Accurate user positioning plays a key role in millimeter wave massive multiple-input multiple-output (mmWaveMIMO) systems by improving beamforming and ensuring reliable connectivity. This paper presents a structured review of positioning techniques with a focus on high-mobility scenarios and challenging non-line-of-sight (NLoS) environments. We first review recent positioning methods that enhance communication speed, reliability, and eficiency in mmWave MIMO systems. It covers angle-based, time-based, and learning-based techniques that help overcome signal blockage, user mobility, and complex environments. We also review real-time positioning methods that support accurate beam alignment, reduce training time, and improve spectral eficiency. Finally, we discuss significant challenges, including beam misalignment, limited positioning accuracy, and high system complexity and present solution strategies that aim to improve system performance.
The ability to accurately determine a user’s location has become a key component of modern
communication systems. This feature supports various applications, including navigation, logistics, emergency
response, and a wide range of location-based services (LBS) [
As wireless networks move toward sixth generation (6G), the need for accurate and fast positioning
is becoming more important. Many researchers have already studied how positioning works in fifth
generation (5G) networks, especially using technologies like millimeter-wave (mmWave), massive
multiple-input multiple-output (mMIMO), and cooperative localization methods [
To support the growing need for accurate positioning, researchers and industry experts are focusing
on advanced technologies such as mmWave and distributed MIMO (D-MIMO). mmWave communication
works in the 30–300 GHz frequency range and ofers high time and spatial resolution for accurate
location tracking. It uses features such as angle of arrival (AOA) and time of flight (TOF) to improve
positioning, especially in places where traditional systems, including global positioning system (GPS)
do not work well [
However, achieving accurate and reliable positioning in high-mobility environments brings several
technical challenges. Although mmWave signals ofer high precision, objects like buildings, vehicles,
and people can easily block them. In high mobility, including trains and autonomous cars, the signal path
changes quickly and often without a clear line-of-sight (LoS). This makes it harder to maintain a stable
connection. The users moving at high speeds also face problems like doppler shift and beam squint.
Doppler shift happens when movement changes the signal frequency at low mmWave frequencies
beam squint afects the direction of the signal across diferent frequencies. Both issues can reduce the
quality of communication and reduce positioning accuracy. To deal with these problems, systems need
to track user movement in real time. This involves quickly estimating the position and direction of
moving users using fast feedback and smart algorithms. One commonly used method is the extended
kalman filter (EKF), which combines diferent signal measurements such as time-diference-of-arrival
(TDOA) to track the user’s location and speed [
In recent years, several studies have explored the role of positioning in supporting beamforming in mmWave and MIMO-based systems. These studies highlight its potential to reduce beam training overhead and improve link reliability. Many of these works focus on either mmWave localization or beamforming separately, often under simplified conditions. However, research that combines highaccuracy user positioning with mmWave-distributed MIMO beamforming in a unified framework remains limited. While some initial eforts have introduced context-aware beamforming strategies using location data, a detailed survey specifically targeting user positioning in mmWave MIMO systems is still lacking. Additionally, few studies address practical concerns such as implementation complexity, error sensitivity, and the impact of user mobility on system performance. This survey aims to fill that gap by reviewing existing positioning techniques in the context of the mmWave MIMO system and analyzing their impact on beamforming performance across various deployment scenarios. Figure 1 shows the positioning scenarios in the mmWave MIMO system. In this paper, we provide a detailed review of the current positioning techniques used in mmWave MIMO systems and examines their role in improving beamforming performance. The review categorizes these techniques into two main groups: those that rely on traditional positioning methods and those that integrate machine learning to enhance accuracy and adaptability. Additionally, the paper discusses the challenges in mmWave positioning, including issues related to positioning errors, orientation inaccuracies, obstacles in the environment, and storage requirements for positioning data. This study also explores the impact of these challenges on the beamforming process and focuses on their efect on the overall performance and eficiency of the system.
The rest of the paper is organized as follows. Section 2 provides a literature review covering key areas such as positioning in mmWave systems, location-aware beamforming, high-accuracy positioning in high-mobility scenarios, and techniques to reduce complexity in non-line-of-sight (NLoS) environments. Section 3 discusses the main challenges in mmWave positioning. Section 4 concludes the paper.
