Situational Awareness Tool for Emergency Operators in the Field

Abstract

This paper presents a mobile application designed to support emergency operators during indoor missions, where Global Positioning System (GPS) often fails. The system combines a wearable waist-mounted Inertial Measurement Unit (IMU) with a network of pre-installed Radio Frequency Identification (RFID) tags, enabling robust, real-time geo-referenced tracking of both personnel and critical Points of Interest (PoIs), such as resources and threats. Development was guided by interviews and surveys with emergency professionals, ensuring the tool addresses real operational needs. Key features include dynamic updates of operator positions and nearby hazards, enabled by an Indoor Positioning System (IPS) that fuses IMU and RFID data to improve accuracy in position and heading estimation. The application also offers a user-friendly Human–Environment Interface (HEI) displaying information on a spatially referenced map. By merging advanced technology with expert feedback, this system enhances safety and coordination in critical scenarios, offering a promising solution for indoor navigation and Situational Awareness (SA) in emergency response.

1. Introduction

Regardless of the origin of the emergency, a fast and effective response not only can dramatically reduce losses but may also keep the accident from becoming a large-scale phenomenon [1]. Managing emergencies in large and complex indoor environments poses unique challenges, as responders need to know their own locations, possible escaping routes, and potential risks (industrial machineries, hazardous materials, etc.) since their and rescued people’s life and safety depend on those Points of Interest (PoIs) [2,3,4,5]. Accurate localization and risk awareness enable rescuers to be better coordinated, guided, and informed, reducing the likelihood of disorientation and improving victim localization and rescue outcomes.

For these reasons, Situational Awareness (SA) is a critical tool for supporting rescue teams, particularly when operating in unfamiliar or high-risk environments. SA entails a comprehensive understanding of what is happening in a specific environment, as well as knowledge about the threats and available resources in the surrounding area [6,7,8]. For instance, during a fire in a densely packed industrial setting (e.g., refineries, oil depots, or petrochemical plants), such as the infamous Seveso (Monza, Italy) disaster [9], internal emergency squads face adverse conditions including limited visibility due to smoke, extreme temperatures, and physical obstructions. This is especially true for external teams that have to operate in an unknown and very critical environment. In such scenarios, early identification of emergency equipment (e.g., fire extinguishers, hydrants, or cabinets) and actuators to halt hazardous material leaks is crucial to preventing escalation into major disasters. As noted in [10,11], beyond the technical challenges of developing SA tools, a fundamental aspect is the nature of the information required by the on-field operator during the emergency.

While most of the existing literature focuses on interface design for remote team coordination [12,13,14,15], which typically manages emergency operators from outside the operational perimeter, tools tailored to the specific needs of field personnel are rarely explored. This gap is particularly significant, as on-site emergency responders must prioritize their surroundings and require information that is both rapidly accessible and presented in an intuitive format. Geo-referenced data, for instance, play a pivotal role in aiding navigation during evacuations and guiding movement toward PoIs [16]. However, such data must be processed and visualized in real time to ensure their relevance and usability under critical conditions. Moreover, reliance on cloud-based solutions often proves impractical in emergency contexts due to unreliable connectivity and power supply. These considerations underscore the need for dedicated tools designed specifically for on-field operators, which can enhance SA and improve operational efficiency.

1.1. Related Work on SA Tools

Diverse SA tools have been developed to address a wide range of needs, from managing large-scale public safety incidents to enhancing coordination among field teams and supporting health-specific emergencies. In the realm of public safety and large-scale disaster response, tools like the Community Seismic Network provide essential situational data by utilizing a network of low-cost sensors to monitor seismic activity and generate comprehensive reports during earthquakes. These systems demonstrate the value of real-time data aggregation in facilitating rapid and accurate disaster responses [17]. Similarly, social media platforms have become a critical resource for SA, as exemplified by the Emergency SA-Automated Web Text Mining (ESA-AWTM) system. This tool analyzes Twitter data in real time to detect, assess, and report emergency incidents, offering actionable insights to crisis managers [12]. Moreover, public engagement and education also remain essential components of SA. The Mobile Applications for Public Safety (MAPS) program, launched in 2014 by the US Homeland Security Department [18], aims to enhance public awareness and support the development of user-friendly mobile applications for first responders. This initiative highlights the importance of equipping both professionals and the general public with tools that increase preparedness and operational efficiency, particularly in high-stakes environments like firefighting. In line with this vision, FirstNet has emerged as a transformative platform, leveraging Location-Based Services (LBS) to provide real-time geographic and spatial data. With advanced capabilities such as Z-Axis tracking, FirstNet enables vertical location mapping, offering a new level of indoor spatial awareness critical for complex environments. Tools like Intrepid Networks’ Response for FirstNet allow responders to visualize personnel locations in real time, ensuring improved safety, accountability, and coordination in diverse operational scenarios [19]. For field operations and coordination, several advanced tools have been developed to enhance the efficiency and safety of emergency response teams. Building on this, the RescueAid platform integrates smartphone capabilities such as Global Positioning System (GPS), cameras, and machine learning to offer spatiotemporal data and real-time tagging. These tools empower both commanders and field operators to make informed decisions quickly, even in dynamic and complex scenarios [20]. Enhancements in visualization technology further support decision-making processes. The Mobile Visual Analytics Tool enables 2D/3D visualization, multimedia playback, and retrospective analysis, while augmented reality solutions like the Mobile AR Application for Emergency Response provide responders with interactive 3D building layouts and geographic data for streamlined navigation and evacuation efforts [21,22]. Health-focused SA tools address the unique challenges of medical emergencies. The Pulse Point application exemplifies this category, alerting nearby individuals to cardiac arrests and guiding them to the closest registered automated external defibrillators. This approach enables rapid interventions that can significantly improve survival rates. Similarly, tools designed for medical dispatch and field triage facilitate seamless communication among healthcare teams, ensuring that patient data and critical updates are efficiently managed during emergencies [23,24].

