Positioning in underground environments poses a fundamental challenge due to the absence of GNSS signals and often limited communication infrastructure. This survey investigates positioning systems in mines, tunnels, and other subterranean settings that involve the client device in a significant way - either by performing the localization directly on the device (fully client-based) or by recording sensor data locally for server-side processing (client-assisted). Based on a structured analysis of over 30 selected systems, we classify and compare approaches by signal technology, algorithmic method, infrastructure requirements, and system topology. The results show a dominance of RF-based solutions, but also highlight promising alternatives like magnetic methods. While fully client-autonomous systems are still rare, recent advances in onboard processing, sensor fusion and SLAM demonstrate the increasing potential of client-side localization in safety-critical and infrastructure-poor underground scenarios.
Positioning in underground environments such as mines, tunnel systems, or caves poses considerable challenges to current research. Unlike above-ground or urban settings, no external signals such as GPS are available underground. Additional constraints such as darkness, moisture, narrow geometries, and highly variable material structures further limit conventional indoor localization approaches. This becomes particularly critical in emergency situations: power outages, damaged infrastructure, or lack of communication technology may render traditional systems completely inoperative—precisely when reliable positioning is needed most.
Compared to conventional indoor environments, underground localization scenarios difer not only due to the absence of GNSS signals but also in terms of technical boundary conditions: Tunnel geometries—particularly for radio-based systems—are prone to multipath propagation and signal reflections. Lighting conditions and line-of-sight paths for radio-frequency systems are severely limited or subject to significant attenuation. Moreover, infrastructure failure can quickly escalate into an emergency situation, which is why special attention must be given to this aspect.
In well-equipped mines, advanced infrastructure-dependent localization systems are already in use to reliably track personnel and machinery. However, these systems typically rely on complex installations such as wireless networks, RFID tagging, or so-called leaky-feeder cables—solutions that require continuous maintenance and substantial investment. Such systems are unavailable for visitor mines, temporary operations, scientific expeditions in cave systems, or unauthorized entries. Even in well-equipped mines, infrastructure cannot be assumed to remain functional in real emergency scenarios.
Accordingly, the development of systems that determine and display positions on the user side (client-side) is of high importance. The degree of operational independence—possibly even without communication links or network coverage—must be critically examined. Such infrastructure-resilient solutions are essential not only for safe return from uncharted or damaged areas, but also for explorative scenarios with minimal equipment.
This survey provides a systematic overview of current approaches to underground localization, focusing on systems that operate on the client side. The aim is to classify existing methods, identify key challenges, and highlight research gaps. Particular attention is given to the question of whether, and to what extent, localization can be achieved without external infrastructure, and which technologies enable reliable positioning under extreme conditions.
Unlike typical indoor environments, underground settings are characterized by extreme attenuation and
lack of fallback infrastructure. Traditional infrastructures such as GNSS, WLAN, or mobile networks
are generally unavailable in mines, tunnels, or cave systems—or become non-functional in emergency
scenarios. In short: there is “no longer any help from above” [
Typical application scenarios include: (a) Routine operation in industrial settings (e.g., navigation of personnel, machine guidance, documentation of tunneling progress), (b) Emergency operation in the event of power failure or structural collapse, such as for self-rescue or locating missing persons, (c) Exploratory use in research, e.g., in unmapped mines or caves where no technical infrastructure exists, can be provided, or is allowed (e.g., visitor mines, cave expeditions, or unauthorized entries).
In routine operation, underground positioning systems are typically designed to transmit the position of workers and machines to a central control unit. Seguel et al. refer to such systems as “Remote Positioning” when the position is determined externally to the client, and as “Indirect Remote Positioning” when the client determines its position but transmits it to a central system [2, p. 9]. A variety of commercial solutions exist for this use case, predominantly based on radio-frequency technologies. A key example are leaky feeder systems, also referred to as “radiating cables” [3, p. 1], [4, p. 15]. These systems use special cables with periodic openings (“leaks”) to emit signals at regular intervals along a tunnel (first core) and simultaneously receive signals (second core).
Originally designed for communication, these systems are now also being explored for positioning
purposes using spectral analysis of dedicated chirp signals [
In a review study, Yarkan et al. point out the advantages of coverage, but also highlight the need for line-of-sight (LOS), power supply, and the susceptibility to single cable cuts as major disadvantages [7, p. 136]. If the cables remain intact, leaky feeder systems can provide cost-efective communication even in emergencies [“£10/100 m” 8, p. 226]. Bedford et al. report a “useable signal strength [...] to a range of 800 m” [8, p. 224].