In recent years, various positioning techniques for mmWave MIMO systems have been proposed to address the challenges of accurate localization. This section provides a summary of the diferent approaches, categorizing them into angle-based, time-based, and signal-strength-based methods, and discusses the improvements made in positioning accuracy and system performance in these challenging environments. The table 1 shows the comparison of paradigm with diferent criteria.
Positioning methods in mmWave systems typically fall into three categories: angle-based, time-based,
and signal strength-based. Techniques like AoA and angle-of-departure (AoD) achieve high angular
resolution using narrow beams and large antenna arrays, but their performance drops in
multipathdense environments. To overcome this, recent work explores advanced models that leverage difuse
scattering and tensor decomposition to extract positional information even under NLoS conditions. To
improve positioning in high-speed railway scenarios, the authors [
In [
In [
In mmWave MIMO systems, beamforming is used to reduce signal loss and keep connections focused in the right direction. Traditional beam training methods can be slow and resource-heavy, especially when users are moving or there are many users in the area. Location-aware beamforming uses information about the user’s position from GPS, motion sensors, or radio signals to select and adjust the best beam direction. This makes it easier and faster to find the right beam, improving connection speed and reducing delay. In [14], the authors proposed a method to enhance indoor positioning accuracy by combining the observed time diference of arrival (OTDOA) with beamforming. OTDOA estimate the user position based on the time diference in receiving reference signals from multiple base stations. However, its accuracy is limited by quantization errors, particularly in environments where many user devices are closely spaced leading to overlapping estimated positions. To address this, the authors used beamforming techniques that rely on measuring the reference signal received power (RSRP) from the two strongest beams. These RSRP values are used to estimate the angle between the user and the serving base station. The system updates the initial OTDOA-estimated position by projecting it along this angle, which gives more accurate location.
In [15], a positioning-assisted three-dimensional (3D) beamforming system was designed to improve the reliability and eficiency of millimeter-wave communication. Traditional beamforming based on channel state information (CSI) requires high training overhead and power consumption, which becomes impractical with large antenna arrays. To address this, the authors used user positioning data to steer beams toward the estimated receiver location and reduce the need for complex channel estimation. The paper develops a closed-form expression for outage probability that considers factors such as positioning errors, transmission power link distance, and beamwidth. The paper [16] introduced a method to improve positioning accuracy in cellular MIMO systems by using Hybrid Analog and Digital Beamforming (HBF). With the growing need for high-accuracy positioning, the authors focus on a novel technique called sensing beamforming, which enhances positioning while maintaining eficient communication. This approach optimizes fisher information to adjust beam alignment and allocate power across multiple resources, including the time frequency and beam dimensions. The method starts by estimating sensing elements such as the AoA, AoD, and time of arrival (ToA) from multiple signal paths. A newton-based algorithm then refines these estimates to filter out multiple path clutters caused by NLoS conditions, making the positioning more robust.
The authors [17] proposed a method to optimize intelligent reflecting surface-assisted multiple-input single-output (IRS-assisted MISO) systems by jointly designing beamforming and IRS positioning in multi-access point environments. They introduced algorithms based on generalized benders decomposition (GBD) for beamforming, mixed-integer semidefinite programming (MISDP) for access point (AP) user pairing, and a heuristic iterative link removal (GBD-ILR) to reduce computational complexity. The paper also explores aerial IRS positioning using a successive convex approximation (SCA) method to improve performance. Finally, [18] proposed a machine learning-based fingerprint positioning method for massive MIMO systems in NLoS environments. It is considered a spatially refined beam-based channel model to improve angle resolution and extract a beam domain channel amplitude matrix as a location-related fingerprint, which includes multi-path information including amplitude, AoA, and DoA. This work introduced two machine learning models. The first is a classification-based model that categorizes user terminal (UT) fingerprints to reduce search space. The second is a regression-based model that directly estimates position coordinates and eliminates the need for similarity searches.