1.2. Related Work on Indoor Positioning System

Properly localized and well-informed rescuers can be better coordinated, commanded, and guided, reducing the chance of disorientation and failure in victims’ localization. Over the years, significant research has focused on these challenges; however, off-the-shelf solutions tailored to comprehensively address such issues are still lacking. In the field of Indoor Localization Systems, numerous technologies and strategies have been explored, each offering distinct advantages for indoor location sensing depending on the specific use case. Technologies with shorter ranges, such as infrared (IR) and magnetic technology, provide high precision for close-proximity applications. Ultrasound and Bluetooth extend the range slightly, making them suitable for room-level localization. Recent innovations have further expanded the scope of indoor localization, employing Bluetooth Low Energy (BLE) and location beacons to provide room-level accuracy even in large public buildings [25]. For fire emergencies, an iBeacon-based Indoor Positioning System (IPS) has been developed to locate firefighters and trapped individuals in real time as shown by [3]. The system integrates BLE devices with fire-retardant materials to withstand challenging environments, offering an accuracy of approximately 3 m and enhancing operational safety. In [26], a beacon deployment algorithm was proposed to enhance room-level localization accuracy while minimizing the effort required to deploy the beacon network. Further advancing the state of the art, a system combining 3D visual reconstruction and Simultaneous Localization And Mapping (SLAM) has been proposed [4]. This approach integrates detailed 3D building maps with Inertial Measurement Units (IMUs) to achieve sub-meter accuracy. Lightweight and cost-effective, this system provides a comprehensive spatial understanding of complex multi-floor environments, aiding first responders in challenging conditions. Furthermore, in caregiving and emergency scenarios, wearable platforms integrating BLE detected from IMUs to monitor individuals in indoor and outdoor environments proved to be effective in emergencies where patients might wander off and require quick assistance [27]. On larger scales, Radio Frequency Identification (RFID) and Wireless Local Area Network (WLAN) technologies provide a balance between range and accuracy, while Ultra-Wideband (UWB) offers a longer range with high precision, enabling effective localization over larger indoor spaces. For instance, a UWB-based system, such as the one proposed in [28], utilized a Time Difference of Arrival (TDOA) algorithm for 3D location estimation.

While the system demonstrated accuracy ranging from 1 to 2 m, its reliance on deploying a large number of sensors posed a limitation. Additionally, it could not track building occupants unless they carried specialized mobile units as part of the sensing network. In [29], a hybrid indoor–outdoor localization system was introduced, where radio nodes were attached to doctors and patients during emergencies. The system employed Monte Carlo and Unscented Kalman Filter techniques, achieving accuracy levels between 5 and 10 m through simulations. Another approach in [30] presented an RFID-based framework to localize first responders. Unlike traditional systems that utilize fixed readers in known locations to track moving nodes, this framework employed fixed nodes with attached readers for tracking occupants. However, no prototype was built, and the system remained unevaluated. A more integrated approach was explored in [31], which reviewed various indoor localization technologies and proposed a system combining foot-mounted IMU sensors and UWB sensors to support first responders. Field tests indicated an accuracy of 1 to 4 m, though heading drifts were observed, increasing with travel distance. Further advancing this field, the Geospatial Location Accountability and Navigation System for Emergency Responders (GLANSER), developed under sponsorship from the Department of Homeland Security, utilized a wearable electronic unit incorporating GPS, an IMU, UWB, Doppler radar, magnetometer, compass, pedometer, and altimeter. While the system achieved an accuracy of approximately 3 m in field tests, the specific algorithm details were not disclosed [32]. Building on this landscape, the present work (included in the so called ‘RISING’ project) introduces a pragmatic, infra-light system tailored to deep-indoor operations: a waist-mounted IMU drives a Pedestrian Dead Reckoning (PDR) loop for continuous motion updates, while opportunistic encounters with passive Ultra-High Frequency (UHF) RFID tags both bound drift and deliver on-tag SA metadata (category, type, optional value) for offline operation. The operator interacts through a geo-referenced, glanceable interface that emphasizes low cognitive load and PoI-centered filtering. This article focuses on the operator-facing workflow and the fusion timing as experienced in the interface.

1.3. Paper Contribution and Outline

To improve personnel safety in the working environment in standard and/or emergency operating conditions, the RISING project integrates RFID devices with a waist-mounted IMU. RFID is a low-cost technology that uses communication via radio waves to identify and track objects. Specifically, in an RFID tag, it is possible to store a small quantity of data that can be retrieved on the fly via RFID readers. Passive RFID technology, in particular, does not require an external power source, as the reader provides the necessary energy for data exchange. On the other hand, an IMU includes devices equipped with triaxial accelerometers, gyroscopes, and magnetometers. Unfortunately, many approaches documented in the literature are hampered by the presence of bias and noise, significantly degrading estimation accuracy over time and rendering the system ineffective after approximately 20 min.

