous environments. Thus, this study offers a foundation for building such systems by placing expert users at the center of every interface decision, as shown in Figure 3.

FIGURE 3

6 Discussion

Despite the valuable insights generated by this study, some limitations should be acknowledged. First, the sample size was relatively small. While this group provided deep and field-grounded expertise, the findings may not fully capture the diversity of perspectives found in other mining regions, operational contexts, or organizational cultures.

Second, the results are context-specific, reflecting the unique cognitive and environmental demands of mine rescue. While some design recommendations may be transferable to other domains involving human-robot interaction in high-risk environments, caution should be exercised in generalizing these findings beyond the immediate context of underground mining emergencies. Furthermore, this study relied on qualitative interviews, focusing on user perceptions, preferences, and retrospective accounts rather than direct observations or empirical testing of interface performance. As such, the cognitive alignment and design recommendations identified here require further validation through applied experiments, simulations, or field trials.

In particular, further research is needed to deepen and broaden the six thematic categories identified here—mapping, gas and environmental data, trust and reliability, update frequency and autonomy, interface appearance, and human input and map editing—across different mine types, organizational structures, and technology levels. The design recommendations proposed in this study should be empirically verified by embedding them in prototype interfaces and evaluating their impact on situational awareness, workload, decision-making quality, and team coordination in realistic simulations, training exercises, and field trials. Such multi-method validation will clarify which recommendations generalize, which require adaptation, and where additional cognitive or domain-specific factors should be incorporated.

Finally, while features like real-time mapping and adaptive alerts are highly desirable, their implementation is limited by technical constraints. Underground environments often lack reliable high-bandwidth communication, making real-time data transfer difficult this is precisely why configurable update intervals become essential, allowing systems to balance data delivery with connectivity constraints.

7 Conclusion

This study offers an in-depth exploration of human-robot information exchange in underground mine rescue operations. Grounded in structured thematic analysis and informed by cognitive models, the findings move beyond general recommendations to offer targeted and field-validated design strategies. Through interviews with experienced mine rescue personnel, several core themes emerged: the need for real-time maps, gas and hazard alerts that support action—not just awareness, flexible update pacing, intuitive visual layouts, and conditional trust in robotic systems and users’ input. These insights reflect not only practical concerns in the field but also deeper cognitive demands for filtering, focus, and control under extreme pressure.

Participants consistently stressed that interfaces should function as cognitive partners, not passive data screens. Information should be filtered and contextually prioritized to support decision-making. Systems should highlight danger thresholds, synchronize with operational rhythm, and visualize spatial and environmental risks in formats responders already understand. This implies a design strategy rooted in modular displays, editable map layers, and multimodal alerts (e.g., sound + color), all tailored to reduce mental workload and improve comprehension.

Trust and autonomy should be earned, not assumed. Users expressed a strong desire to retain control, verify information, and contribute field knowledge through manual edits and annotations. Interfaces that prevent user input or fail to reflect evolving field conditions risk becoming irrelevant or even hazardous. In conclusion, this study provides a practical framework for designing cognitively aligned human-robot interfaces. By incorporating shared cognition, flexible control, familiar visual elements, and alert prioritization—while avoiding complexity, rigidity, and automation-overreach—developers can create systems that potentially enhance mine rescue effectiveness, safety, and human-robot synergy when it matters most.

Future research should focus on implementing the identified design recommendations into a functional prototype interface tailored for underground mine rescue operations. The next step should involve usability testing with mine rescue personnel under simulated and controlled conditions. These tests would help evaluate the effectiveness of features such as layered maps, danger-triggered alerts, flexible update pacing, and manual override controls. Iterative improvements should be driven by expert feedback, with particular attention to optimizing real-time data access, reducing cognitive load, and improving interface clarity. Such evaluations will be essential for developing user-centered, cognitively aligned robotic tools that enhance situational awareness, decrease workload, improve decision-making, and operational trust in high-stress rescue environments.