Achieving high positioning accuracy in high-mobility scenarios is a critical challenge for modern wireless systems, especially in applications such as autonomous vehicles and high-speed trains. Rapid movement, frequent handovers, and changing signal conditions require robust and responsive positioning techniques. The integration of high-accuracy positioning services with high-speed vehicles in mmWave communications was discussed in [19]. It focuses on simultaneous localization and communication (SLAC) techniques that use CSI to track vehicles in dynamic environments. While Doppler and spatial wideband efects challenge positioning accuracy, the paper shows how these efects can be used to improve vehicle tracking. By using parameters such as ToA, AoA, and Doppler shifts, the study demonstrates that high-accuracy positioning can be achieved with low-complexity algorithms that approach the cramér-rao lower bound (CRLB) performance. The work in [20] explores high-accuracy positioning for urban road trafic using 5G mmWave technology. It evaluates various positioning techniques, such as multilateration (based on time-of-arrival and signal strength) and triangulation (using angle-of-arrival), as well as hybrid methods combining both. The authors used ray-tracing simulations to evaluate performance in real-world conditions taking into account errors and network synchronization. This paper [21] focused on positioning high-speed trains (HSTs) using 5G new radio (NR) synchronization signals. The authors explore how time-of-arrival (TOA) and angle-of-departure (AOD) measurements from 5G synchronization signals can be used to track the train’s position, velocity, and acceleration in real time. Using an extended kalman filter (EKF), the train’s position is estimated even at high speeds (up to 500 km/h). The results show that combining both TOA and AOD measurements provides sub-meter accuracy for over 75% of the tracking time, which is essential for mission-critical applications like autonomous trains.
In [22], they presented a real-time fusion positioning system for urban rail transit that combines ultra-wideband (UWB) and inertial measurement unit (IMU) data using an error-state kalman filter (ESKF) algorithm. The system aims to improve the accuracy of the train position in tunnels where satellite-based positioning is inefective. The authors proposed a UWB ranging error model based on both static and dynamic measurements and validated the fusion algorithm through simulations and real-world tests. The results showed that the ESKF-based fusion system greatly improves positioning accuracy compared to standalone UWB that achieves an error of less than 0.4 meters under normal conditions. High-accuracy positioning method that combines carrier phase measurements and Bayesian estimation to enhance mobile features for positioning was presented in [23]. The authors propose a system that estimates the mobile feature using time-diferential carrier phase measurements and utilizes this information along with distance and carrier phase measurements to enhance positioning accuracy. The method applies bayesian estimation to calculate the posterior probability, which is then solved using a factor graph and sum-product algorithm.
Accurate positioning in mmWave systems with low hardware requirements requires reducing system complexity in NLoS or partially blocked scenarios without sacrificing performance. In [ 24], the lowcomplexity method was presented for positioning single-antenna users in mmWave systems using downlink (DL) signals. Unlike traditional uplink (UL) methods that rely on AoA and require more power from user devices, this approach uses AOD and adaptive beamforming from the base station. A two-step process is proposed: first, the base station sends uniform signals, and the user estimates its rough direction and feeds it back and second, the base station steers focused beams toward that direction, improving accuracy. The authors [25] proposed a low-complexity position estimation method for beyond 5G and 6G networks using angle-based localization. It combines downlink AoD and uplink AoA measurements in a massive MIMO-OFDM system. Instead of relying on complex iterative methods like gradient descent, it uses a simplified least squares (LS) approach to estimate user location with reduced computation.
Machine learning methods to improve radio positioning in NLoS environments within beamforming networks were discussed in [26]. This work compares three techniques: k-nearest neighbors (k-NN), deep neural networks (DNN), and long short-term memory networks (LSTM) for estimating user positions using signal features from beamforming transmissions. This paper [27] presents a practical approach for tracking vehicles in multipath NLoS environments using mmWave MIMO systems. The method relies on a two-stage Kalman filter. In the first stage, the system estimates a rough position by tracking signal directions from the base station. In the second stage, it improves the estimate by using motion data from the vehicle. By treating both the wireless channel and the vehicle’s movement as evolving patterns over time, the system adapts well to sudden changes in the environment.