The application contexts of “emergency operation” and “exploration” in particular lead to a number
of specific requirements for localization systems. The main focus is on resilience to infrastructure
failure, robustness to environmental conditions (e.g., darkness, moisture, dust, unstable geometries),
and intuitive usability—even under stress, systems must be operable by non-expert users [
In routine underground operations, it is essential to know the location of personnel at all times in
order to initiate targeted rescue measures in the event of an emergency. Workers are trained in how to
act. In non-productive settings—such as visitor mines or cave exploration by laypersons—client-side
localization is often the only way to inform the user of their position. Seguel et al. refer to this approach
as “Self Positioning”, but describe the clients as “dumb node[s]” [2, p. 9]. This terminology suggests that
the clients do not perform the position estimation themselves but merely receive externally computed
results. In such underground applications, mines can be retrofitted after their productive phase with
localization technologies. With the increasing substitution of formerly specialized technology by
standardized “of-the-shelf” solutions, positioning functionality is becoming available to clients without
further technical efort. WiFi access points with RSSI-based distance estimation [
The notion of Self Positioning—understood as actual position estimation performed by the client—stands in contrast to the dumb-node paradigm. In exploration scenarios, no external option for position estimation or communication is typically available. Systems should therefore also be evaluated in terms of the degree to which the client is capable of autonomously determining its own position. Table 1 lists works that implement signal acquisition and evaluation on the client side. Summary of Trends and Gaps The literature shows a clear dominance of RF-based approaches, yet true infrastructure-free client-side systems remain rare. While many systems achieve sub-meter accuracy in controlled environments, real-world evaluations—especially under emergency constraints—are scarce. Magnetic methods ofer promising alternatives, particularly when combined with inertial fusion, but their deployment is currently limited to prototypes.
An analysis of the identified signal technologies reveals a clear dominance of radio-frequency (RF)-based systems. A total of 15 of the examined systems utilize RF communication. Another focus is on optical systems, either in the form of visible light communication (VLC) or through sensor technologies such as cameras. These optical methods often achieve very high accuracy in the sub-meter range, but rely on line-of-sight conditions and stable lighting environments.
Magnetic field–based localization is represented by a total of five systems. Two of these systems generate and evaluate artificial magnetic fields: Lin et al. and Abrudan et al. use wireless underground sensor networks (WUSNs) to create time-modulated magnetic fields, which are then used by clients to estimate their position [36, p. 1454], [25, p. 4389].
Of the three systems that rely on the Earth’s magnetic field [ 43, 42, 44], one is used to reconstruct the trajectory of a drill during a boring operation [42]: In this case, the position of the drill head is estimated via dead reckoning and refined using geomagnetic fingerprinting [“dead reckoning is used at ifrst” 42, p. 1379].
Haverinen and Kemppainen recorded magnetic vectors using a sensor array mounted on a board containing 60 sensors distributed over an area of 0.5 × 0.5 m [44]. The authors employed Monte Carlo Localization (MCL) using the measurements from these 60 sensors. Particle filter–based methods often require a prior reference map or rely on server-side computation, as is the case in the work of Haverinen and Kemppainen [44].
Among the surveyed systems, filtering techniques include Kalman filters (e.g., for inertial fusion), Particle filters (e.g., for geomagnetic localization), and SLAM-based graph optimization.
In summary, RF-based systems continue to play a dominant role in underground positioning. While optical methods are being explored, their real-world deployment remains limited. Alternative signal sources such as magnetism, light, or acoustics are currently used primarily in specialized niche applications. Comparative overview of underground self-localization systems based on primarily client-side or clientassisted signal processing and positioning (“client-side” topology means: both, signal acquisition and
The systems considered in this survey can be categorized based on their topology into three types: • fully client-side systems, • systems with a return channel, • and approaches with server-side computation based on client-side data acquisition.
A special case in which the client merely transmits a signal while reception and computation are fully handled on the server side was included as last item in table 1 for completeness, but strictly speaking, it does not qualify as client-side data acquisition or processing.
In the first category—fully client-side systems—all positioning computations are carried out locally
on the device, without any return channel or external servers. These systems are characterized by
high autonomy and are particularly suitable for emergency scenarios and exploration settings without
network connectivity. Examples include systems that rely solely on optical sensing or inertial navigation.