References

• 1

  Almasi S. Bahaadinbeigy K. Ahmadi H. Sohrabei S. Rabiei R. (2023). Usability evaluation of dashboards: a systematic literature review of tools. Biomed. Res. Int.2023, 9990933. 10.1155/2023/9990933

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 2

  Alves-Oliveira P. Orr A. Björling E. A. Cakmak M. (2022). Connecting the dots of social robot design from interviews with robot creators. Front. Robot. AI9, 720799. 10.3389/frobt.2022.720799

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 3

  Arai M. Tanaka Y. Hirose S. Kuwahara H. Tsukui S. (2008). Development of “Souryu-IV” and “Souryu-V:” serially connected crawler vehicles for in-rubble searching operations. J. Field Robot.25, 31–65. 10.1002/rob.20229

  • CrossRef
  • Google Scholar

• 4

  Bakzadeh R. Androulakis V. Khaniani H. Shao S. Hassanalian M. Roghanchi P. (2023). Robots in mine search and rescue operations: a review of platforms and design requirements.

  • Google Scholar

• 5

  Balta H. Bedkowski J. Govindaraj S. Majek K. Musialik P. Serrano D. et al (2017). Integrated data management for a fleet of search-and-rescue robots. J. Field Robotics34 (3), 539–582. 10.1002/rob.21651

  • CrossRef
  • Google Scholar

• 6

  Baumgartner M. Kopp T. Kinkel S. (2022). Analysing factory workers’ acceptance of collaborative robots: a web-based tool for company representatives. Electronics11 (1), 145. 10.3390/electronics11010145

  • CrossRef
  • Google Scholar

• 7

  Bayer J. Faigl J. (2021). Decentralized topological mapping for multi-robot autonomous exploration under low-bandwidth communication, 1–7.

  • Google Scholar

• 8

  Beer J. M. Smarr C. A. Chen T. L. Prakash A. Mitzner T. L. Kemp C. C. et al (2012). The domesticated robot: Design guidelines for assisting older adults to age in place. Proc. ACM SIGCHI2012, 335–342. 10.1145/2157689.2157806

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 9

  Blundell J. Scott S. Harris D. Huddlestone J. Richards D. (2020). With flying colours: pilot performance with colour-coded head-up flight symbology. Displays61, 101932. 10.1016/j.displa.2019.101932

  • CrossRef
  • Google Scholar

• 10

  Braun V. Clarke V. (2006). Using thematic analysis in psychology. Qual. Res. Psychol.3 (2), 77–101. 10.1191/1478088706qp063oa

  • CrossRef
  • Google Scholar

• 11

  Brown L. Clarke R. Akbari A. Jovan F. Marafie J. Marjanovic O. et al (2025). Subterranean mine inspection using the prometheus re‐configurable drone. J. Field Robotics. 10.1002/rob.70112

  • CrossRef
  • Google Scholar

• 12

  Burdick K. J. Gupta M. Sangari A. Schlesinger J. J. (2022). Improved patient monitoring with a novel multisensory smartwatch application. J. Med. Syst.46 (12), 83. 10.1007/s10916-022-01869-1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 13

  Calhoun M. (2022). Mine rescue teams are critical to mSHA’s mission. US Department of Labor Blog. Available online at: https://blog.dol.gov/2022/10/26/mine-rescue-teams-are-critical-to-mshas-mission.

  • Google Scholar

• 14

  Casper J. Murphy R. (2003). Human-robot interactions during the robot-assisted urban search and rescue response at the world trade center. IEEE Syst. Man, Cybern. Soc.33, 367–385. 10.1109/TSMCB.2003.811794

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 15

  Chang Y. Ebadi K. Denniston C. Ginting M. Rosinol A. Reinke A. et al (2022). LAMP 2.0: a robust multi-robot SLAM system for operation in challenging large-scale underground environments. IEEE Robotics Automation Lett.7, 1–8. 10.1109/lra.2022.3191204

  • CrossRef
  • Google Scholar

• 16

  Chiu C. J. Hua L. C. Chiang J. H. Chou C. Y. (2025). User-centered prototype design of a health care robot for treating type 2 diabetes in the community pharmacy: development and usability study. JMIR Hum. Factors12, e48226. 10.2196/48226