The mmWave technologies ofer significant promise in improving user positioning accuracy but face several challenges in real-world environments. This section explores the main research challenges associated with mmWave positioning systems, focusing on their impact on performance and potential solutions.
One of the major challenges in positioning beamforming is maintaining accurate beam alignment in high-mobility environments. When users or vehicles move rapidly, frequent changes in location cause the AoA and AoD to shift quickly. This dynamic behavior leads to beam misalignment, which degrades both communication quality and positioning accuracy. Additionally, high mobility introduces Doppler shifts that distort the received signals and complicate channel estimation. The situation is made worse in wideband systems where beam squint causes diferent subcarriers to experience diferent spatial characteristics, making it dificult to maintain consistent beamforming across the bandwidth. These factors combined make real-time low-latency beam tracking very challenging when limited by computational resources and the lack of precise real-time location information. To address beam misalignment in high-mobility environments, techniques like low-complexity direct position determination (DPD) algorithms and fisher information-based hybrid beamforming optimization can improve positioning accuracy by leveraging doppler shifts and beam squint efects. Machine learning models can further optimize beam alignment by dynamically adapting to movement patterns.
Positioning accuracy in NLoS scenarios is significantly impacted by both static and mobile obstacles. In mmWave networks, static obstacles like buildings and walls block the LoS between the transmitter and receiver, leading to signal degradation. Mobile obstacles, such as moving people, are especially problematic in indoor environments. These obstructions generate multipath signals that complicate the localization process and reduce accuracy. GPS becomes unreliable in urban and indoor areas due to poor signal strength, while technologies like Li-Fi, which depend on LoS communication, are highly sensitive to any obstructions. The constant movement of users in environments like vehicles further increases the likelihood of signal blockage, making it more challenging to maintain accurate positioning. To solve positioning accuracy issues in NLoS scenarios, techniques like crowdsensing, machine learning models (k-NN, DNN, LSTM), and sensor fusion can be used to improve accuracy. Kalman filters, dynamic channel estimation, and autoregressive models help track and predict channel changes.
Positioning overhead is also one of the major challenges in beamforming systems that rely on positioningbased beamforming schemes that use large databases for location information. These systems require significant computational resources and storage to manage large amounts of positioning data. Storing current and historical positioning information in databases consumes a lot of memory in dynamic environments with many users. Beam training setup times increase as the system needs to update beam pairs during each beacon frame. This leads to higher power consumption. Additionally, maintaining accurate positioning in environments with mobility and obstacles such as vehicles or urban areas increases the dificulty. Traditional ofline learning methods struggle to adapt to environmental changes, while machine learning methods require careful management of feature selection and network complexity.
In this paper, we presented a systematic review of user positioning methods in mmWave MIMO systems and explained their role in improving beamforming in high mobility and NLoS scenarios. This paper covers angle-based, time-based, and learning-based approaches that improve beamforming accuracy, reduce overhead, and enable real-time communication. It also presented major challenges, including beam misalignment, positioning errors, and system complexity, along with solution techniques from recent research.