Ren and Wang describe a system based on 3D LiDAR and odometry, which operates entirely onboard a
remotely controlled vehicle in a tunnel environment [
An intermediate form between pure client-side and server-based approaches is represented by the
system of Li and Zhan. They use a laser scanner for onboard localization of mining vehicles in a
real-world scenario, while higher-level functions such as path planning, progress monitoring, and data
fusion are handled by a central server [
Client-side systems typically rely on infrastructure-free principles or strategically placed
sensors [40, “distance between nodes”], [
Another category includes systems with a return channel, in which the end device collects data locally, but the actual position estimation is carried out by a central instance. Wu and Zhang implemented a system based on visible light, in which the mobile device communicates with a central unit using Manchester-modulated headlamp signals [37]. In [41], VLC signals are combined with inertial step counting, and a reverse communication path to the base station is integrated. This inverts the traditional principle: headlamps transmit a code, which is received by ceiling-mounted sensors. Lin et al. use a magnetic system enhanced by return-channel communication to improve positioning accuracy [36].
Infrastructure in this class is typically characterized by strategic placement or reuse of existing infrastructure. The resulting accuracies often fall below one meter, especially when UWB or hybrid methods are employed.
In the third category, data is acquired locally, while the actual position computation is carried out on a server. Cypriani et al. refer to the central unit as an “aggregation server” [39, p. 3]. These systems are therefore not suitable for ofline use and are typically intended for production-related applications. Lin et al. present a WiFi fingerprinting system, in which positioning for “moving Android smart phones” is handled by a dedicated server module (“database server”) [38, p. 4]. The system was deployed in a large-scale tunnel setting of a dam project and achieved positional accuracies of around 3–5 meters.
Yinjing et al. developed a method based on feature vectors and matching against a previously recorded tunnel database. Signal characteristics are captured on the mobile device, but the actual matching is performed on a server [45]. The UWB-based system by Song et al. also follows this architecture and uses a particle filter to enhance server-side estimation accuracy [ 46]. However, the test setup was limited in scale [“nine reference points” 46, p. 7]. The underlying infrastructure in this category is typically strategically placed.
The variety of analyzed systems demonstrates that there is no one-size-fits-all solution for underground applications. Systems with fully client-side computation ofer the greatest potential for autonomous operation in infrastructure-poor or emergency-driven scenarios. In contrast, return-channel and servercentric architectures provide high accuracy and feature-rich functionality for production use but rely on intact communication links.
The review of existing systems for client-side underground localization reveals a wide range of technical approaches, whose suitability depends heavily on the specific application context. While infrastructurebased solutions still dominate in productive mining operations, autonomous client-side systems are particularly relevant for exploratory scenarios, temporary deployments, and emergencies. Their ability to operate independently of central infrastructure or communication makes them an essential component of future localization strategies.
The systems analyzed can be grouped into three main categories: fully client-side systems, systems with a return channel, and systems with server-side computation. Notably, there is a growing number of solutions that enable localization directly on the client device. These systems increasingly rely on modern sensor technology (e.g., LiDAR, cameras, inertial units, magnetic sensors) and algorithmic methods such as particle filters, SLAM, or deep learning. The rising computational power of mobile devices is gradually shifting previously server-side computations to the clients themselves, increasing their autonomy—especially in the context of computationally intensive methods such as fingerprinting, particle filtering, and neural networks.
Nevertheless, significant research gaps remain. Many of the systems described in the literature are prototypes, simulations, or were evaluated only under idealized conditions. As a result, their applicability to real, complex underground environments is not always guaranteed. In particular, systems that operate entirely without infrastructure remain rare—despite their especially high value in emergency scenarios.
Future research should increasingly focus on robust localization methods that function without infrastructure and remain operational even under harsh conditions. In addition, systematic evaluation under realistic field conditions is needed to ensure the practical viability of proposed approaches. The combination of standardized sensor technology, eficient signal processing, and adaptive, learning-based algorithms ofers promising perspectives for the development of resilient underground navigation systems.
In accordance with the CRediT (Contributor Roles Taxonomy) taxonomy, the first author was solely responsible for the conceptualization, methodology, investigation, formal analysis, data curation, visualization, and writing of the original draft and revisions of the manuscript. Authors 2 and 3 acted in supervisory and administrative roles, providing institutional support and funding acquisition. They were not involved in the conceptual development, writing, or editing of the manuscript.