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 17

  DeJonckheere M. Vaughn L. (2019). Semistructured interviewing in primary care research: a balance of relationship and rigour. Fam. Medicine Community Health7, e000057. 10.1136/fmch-2018-000057

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 18

  Edworthy J. Edworthy J. Reed D. Wessel C. Lawrence L. (2021). Ergonomic auditory alarm signals for the oil and chemical processing industry. IISE Trans. Occup. Ergonomics Hum. Factors9 (2), 86–95. 10.1080/24725838.2021.2012730

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 19

  Endsley M. (1995). Toward a theory of situation awareness in dynamic systems. Hum. Factors37, 32–64. 10.1518/001872095779049543

  • CrossRef
  • Google Scholar

• 20

  Endsley M. R. (2011). Designing for situation awareness: an approach to user-centered design. Second Edition. Boca Raton: CRC Press, Inc.10.1201/b11371

  • CrossRef
  • Google Scholar

• 21

  Ezenyilimba A. Wong M. Hehr A. Demir M. Wolff A. Chiou E. et al (2023). Impact of transparency and explanations on trust and situation awareness in human–robot teams. J. Cognitive Eng. Decis. Mak.17 (1), 75–93. 10.1177/15553434221136358

  • CrossRef
  • Google Scholar

• 22

  Friedrich M. Vollrath M. (2022). Urgency-based color coding to support visual search in displays for supervisory control of multiple unmanned aircraft systems. Displays74, 102185. 10.1016/j.displa.2022.102185

  • CrossRef
  • Google Scholar

• 23

  Glaser B. G. Strauss A. L. (2017). The discovery of grounded theory. New York: Routledge. 10.4324/9780203793206

  • CrossRef
  • Google Scholar

• 24

  Griffin A. L. Reichenbacher T. Liao H. Wang W. Cao Y. (2024). Cognitive issues of mobile map design and use. J. Locat. Based Serv.18 (4), 350–380. 10.1080/17489725.2024.2371288

  • CrossRef
  • Google Scholar

• 25

  International Society of Automation (2016). Management of alarm systems for the process industries. Research Triangle Park, NC: International Society of Automation.

  • Google Scholar

• 26

  Kilinc F. S. Monaghan W. D. Powell J. B. (2014). A review of mine rescue ensembles for underground coal mining in the United States. J. Eng. Fibers Fabr.9, 174–185. 10.1177/155892501400900120

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 27

  Krotkov E. Bares J. Katragadda L. Simmons R. Whittaker R. (1994). Lunar rover technology demonstrations with dante and ratler. JPL, third international symposium on artificial intelligence, robotics, and automation for space.

  • Google Scholar

• 28

  Kulkarni M. Dharmadhikari M. Tranzatto M. Zimmermann S. Reijgwart V. De Petris P. et al (2022). Autonomous teamed exploration of subterranean environments using legged and aerial robots. 3306-13.

  • Google Scholar

• 29

  Li M. Zhu H. Tang C. You S. Li Y. (2022). “Coal mine rescue robots: development, applications and lessons learned,” in Proceedings of 2021 International Conference on Autonomous Unmanned Systems (ICAUS 2021). Lecture Notes in Electrical Engineering, 2127–2138.

  • Google Scholar

• 30

  Liu J. Wang Y. Li B. Ma S. (2007). Current research, key performances and future development of search and rescue robots. Front. Mech. Eng. China2 (4), 404–416. 10.1007/s11465-007-0070-2

  • CrossRef
  • Google Scholar

• 31

  Maguire M. Delahunt B. (2017). Doing a thematic analysis: a practical, step-by-step guide for learning and teaching scholars. AISHE9 (3).

  • Google Scholar

• 32

  Mehrotra S. Wang M. Wong N. Parker J. Roberts S. C. Kim W. et al (2022). Human-machine interfaces and vehicle automation: a review of the literature and recommendations for system design, feedback, and alerts. Washington, D.C.: AAA Foundation for Traffic Safety.