Thus, the RISING project combines such technologies to overcome their limits in order to develop a system able to do the following:

(a)

Support emergency operators in the field to improve SA.

(b)

Operate in the absence of any external resources (included power network).

(c)

Help to improve navigation and localization.

In this paper, the proposed framework is illustrated as a Human–Environment Interface (HEI), developed in response to unique requirements identified by security managers and emergency operations experts. To achieve this, a specialized questionnaire was distributed to 26 operators involved in the field. The proposed HEI consists of a mobile application composed of two interacting systems:

(I)

IPS: This hybrid IPS employs a prediction–correction scheme to estimate the user’s pose by integrating inertial signals from the operator with exteroceptive data about the environment stored in RFID tags.

(II)

SA System: It exploits the information stored in pre-deployed RFID tags in order to provide data about particular hazards, threats, and resources in the surrounding area. Additionally, the application can localize such items relative to the operator’s current position, offering a basic yet effective navigation tool.

Beyond combining IMU and RFID, this work contributes a workflow and data model tailored to deep-indoor response: (i) RFID tags encode on-tag SA metadata (category, type, optional value) and the PoI’s relative localization, so that opportunistic encounters both bound drift and inject actionable context on device (offline first); (ii) the fusion loop maintains usefulness under degraded visibility and connectivity, with tag-driven resets of pose and heading; (iii) topology-aware tag planning (choke points, stairwells, hazard doors, resource caches) supports sparse deployments instead of dense beacon grids and facilitates multi-level registration; and (iv) an operator-centered, geo-referenced user interface with layer filtering is grounded in input from 26 practitioners prioritizing nearby PoIs and glanceable visuals. These choices differentiate our approach from prior IMU + RFID hybrids that use tags chiefly as absolute anchors without co-located SA context.

The outline of the paper is as follows: In Section 2, we illustrate the information regarding the system requirements collected within the above-mentioned survey. In Section 3, we provide a detailed overview of the proposed framework by analyzing the position estimation process based on the merging of proprioceptive and exteroceptive signals. In Section 4, we describe the main features of the mobile application realized within this project. Results are collected in Section 5 with the aim of validating the proposed framework with an example of the usage of the application; finally, a discussion and conclusive remarks are collected in Section 6 and Section 7.

2. System Requirements Analysis

To design an HEI able to effectively support emergency operators and first responders in the field, it is mandatory to carefully identify the relevant information that needs to be reported to the user and the most suitable way to present it. To properly gather that information, a survey was administered online to operational personnel (Fire Service and Civil Protection) identified through a formal request to the Italian Firefighting Corp and via partner organizations; participation was voluntary and anonymous, informed consent was obtained, and no personal data were collected. A total of 26 experts completed the survey. The instrument comprised 5-point Likert items on information priorities and presentation modes plus two brief free-text prompts on map symbology and acceptable glance time. The purpose was to identify the most pertinent information to provide and the best way for displaying it on the fly. The questionnaire was divided into three main parts: the first part (questions Q1 to Q4) aimed to assess the importance of collecting information related to the operator’s absolute or relative pose within the surrounding area. The second section (questions Q5 to Q10) explored the relevance of the nature and location of PoIs (i.e., hazards and/or resources) in the surrounding area. The final part (questions Q11 to Q19) focused on the preferred methods for delivering this information to field operators. A complete list of the questions is provided in Appendix A.

By analyzing the results from Q1 to Q4 (see

Table 1. Questionnaire results.

Topic	Results
Absolute/relative pose of the operator and team members within the environment	Be aware about the real-time relative pose of team members was essential for 60% of responders. For 55% of the interviewees, communicating the operator position to the command center was essential. Knowing the operator absolute pose in real time was essential for 35% of the operators. Having the chance to acquire the operator pose in real time on a pseudo map of the environment was essential for 45% of the responders.
Knowledge about the position of potential PoIs in the surrounding area	For 60% of the operators, knowing PoIs’ relative positions with respect to the operator pose was essential. About 40% of interviewees thought that having knowledge about PoIs’ absolute pose in the environment was essential. Knowing PoIs’ locations on a map of the environment was essential for 40% of the operators.
About the distance of the PoIs to be aware of	In total, 52.6% considered essential knowing the positions of the PoIs located in a range of about 16 m with respect to the user position. Information about PoIs’ relative distances when greater than 25 m with respect to the operator was assumed to be quite useless by 36.8% of the respondents.
Preference about technologies for information providing	Acoustic messages through earphones (40%), visual information by means of a tablet (30%), holographic images through smart glasses (15%), smart watches (10%), tactile feedback (5%).
Preference about information visualization	Graphic representation (70%), grid (20%), textual representation (10%).
Preference about PoIs’ descriptions	Icon with short description (70%), short textual description (15%), icon (15%).
Potential hazards, emergency devices, plants, and environmental design elements to be aware	Explosive materials (75%), high-voltage points (45%), and fire spots (60%). Chemical (55%) and bacteriological materials (45%). Emergency exits (80%), hydrants (55%), fire extinguishers (40%), and first-aid kit (40%). Electrical panel (60%) and high-voltage socket (50%). Water plants (40%), central fire detection points (50%), UPS (50%), and stairs (45%). Local heating–cooling systems (35%) and thermal power plants (35%). The opportunity to select on a GUI the PoIs to display and to navigate to was considered essential by 65%.