In line with the CEUR-WS Taxonomy for the Use of Generative AI in Scientific Writing 1, the author(s) used GPT-4 during the preparation of this work for piecewise translation of text segments (T1.1) and for grammar and spelling correction (R1.1). These uses were limited to language-level assistance; all intellectual contributions remain the sole responsibility of the human authors. tioning, thesis, Toronto Metropolitan University, 2023. doi:10.32920/24050727.v1. [35] U. I. Minhas, I. H. Naqvi, S. Qaisar, K. Ali, S. Shahid, M. A. Aslam, A WSN for Monitoring and Event Reporting in Underground Mine Environments, IEEE Systems Journal 12 (2018) 485–496. doi:10.1109/JSYST.2016.2644109. [36] S.-C. Lin, A. A. Alshehri, P. Wang, I. F. Akyildiz, Magnetic Induction-Based Localization in Randomly Deployed Wireless Underground Sensor Networks, IEEE Internet of Things Journal 4 (2017) 1454–1465. doi:10.1109/JIOT.2017.2729887. [37] G. Wu, J. Zhang, Demonstration of a Visible Light Communication System for Underground Mining Applications, DEStech Transactions on Engineering and Technology Research (2016). doi:10.12783/dtetr/iect2016/3708. [38] P. Lin, Q. Li, Q. Fan, X. Gao, S. Hu, A Real-Time Location-Based Services System Using WiFi Fingerprinting Algorithm for Safety Risk Assessment of Workers in Tunnels, Mathematical Problems in Engineering 2014 (2014) 371456. doi:10.1155/2014/371456. [39] M. Cypriani, G. Delisle, N. Hakem, Wi-Fi-based positioning in underground mine tunnels, in: International Conference on Indoor Positioning and Indoor Navigation, 2013, pp. 1–7. doi:10. 1109/IPIN.2013.6817894. [40] Y. Qin, C. Zhou, S.-H. Yang, F. Wang, A distributed newton iteration based localization scheme in underground tunnels, in: Proceedings of 2012 UKACC International Conference on Control, 2012, pp. 851–856. doi:10.1109/CONTROL.2012.6334743. [41] M. Pang, K. Zhang, X. Yang, Y. Yin, S. Gao, P. Chen, Reliable Visible Light-Based Underground Localization Utilizing a New Mechanism: Reverse Transceiver Position, Lecture Notes in Computer Science (2020). doi:10.1007/978-3-030-59019-2_7. [42] B. Park, H. Myung, Resilient Underground Localization Using Magnetic Field Anomalies for Drilling Environment, IEEE Transactions on Industrial Electronics 65 (2018) 1377–1387. doi:10. 1109/TIE.2017.2733420. [43] T. Makkonen, R. Heikkilä, A. Kaaranka, The applicability of a geomagnetic field based positioning technique with mobile phone to underground tunnels, 31st International Symposium on Automation and Robotics in Construction and Mining, ISARC 2014 - Proceedings (2014) 953–959. [44] J. Haverinen, A. Kemppainen, A geomagnetic field based positioning technique for underground mines, in: 2011 IEEE International Symposium on Robotic and Sensors Environments (ROSE), 2011, pp. 7–12. doi:10.1109/ROSE.2011.6058533. [45] G. Yinjing, S. Xianqi, Y. Lei, L. Wenhong, A Coal Mine Underground Localization Algorithm Based on the Feature Vector, Journal of Engineering and Technological Sciences 51 (2019) 184–197. doi:10.5614/j.eng.technol.sci.2019.51.2.3. [46] B. Song, S. Zhang, J. Long, Q. Hu, Fingerprinting Localization Method Based on TOA and Particle Filtering for Mines, Mathematical Problems in Engineering 2017 (2017) 3215978. doi:10.1155/ 2017/3215978. [47] D. Larionov, O. Lukashenko, A. Moschevikin, R. Voronov, A. Sikora, Improving the Accuracy for Radio-Based Positioning in Mines Using SLAM, in: 2020 IEEE 5th International Symposium on Smart and Wireless Systems within the Conferences on Intelligent Data Acquisition and Advanced Computing Systems (IDAACS-SWS), 2020, pp. 1–7. doi:10.1109/IDAACS-SWS50031. 2020.9297088. [48] Z. Li, Z. Li, Reference coordinate based Chan algorithm for UWB personnel localization in underground coal mines, Scientific Reports 15 (2025) 17922. doi:10.1038/s41598-025-03007-6.