  • Google Scholar

• 33

  Minz I. M. (2024). Exploring cognitive models in human-robot interaction. Int. J. For Multidiscip. Res.6 (5). 10.36948/ijfmr.2024.v06i05.28450

  • CrossRef
  • Google Scholar

• 34

  Moczulski W. Cyran K. Novak P. Rodriguez A. Januszka M. (2014). TeleRescuer-a concept of a system for teleimmersion of a rescuer to areas of coal mines affected by catastrophes. Mechatronics Syst. Veh. Constr. Mach.

  • Google Scholar

• 35

  Murphy R. Kravitz J. Stover S. Shoureshi R. (2009). Mobile robots in mine rescue and recovery. IEEE Robotics and Automation Mag.16 (2), 91–103. 10.1109/mra.2009.932521

  • CrossRef
  • Google Scholar

• 36

  Norman D. A. (2002). The design of everyday things. Basic Books, Inc.

  • Google Scholar

• 37

  Novák P. Kot T. Babjak J. Konečný Z. Moczulski W. Rodriguez López Á. (2018). Implementation of explosion safety regulations in design of a Mobile robot for coal mines. Appl. Sci.8 (11), 2300. 10.3390/app8112300

  • CrossRef
  • Google Scholar

• 38

  Ohneberg C. Stöbich N. Eberl I. (2025). Possible applications and conditions for the development and integration of robotic assistance systems in the acute inpatient setting from the perspective of nursing professionals: a qualitative study. Int. J. Soc. Robotics. 10.1007/s12369-025-01289-4

  • CrossRef
  • Google Scholar

• 39

  Pruitt Z. Bocknek L. Busog D.-N. Spaar P. Milicia A. Howe J. et al (2023). Informing healthcare alarm design and use: a human factors cross-industry perspective. Patient Saf., 6–14. 10.33940/med/2023.3.1

  • CrossRef
  • Google Scholar

• 40

  Ralston J. C. Hainsworth D. W. Reid D. C. Anderson D. L. McPhee R. J. (2001). Recent advances in remote coal mining machine sensing, guidance, and teleoperation. Robotica19 (5), 513–526. 10.1017/s0263574701003447

  • CrossRef
  • Google Scholar

• 41

  Ray N. Dalui R. Maity A. Majumder S. (2010). Sub-terranean robot: a challenge for the Indian coal mines. Online J. Electron. Electr. Eng. (OJEEE).

  • Google Scholar

• 42

  Reddy A. H. Kalyan B. Murthy C. S. N. (2015). Mine rescue robot system - a review. Proced. Earth Plan. S. C.11, 457–462. 10.1016/j.proeps.2015.06.045

  • CrossRef
  • Google Scholar

• 43

  RoboCup Federation (2023). RoboCupRescue. Available online at: https://www.robocup.org/domains/2 (Accessed date September 10, 2025).

  • Google Scholar

• 44

  Rouček T. Pecka M. Čížek P. Petříček T. Bayer J. Šalanský V. et al (2020). DARPA subterranean challenge: multi-robotic exploration of underground environments. Modelling and simulation for autonomous systems. Lect. Notes Comput. Sci., 274–290. 10.1007/978-3-030-43890-6_22

  • CrossRef
  • Google Scholar

• 45

  Seyitoğlu F. Ivanov S. (2024). Robots and emotional intelligence: a thematic analysis. Technol. Soc.77, 102512. 10.1016/j.techsoc.2024.102512

  • CrossRef
  • Google Scholar

• 46

  Shah B. M. Choset H. (2004). Survey on urban search and rescue robots. J. Robotics Soc. Jpn.22, 582–586. 10.7210/jrsj.22.582

  • CrossRef
  • Google Scholar

• 47

  Søraa R. A. Nyvoll P. Tøndel G. Fosch-Villaronga E. Serrano J. A. (2021). The social dimension of domesticating technology: interactions between older adults, caregivers, and robots in the home. Technol. Forecast. Soc. Change167, 120678. 10.1016/j.techfore.2021.120678