), it transpired that all the experts were interested in the realization of a tool able to assist the rescuers in the field during their missions. In more detail, all the proposed features were considered useful in order to assist the navigation of the single operator and to coordinate a team of rescuers from a remote control room. The answers to questions from Q5 to Q10 indicated that end users were interested in incorporating an SA module based on PoIs into the system. Specifically, there was significant interest in the ability to provide information about the position of PoIs relative to the rescuer’s location. The report of the answers to Q9 and Q10 is shown in Figure 1a,b. The results demonstrate the critical nature of knowing the PoIs located within approximately 16 m of the first responders. Furthermore, it was found that environmental elements beyond 25 m were considered too distant to be identified.

Concerning the preferred applicable technologies (results summarized in Table 1), the adoption of acoustic messages combined with the use of visual information emerged as one of the most interesting solutions. Despite the high percentage of preferences in the use of acoustic messages (40%) some other operators excluded the use of this kind of solution due to logistic reasons, especially in loud environments. Conversely, operators showed a preference for tablet or PC-based solutions; specifically, these operators valued graphic visual representations of PoIs in the surrounding area, which are associated with icons, names, or detailed descriptions of the environmental elements. Essentially, this approach involves a visual representation where a geo-referenced marker is used to indicate a PoI on a map of the scenario. These results are reviewed in Figure 1c. Additionally, people who answered this survey were asked to specify what kind of point of danger, plant characteristics, or emergency devices were considered the most relevant. Each answer was associated to a rate and results are summarized in Table 1. Focusing on the points of danger, the responders considered explosive materials, flammable materials, voltage elements, and chemicals as the most relevant hazards. Likewise, responders appreciated the visualization of emergency devices such as emergency exits, first-aid kits, hydrants, and fire extinguishers.

Moreover, the visualization of plant characteristics, such as switchboards, high-voltage sockets, and ladders during environmental exploration was deemed useful. Responses to Q18 underscored the necessity to enable or disable the display of certain categories of PoIs on the map, depending on the emergency type. In summary, the results indicate strong interest from end users in a product that combines an IPS with a visual tool for SA. Regarding the IPS requirements, operators require a real-time system that accurately identifies their position on a map of the indoor infrastructure. For SA, the most relevant PoI categories were identified, and user preferences for their visualization were gathered.

3. RISING Architecture

3.1. Project Guidelines

In addition to the results of the questionnaire, a detailed study of the equipment supplied to emergency managers was performed in order to collect guidelines. In this section, we collect all the requirements about the implementation of the proposed system by considering infrastructural, usability, and technological aspects. The framework must be a flexible and scalable cutting-edge solution that allow an increase in worker safety, both in emergencies and normal operating conditions, with a minimum investment. The framework must be able to provide real-time assistance to the operator during first-responder activities by providing an overview of the surrounding environment and by analyzing hazards, threats, and other additional information such as their position.

From an infrastructural point of view, the proposed framework must be easily integrated with the existing equipment of first responder and, at the same time, with the environment. In more detail, considering the operator equipment, the proposed solution must represent an enhancement or be integrated with it without interfering with the usage of other tools. With regards to the integration with the indoor environment, the proposed architecture should be easily installable without requiring infrastructural changes or significant set-up work.

Concerning the usability of the system, there is a requirement for the ability to display pseudo-maps of the environment on smart devices. Providing information about the location of specific resources or hazards helps to enhance the SA of first-aid staff in emergencies and supports specialists during maintenance operations.

From an economic perspective, the system must be cost-effective. Utilizing commercially available mobile devices and low-cost technologies for indoor localization is recommended to avoid the need for expensive, dedicated infrastructure or facilities that are costly to purchase, install, and maintain. IMU sensors and smart devices, which are economical and widely available, often already form part of the existing equipment for workers.

Concerning the technological perspective, the development of solutions utilizing Micro-Electro-Mechanical Systems (MEMSs), such as compasses, accelerometers, and gyroscopes, is favored due to their low cost, compact size, and minimal power requirements. Additionally, there is a focus on developing technologies that provide feedback about the surrounding environment to the operator. This approach aims to communicate critical information without requiring direct visual contact, thereby minimizing the need for close proximity to the inspected area.

Additional requirements include the system being user-friendly without necessitating extensive training or prior specific knowledge. It should be accessible even to personnel who are unfamiliar with the proposed technologies. Given the dynamic nature of environments like construction sites, waste depositories, or chemical plants, another critical requirement is the system’s adaptability. It must be easily updateable to promptly provide operators with current data on new PoIs, including emerging threats and hazards.

3.2. System Overview

The RISING solution consists in a system based on the cooperation of three main actors: the operator, the RFID tags deployed in the scenario, and a mobile application, as depicted in Figure 2. The environment is equipped with RFID tags that not only store data about the PoIs and their locations but also assist in refining the operator’s estimated position.