This work has received funding from the Marie Sklodowska-Curie Actions (MSCA) under the project "ultra-massive MIMO for future cell-free heterogeneous networks" (MiFuture, grant agreement ID 101119643). This work was also supported by FCT - Fundação para a Ciência e Tecnologia, I.P. by project reference UIDB/50008, and DOI identifier 10.54499/UIDB/50008”
During the preparation of this work, the authors used Grammarly in order to spelling check. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the publication’s content.. [13] Y. Zhang, G. Wu, X. Li, T. Xu, M. Zhang, F. Guo, Direct position determination based on passive synthetic aperture for coherent receivers, IEEE Sensors Journal 24 (2024) 17917–17925. doi:10. 1109/JSEN.2024.3386869. [14] H. Seo, H. Kim, T. Kim, D. Hong, Accurate positioning using beamforming, in: Proceedings of the 2021 IEEE 18th Annual Consumer Communications and Networking Conference (CCNC), Las Vegas, NV, USA, 2021, pp. 1–4. doi:10.1109/CCNC49032.2021.9369486. [15] Y. Zhang, K. Zhang, S. Huang, B. Zhu, Z. Zhang, Positioning-assisted 3d beamforming systems and their outage analysis, in: Proceedings of the 2024 IEEE/CIC International Conference on Communications in China (ICCC), Hangzhou, China, 2024, pp. 72–77. doi:10.1109/ICCC62479. 2024.10681753. [16] X. Chu, Z. Lu, J. Kang, Y. Zou, H. Zhang, X. Qiu, Hybrid beamforming toward positioning enhancement under cellular mimo systems, IEEE Transactions on Wireless Communications 23 (2024) 13545–13561. doi:10.1109/TWC.2024.3403568. [17] C. Tang, C. C. Fung, Z.-E. Ni, M. Servetnyk, Joint beamforming and aerial irs positioning design for irs-assisted miso system with multiple access points, IEEE Open Journal of the Communications Society 5 (2024) 612–632. doi:10.1109/OJCOMS.2023.3346895. [18] X. Gong, X. Yu, X. Liu, X. Gao, Machine learning-based fingerprint positioning for massive mimo systems, IEEE Access 10 (2022) 89320–89330. doi:10.1109/ACCESS.2022.3199728. [19] Z. Gong, X. S. Shen, C. Li, Y. Song, R. Su, High-accuracy positioning services for high-speed vehicles in wideband mmwave communications, IEEE Transactions on Signal Processing 71 (2023) 3867–3882. doi:10.1109/TSP.2023.3324727. [20] S. Häger, N. Gratza, C. Wietfeld, Characterization of 5g mmwave high-accuracy positioning services for urban road trafic, in: Proceedings of the 2023 IEEE 97th Vehicular Technology Conference (VTC2023-Spring), 2023, pp. 1–7. doi:10.1109/VTC2023-Spring57618.2023.10200661. [21] J. Talvitie, T. Levanen, M. Koivisto, K. Pajukoski, M. Renfors, M. Valkama, Positioning of high-speed trains using 5g new radio synchronization signals, in: Proceedings of the 2018 IEEE Wireless Communications and Networking Conference (WCNC), 2018, pp. 1–6. doi:10.1109/WCNC.2018. 8377225. [22] R. Wang, H. Jiang, Q. Zhang, G. Liu, F. R. Yu, Real-time uwb and imu fusion positioning system for urban rail transit with high mobility, in: Proceedings of the 2024 IEEE 99th Vehicular Technology Conference (VTC2024-Spring), 2024, pp. 1–7. doi:10.1109/VTC2024-Spring62846.2024. 10683298. [23] S. Fan, R. Zeng, H. Tian, Mobile feature enhanced high-accuracy positioning based on carrier phase and bayesian estimation, IEEE Internet of Things Journal 9 (2022) 15312–15322. doi:10. 1109/JIOT.2022.3147949. [24] A. Fascista, A. Coluccia, H. Wymeersch, G. Seco-Granados, Low-complexity accurate mmwave positioning for single-antenna users based on angle-of-departure and adaptive beamforming, in: Proceedings of the 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 4866–4870. doi:10.1109/ICASSP40776.2020.9053493. [25] A. A. Masal, V. Singh, R. Ganti, J. K. Milleth, Low complexity position estimator for dl-aod and ul-aoa for b5g and 6g networks, in: Proceedings of the 2024 IEEE 35th International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 2024, pp. 1–7. doi:10.1109/ PIMRC59610.2024.10817315. [26] R. Klus, J. Talvitie, J. Vinogradova, J. Torsner, M. Valkama, Machine learning based nlos radio positioning in beamforming networks, in: Proceedings of the 2022 IEEE 23rd International Workshop on Signal Processing Advances in Wireless Communication (SPAWC), 2022, pp. 1–5. doi:10.1109/SPAWC51304.2022.9834010. [27] Z. Ye, J. Vinogradova, G. Fodor, P. Hammarberg, Vehicular positioning and tracking in multipath non-line-of-sight channels, in: Proceedings of the 2022 IEEE 95th Vehicular Technology Conference (VTC2022-Spring), IEEE, 2022, pp. 1–5. doi:10.1109/VTC2022-Spring54318.2022.9860711.