  • CrossRef
  • Google Scholar

• 48

  Sørensen L. Sagen Johannesen D. T. Melkas H. Johnsen H. M. (2025). User acceptance of a home robotic assistant for individuals with physical disabilities: explorative qualitative study. JMIR Rehabil. Assist. Technol.12, e63641. 10.2196/63641

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 49

  Styler B. K. Wang E. Ding D. (2025). Qualitative perspectives from kinova(®) jaco(®) robotic arm owners: understanding daily device usage. Disabil. Rehabil.47, 1–12. 10.1080/09638288.2025.2466725

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 50

  Sweller J. (1988). Cognitive load during problem solving: effects on learning. Cognitive Sci.12 (2), 257–285. 10.1207/s15516709cog1202_4

  • CrossRef
  • Google Scholar

• 51

  Sweller J. Ayres P. Kalyuga S. (2011). Cognitive load theory.

  • Google Scholar

• 52

  Tabrez A. Luebbers M. Hayes B. (2020). A survey of mental modeling techniques in human–robot teaming. Curr. Robot. Rep.1 (4), 259–267. 10.1007/s43154-020-00019-0

  • CrossRef
  • Google Scholar

• 53

  Tardioli D. Riazuelo L. Sicignano D. Rizzo C. Lera F. Villarroel J. et al (2019). Ground robotics in tunnels: keys and lessons learned after 10 years of research and experiments. J. Field Robotics36, 1074–1101. 10.1002/rob.21871

  • CrossRef
  • Google Scholar

• 54

  Teaster E. C. Jr. Pavlovich J. W. (2008). Independent review of mSHA’s actions at crandall canyon mine, genwal resources, incorporated, huntington, emery county, Utah. Washington, DC: United States Department of Labor.

  • Google Scholar

• 55

  U. S.Department of Labor, Mine Safety Health Administration, National Mine Health Safety Academy (2013). Advanced mine rescue training: Coal mines. U.S. department of labor, mine safety and health administration. National Mine Health and Safety Academy.

  • Google Scholar

• 56

  Wang W. Dong W. Su Y. Wu D. Du Z. (2014). Development of search-and-rescue robots for underground coal mine applications. J. Field Robotics31, 386–407. 10.1002/rob.21501

  • CrossRef
  • Google Scholar

• 57

  Wang Y. Tian P. Zhou Y. Chen Q. (2018). The encountered problems and solutions in the development of coal mine rescue robot. J. Robotics2018 (1), 8471503–8471511. 10.1155/2018/8471503

  • CrossRef
  • Google Scholar

• 58

  Warden A. C. Wickens C. D. Clegg B. A. Ortega F. R. (2023). Effects of color commonality of overlay clutter and information access effort on tasks requiring visual search. Proc. Hum. Factors Ergonomics Soc. Annu. Meet.67 (1), 650–656. 10.1177/21695067231192615

  • CrossRef
  • Google Scholar

• 59

  Wickens C. (2002). Multiple resources and performance prediction. Theor. Issues Ergonomic Sci.3, 159–177. 10.1080/14639220210123806

  • CrossRef
  • Google Scholar

• 60

  Wickens C. D. (2008). Multiple resources and mental workload. Hum. Factors50 (3), 449–455. 10.1518/001872008X288394

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 61

  Wickens C. D. Carswell C. M. (1995). The proximity compatibility principle: its psychological foundation and relevance to display design. Hum. Factors37 (3), 473–494. 10.1518/001872095779049408

  • CrossRef
  • Google Scholar

• 62

  Wickens C. D. Helton W. S. Hollands J. G. Banbury S. (2021). Engineering psychology and human performance. 5th ed. Routledge.

  • Google Scholar

• 63

  Zhai G. Zhang W. Hu W. Ji Z. (2020). Coal mine rescue robots based on binocular vision: a review of the state of the art. IEEE Access8, 130561–130575. 10.1109/access.2020.3009387

  • CrossRef
  • Google Scholar

• 64

  Zhao J. Gao J. Zhao F. Liu Y. (2017). A search-and-rescue robot system for remotely sensing the underground coal mine environment. Sensors17 (10). 10.3390/s17102426

  • Pubmed Abstract
  • CrossRef
  • Google Scholar