RFID tags, consisting of a low-cost infrastructure of passive and semi-passive/active types, are pre-installed throughout the indoor environment. They are strategically placed near resources, plants, or key points crucial for emergency operators’ localization, such as emergency lights and signs, and are embedded into emergency lamps to ensure effective coverage and power supply. Each tag contains the following data:

• Tag Id: a unique code to identify the tag;
• Tag absolute position: expressed in Cartesian coordinates relative to an absolute reference frame attached to the environment;
• Tag data for initialization, updates and inspection.

Furthermore, the PoI message descriptors, which comprise only a few bits, encode both the type of PoI and its relative position to the tag’s absolute location. These descriptors are crucial for enhancing SA and improving operational efficiency. The descriptors used inside the Tag data field are:

• Category: specifies the category of the PoI (i.e., hazard, resource, facility, etc.).
• Type: defines the type of PoI, related to its category. For example, for the category Fire Hazard, the type specify the fire hazard, e.g., oxygen fire hazard, methane fire hazard, etc.
• Sub-Type: defines the type of PoI (if necessary).
• Value: defines the range of the PoI (if available), relating to its category, type, and sub-type (when available).
• Localization: Indicates the relative position of the PoI with respect to the tag’s absolute pose within the environment, with this information being presented in Cartesian coordinates.

The installation of RFID tags within the environment requires careful consideration of the trade-off between the number of PoIs and the number of tags. A higher number of tags not only ensures a more uniform distribution of data concerning PoIs but also provides numerous fixed reference points that enhance the accuracy of the operator’s position on the map. On the other hand, a higher number of tags can increase set-up and maintenance cost and may enhance the problem of incoherent data (because the different tags are set via uniform updates). However, a large number of tags can make the system more resilient. Additionally, a greater density of tags can significantly reduce the time required for reading. To optimize the reading process, the full memory of a tag is only read and stored on the mobile device when a new tag is encountered. Subsequently, for tags that have been read previously, the operator needs to read only the Tag Id field, retrieving the rest of the information from local memory. Encountering a new tag slightly delays the reading process by a few seconds, during which there is an opportunity to refine or reset the operator’s position on the map. More details about the reset and refinement functions are discussed in Section 4. Following all these considerations, operators (i.e., firefighters, technical inspectors, and specialized personnel) are equipped with rugged mobile devices, RFID readers, and a waist-mounted IMU. The mobile application installed on these devices is pivotal, functioning as a central hub for collecting data from both the inertial platform and the RFID reader and executing several critical tasks:

(1)

Manage the serial/Bluetooth connections with the inertial platform and the RFID reader.

(2)

Compute the operator position in real time providing an estimation based on the IMU signals and the data acquired from the RFID tags.

(3)

Communicate to the operator useful data for the SA.

4. RISING Mobile Application

The RISING mobile application comprises an IPS and an SA module, designed for deep-indoor response with an infra-light, offline-first workflow. The IPS leverages PDR from IMU signals to provide continuous motion updates, while opportunistic encounters with passive UHF RFID tags both bound drift and deliver on-tag SA metadata (category, type, optional value). In practice, when the wearable reader enters the tag’s main lobe, reads are accepted through a Received Signal Strength Indicator (RSSI) gate with a short dwell (hysteresis); upon a valid read, the user’s absolute position is refreshed according to the tag’s Cartesian coordinates and, when available, the tag orientation provides a heading cue. This enables corrections in motion without a fixed standoff, aligning with the fusion strategies documented in our prior technical publications on the same PDR+IMU–RFID backbone [5,7,8,33].

Because the tags are passive (no power/batteries), deployment scales with decision points (doorways, stairwells, hazard/resource caches) rather than with floor area; commissioning is scan–write and can be rolled out incrementally by zone/floor without backend connectivity. The SA module presents a geo-referenced map centered on the operator, with filterable PoIs and concise iconography to support glanceable, low-cognitive-load operation [12,13,14,15]. In this manuscript, we focused on the operator-facing workflow and fusion timing as experienced in the interface; quantitative accuracy and site-level specifications for the underlying IPS are reported in the cited technical articles and are not re-stated here [5,7,8,33].

4.1. IPS Description

The proposed IPS builds on prior work by the authors; quantitative error analyses and site specifications for the same PDR with IMU–RFID fusion backbone are reported in [5,7,8,33]. As outlined in Algorithm 1, the IPS operates through a continuous prediction–correction loop that updates the operator’s state at each detected step. At each step k, the algorithm maintains the operator state vector (see Equation (1)):

$${\mathbf{x}}_k={\left[\begin{array}{cccc}{p}_{x,k}& p_{y,k}& p_{z,k}& {\theta }_k\end{array}\right]}^{\top },$$

(1)

where $(p_{x,k},p_{y,k},p_{z,k})$ denotes the position and ${\theta }_k$ the heading.

Algorithm 1 IPS:  timing and fusion

Input: ${\mathbf{x}}_0$                                       ▹ initial state 1: while true do 2:     $\theta \leftarrow$ATTITUDE_ $\mathrm{EKK}(\omega ,\mathbf{a},\mathbf{m})$                           ▹  runs at IMU rate 3:     if STEP_DETECTED then 4:         $\mathcal{P}\leftarrow$PATTERN_ID$\left(\mathbf{a}\right)$                             ▹ flat/stairs/elevator 5:         $s_k\leftarrow$STEP_LENGTH$\left(\mathbf{a}\right)$                             ▹ uses Equation (3) 6:         $\mathbf{x}\leftarrow$PDR_PROPAGATE$(\mathbf{x},\theta ,s_k,\mathcal{P})$ 7:     end if 8:     if RFID_AVAILABLE_RSSI_GATED then 9:         $(T_{xyz},T_{\theta })\leftarrow$READ_TAG 10:         $\mathbf{x}\leftarrow$RFID_CORRECT$(\mathbf{x},T_{xyz},T_{\theta })$             ▹ position reset; optional heading realignment 11:     end if 12: end while

Heading is provided continuously by a quaternion-based Extended Kalman Filter (EKF) that fuses gyroscopes $\omega$ (prediction) with accelerometers $\mathbf{a}$ and magnetometers $\mathbf{m}$ (measurement updates). The quaternion representation avoids singularities and supports robust attitude tracking under motion [8,33]; the filter follows the standard Kalman framework [34]. Figure 3 provides a schematic overview of the IPS architecture, showing how the heading EKF interacts with the tracking modules that compose the overall sensory system.

4.1.1. Prediction

During prediction, inertial signals are analyzed to identify walking patterns (flat walk, stairs, elevator) as in [5]. The PDR updates the position at each detected step event using step-length and heading kinematics, as shown in Equation (2):

$${\mathbf{p}}_k={\mathbf{p}}_{k-1}+s_k\left[\begin{array}{c}cos{\overline{\theta }}_k\\ sin{\overline{\theta }}_k\\ \Delta z_k\end{array}\right],$$

(2)

where ${\overline{\theta }}_k$ is the current heading from the attitude EKF, and $\Delta z_k$ depends on the recognized pattern (e.g., stairs or elevator). Between steps, the pose is held constant while uncertainty grows. The step length $s_k$ is computed according to Equation (3):

$$s_k=\beta \sqrt[4]{a_{v,k}^M-a_{v,k}^m},$$

(3)

where $a_{v,k}^M$ and $a_{v,k}^m$ are the maximum and minimum vertical accelerations within step k, and $\beta$ is a user-dependent parameter [35,36]. Because pre-mission calibration is not permitted, $\beta$ is adapted online using tag-to-tag constraints, as described later (Equation (4)).

4.1.2. Correction

RFID encounters are handled opportunistically: when the wearable reader enters the antenna main lobe, reads are accepted through an RSSI gate with a short dwell (hysteresis) to suppress brief fades; no deliberate stop or fixed standoff distance is required. Upon a valid read, the tag provides an absolute Cartesian anchor $(T_x,T_y,T_z)$ and, when available, an orientation cue $T_{\theta }$. The filter always corrects position and covariance with $(T_x,T_y,T_z)$; heading is re-aligned only if $T_{\theta }$ is present, otherwise it remains governed by the attitude EKF. When two consecutive tag encounters, $(T_{xyz}^{\left(i\right)},T_{xyz}^{\left(j\right)})$, are observed, the optimizer updates $\beta$ so that the integrated PDR displacement between the two encounters matches the mapped geodesic from $T^{\left(i\right)}$ to $T^{\left(j\right)}$ within the walkable space, as shown in Equation (4):

$${\beta }_k=arg\underset{\beta }{min}∥\left({\mathbf{p}}_{k-1}+\sum _{t\in (i,j]}\Delta {\mathbf{p}}_t\left(\beta \right)\right)-T_{xy}^{\left(j\right)}∥.$$

(4)

This yields drift control without imposing fixed poses or distances at tags. The overall execution loop combining these prediction updates and opportunistic RFID corrections is summarized in Algorithm 1.

The simulations illustrated in this paper are intended to clarify fusion sequencing and user-facing behavior under controlled perturbations; head-to-head benchmarks against deployed systems and full site parameterizations are reported in the cited previous technical literature [8,33].

4.2. SA System

Thanks to the SA system, the operator is kept informed about the presence of hazards, potential risks, and technical tools in the surrounding area.

The main screen of the application reports a spatially referenced map of the explored scenario. In more detail, as depicted in Figure 4, this consists of an image with an overlaid grid, allowing any image that represents the planimetry of the structure to be used as a map, regardless of image format. During scenario inspections, the map’s position and orientation are continuously updated to align with the operator’s estimated position and heading. Specifically, a central arrow remains fixed at the center of the screen, while the map adjusts in terms of position and orientation. When the user is in the tag radiation area and taps on the read tag button, the device reads the data stored in the tag memory and the map is populated with the markers. As shown in Figure 4, they are anchored to fixed cells according to the localization field of the tag. Actually, the application considers the categories listed in

Table 2. Categories of supported markers.

Category	Marker	Category	Marker
Emergency exit		Biological
Fire hose		Toxic gas
Extinguisher		Flammable gas
Fire ax		Energy
Radioactivity

and the associated graphic solution as required by the end users. At the same time, it is possible to retrieve additional information about the selected PoI by tapping it. As required from the answers to Q18, each operator can customize the application settings to tailor the functionality to their specific interests, selecting which categories are displayed on the map. This customization also helps reduce the reading time of the tag memory, as the application will skip reading additional information about a PoI if the operator deems it irrelevant.

5. Results

This section outlines the main features of the proposed mobile application by analyzing both the indoor localization backbone and the SA functionality. A validation mission was conducted to illustrate how the RISING application communicates navigational information to the operator. During the simulation campaign, a building basement was explored—an ideal deep-indoor setting without GPS, cellular, or Wi-Fi coverage. The floor spanned $19,000{\mathrm{m}}^2$ and was cluttered with hazmat conduits along the ceiling that significantly hindered radio propagation. Chemical and biological materials in the area constituted critical PoIs. The following walkthrough demonstrates interface behavior and fusion timing; quantitative accuracy for the adopted backbone is documented elsewhere [8,33].

Initially, the floor was surveyed to catalog PoIs. The number and placement of RFID tags were then determined to obtain a topology-aware distribution (doorways, stairwells, choke points, hazard/resource caches), maximizing natural encounters along likely routes while keeping infrastructure sparse. Each PoI was associated with the nearest tag. Tags were programmed with on-tag SA metadata using a project protocol tailored to Omni-ID passive UHF tags (CAEN), which provided a 28-word user memory. Because tags are passive (no power or cabling), they could be affixed or embedded in doorframes, luminaires/emergency lights, signage housings, conduits, or under raised-floor tiles; on-metal form factors were used where required, and floor/zone identifiers were encoded to support multi-level deployments. Before starting the session, the operator enabled communications among the tablet, RFID reader, and IMU (Figure 5). The map of the environment was loaded, and a grid cell size was specified for the overlaid reference. Once the map was visible on the main screen (Figure 6), the initial position and heading could be adjusted by dragging and rotating the map; after this, the operator began the mission.

As the operator proceeded along a corridor, turned left, and passed a doorway, the estimated position was shown on screen. When the wearable reader opportunistically detected the RFID tag installed at the doorway, the system automatically refreshed pose and context: the reset was triggered when the reader’s RSSI crossed a calibrated threshold and remained above it for a short dwell (hysteresis); the RSSI trend also yielded a coarse proximity class (near/medium/far) used to weight the update. No deliberate stop or fixed standoff distance was required.

As depicted in Figure 6a,b, upon a successful read, the application resets the predicted position and, when an orientation cue is available, realigns the heading; the accumulated error (blue circle) shrinks, and the step-length parameter $\beta$ is updated. Concurrently, the SA module populates the map with markers for hazards, risks, and emergency tools. This walkthrough illustrates the in-motion correction mechanism (interface behavior and fusion timing), not a fixed-distance operating protocol. Finally, the operator requested additional information about a gas cylinder located in the room behind them: Figure 6c shows the pop-up displayed when tapping the gas marker, with provides details about the presence of propane. This study provides a qualitative site description; future evaluations will include a structured reproducibility log covering overall and corridor-level dimensions (including ceiling heights), wall/door materials and metal density by segment, electromagnetic context (RF clutter and interference sources such as HVAC motors/transformers), full tag coordinates and configuration (including power settings), and sensor health/noise summaries (e.g., IMU bias trends) to support external comparison and replication.

Regarding the operator equipment, a rugged Getac Z710 tablet was used to ensure usability with both clean and greasy gloves. The application user interface (UI) was iterated to support glanceable interaction and minimal cognitive load. The wearable reader, mounted on the left shoulder (see Figure 2), was a CAEN A528 OEM UHF multi-regional compact reader connected to the tablet via Bluetooth. The waist-mounted IMU (STMicroelectronics iNEMO STEVAL-MKI062V2) interfaced with the tablet through a high-speed USB connection.

6. Discussion

Emergency response in large, complex deep-indoor facilities must contend with unreliable GPS and networking, radio-frequency attenuation due to metallic clutter, smoke, heat, and occlusions, while still requiring dependable self-localization and timely awareness of nearby hazards and resources. The proposed system adopts an infra-light configuration in which a waist-mounted IMU drives a PDR loop that provides continuous updates, and opportunistic RFID tag encounters both reset drift and carry SA metadata. This enables offline-first operation with modest infrastructure and cost. In contrast to earlier IMU–RFID hybrids that employ tags primarily as absolute anchors, here, each opportunistic encounter also delivers on-tag SA metadata (category, type, optional value) and relative localization, so that drift bounding and contextualization occur in a single offline-first step. The workflow is deliberately infra-light and visibility-agnostic: the PDR loop bridges gaps between encounters, and tag-driven resets refresh both pose and context without a backend. Accordingly, we do not re-report error metrics in this interface-focused article and refer the readers to [8,33] for experimentally validated accuracy and site-level details of the same backbone. Topology-aware planning (stairwells, choke points, hazard-doorways, resource caches) is used instead of uniform dense grids, prioritizing natural encounters at decision points.

A technical comparison clarifies the operational niche. BLE and iBeacon solutions [3,25,26] provide room-level awareness under dense beaconing but rely on sustained broadcast coverage; here, uninterrupted motion tracking is maintained by the inertial loop, while tag encounters bound drift and supply on-device contextual information. UWB with time difference of arrival [28,31] can reach sub-meter accuracy with dense anchors; in contrast, the present configuration reduces infrastructure and set-up time and remains resilient to partial infrastructure loss, at the expense of absolute accuracy between tag encounters unless tag density is increased. Vision and SLAM pipelines [4] yield rich maps and object-level semantics under adequate visibility; the approach here remains functional in smoke, low light, and occlusions because it does not require persistent visual features, trading dense mapping for robustness. Multi-sensor responder platforms [11,32] achieve strong performance through extensive sensing and supporting infrastructure; the design presented remains wearable with standard gear and emphasizes low cognitive and logistical burden. Methodologically, tag-driven pose resets coupled with online adaptation of the step-length parameter $\beta$ [5,7,8,33,35,36] address heading drift and user-specific variability without dedicated calibration sessions. Overall, the distinctive contribution lies in coupling on-tag SA metadata with opportunistic tag-driven resets within an offline-first workflow, complemented by topology-aware placement and an operator-centered, glanceable user interface. These choices target deep-indoor constraints (limited visibility, partial infrastructure, low interaction budget) rather than maximizing absolute accuracy in fully instrumented sites.

Evidence from a survey of 26 practitioners indicates that information needs are spatially bounded: PoIs and team positions within roughly 16 m are prioritized, whereas elements beyond approximately 25 m are often not actionable. A geo-referenced map with concise iconography and short text is preferred to purely acoustic feeds, in line with design principles for emergency human–machine interfaces that emphasize low cognitive load, glanceability, and spatial grounding [12,13,14,15]. These observations inform an operator-centered interface with filterable layers to control clutter during operations. In routine use, tag encounters are automatic and do not require deliberate stops or fixed standoff distances; encounters occur within typical read ranges during normal motion along planned routes, and the next encounter bounds any interim drift. Given the online modality and recruitment through operational channels, the sample is convenience-based and generalizability is limited; a larger task-based usability evaluation is planned as follow-up.

While the present study did not explicitly evaluate metal-rich settings, multi-floor transitions, or multi-person coordination, the architecture was conceived for these scenarios. In metal-intensive areas, magnetic disturbances and radio occlusions can affect heading estimates and reduce tag read range; the inertial loop preserves short-term continuity through gyroscopes and step-based constraints, and tag encounters act as opportunistic resets when available. Practical deployments should tune tag density and placement near decision points to mitigate local radio shadowing and add redundancy in critical areas. From a deployment standpoint, passive UHF tags minimize operational burden (no power/batteries), allow on-metal form factors where needed, and support periodic audits; scalability is achieved by zoning and by encoding floor/zone IDs, so maintenance remains lightweight even over large multi-floor areas. In multi-floor facilities, vertical motion cues and barometric trends can provide transition hints, while floor identifiers encoded on tags enable unambiguous registration across levels. For multi-person collaboration, the system supports glanceable, geo-referenced awareness with filterable layers and can share context locally through tags, with an optional uplink to command centers when connectivity is available. These are design-oriented considerations rather than validated results and define specific aspects for future trials.

Operationally, encoding PoI category, type, and relative localization directly on tags supports an offline-first workflow in which pose can be reset and context retrieved locally without a backend, then visualized on a layered and filterable map. At the same time, operating at scale will require authoring and quality-control procedures for tag programming to prevent inconsistent metadata and to sustain reliability over time. To support reproducibility and external comparison, future trials will maintain a standardized site log covering floor- and corridor-level dimensions, materials and fixtures (including metal structures), electromagnetic context (RF clutter and interference sources), and environmental conditions; we will also report sensor health/noise (e.g., IMU bias trends) and RFID read-rate versus orientation, together with full tag coordinates and configuration.

The present study does not yet report a head-to-head benchmark. A fair experimental comparison will require matched trajectories, shared maps, and comparable infrastructure budgets to assess accuracy, robustness, and cognitive load across alternatives. Within this scope, the present contribution is a field-introductory blueprint: a deployable workflow and data model that integrate positioning and SA under degraded visibility/connectivity, to be followed by controlled head-to-head evaluations once standardized, scenario-matched protocols are in place [4,5,8,28].

7. Conclusions

This work introduces a practical system for SA tailored to first responders in deep-indoor environments, combining IMU-based PDR with opportunistic RFID tag encounters to deliver both localization and contextual information on a geo-referenced, glanceable interface. Our contribution is threefold. First, the hybrid IMU + RFID design bounds drift through in-field resets of position and heading, enabling reliable operation without continuous connectivity. Second, storing essential SA metadata on tags supports an offline-first workflow that keeps hazards, resources, and facilities accessible under degraded communications. Third, the interface is expressly shaped by end-user evidence showing that actionable information is concentrated within about 16 m, which guided design choices that minimized cognitive load during time-critical tasks.

Future work will prioritize a systematic and quantitative assessment of accuracy and robustness across materials and architectural layouts, with controlled comparisons against BLE, UWB, and vision-based baselines. Additional efforts will refine vertical-motion handling and multi-floor registration, introduce multi-user awareness with optional uplinks to command centers, and explore route suggestions that account for hazards and available resources while preserving low cognitive load. Finally, we will develop authoring and quality-control tools for tag programming and maintenance at scale, to ensure consistent metadata and dependable operation throughout deployment. Consistent with these considerations, we will carry out targeted evaluations at metal-intensive industrial sites to quantify resilience under magnetic and radio disturbances, extend validation to multi-floor buildings including stair and elevator transitions, and assess coordination in multi-person teams with attention to identity management and deconfliction. As such, the present work should be read as a field-introductory blueprint that operationalizes a pragmatic, offline-first integration of positioning and SA; subsequent phases will translate this blueprint into standardized comparative trials and expanded deployments across diverse, metal-rich, and multi-floor facilities.