Advancements in Indoor Precision Positioning: A Comprehensive Survey of UWB and Wi-Fi RTT Positioning Technologies

Abstract

High-precision indoor positioning is essential for various applications, such as the Internet of Things, robotics, and smart manufacturing, requiring accuracy better than 1 m. Conventional indoor positioning methods, like Wi-Fi or Bluetooth fingerprinting, typically provide low accuracy within a range of several meters, while techniques such as laser or visual odometry often require fusion with absolute positioning methods. Ultra-wideband (UWB) and Wi-Fi Round-Trip Time (RTT) are emerging radio positioning technologies supported by industry leaders like Apple and Google, respectively, both capable of achieving high-precision indoor positioning. This paper offers a comprehensive survey of UWB and Wi-Fi positioning, beginning with an overview of UWB and Wi-Fi RTT ranging, followed by an explanation of the fundamental principles of UWB and Wi-Fi RTT-based geometric positioning. Additionally, it compares the strengths and limitations of UWB and Wi-Fi RTT technologies and reviews advanced studies that address practical challenges in UWB and Wi-Fi RTT positioning, such as accuracy, reliability, continuity, and base station coordinate calibration issues. These challenges are primarily addressed through a multi-sensor fusion approach that integrates relative and absolute positioning. Finally, this paper highlights future directions for the development of UWB- and Wi-Fi RTT-based indoor positioning technologies.

1. Introduction

The advancement of the Internet of Things, robotics, smart manufacturing, and location-based services, such as personnel security monitoring within power substations, has spurred a growing demand for indoor positioning [1,2]. Although the global navigation satellite system (GNSS) offers high-precision navigation and positioning in outdoor environments regardless of weather conditions, it falls short in indoor and indoor–outdoor transition areas. Consequently, a multitude of indoor positioning technologies have emerged in recent years, including ultra-wideband (UWB), Wi-Fi, Bluetooth, vision-based systems, LiDAR, and inertial navigation sensors [3].

Among these technologies, UWB and Wi-Fi Round-Trip Time (RTT, also referred to as Fine Timing Measurement, FTM) have gained prominence, supported by industry giants like Apple and Google, respectively. UWB technology stands out for its cost-effectiveness, robust anti-interference capabilities, and high transmission rates. In 2019, Apple integrated the U1 UWB chip into the iPhone11, further improving the positioning function of the mobile phone, which can not only sense the location of its own mobile phone, but also the location of other mobile phones in the surrounding area. The DW3000 series products launched by Qorvo can also be used with the supported Apple chips to realize nearby interaction functions [4]. This underscores the widespread adoption of UWB ranging technology for achieving centimeter-level positioning accuracy in industrial applications. On the other hand, as Wi-Fi signals are ubiquitous across various settings and mobile devices, Wi-Fi-based indoor positioning is considered a highly promising technology, and Google is promoting its broad applications with smart devices [5].

In a multipath environment, signals from different paths often experience time delays due to propagation distance and obstacles. For narrowband signals, these signals from different paths can easily overlap each other because the bandwidth of narrowband signals is very small, and the delays of each signal cannot be clearly distinguished, resulting in overlapping multipath signals and difficulty in separation. Unlike traditional radio signal transmission that uses high-frequency carriers to modulate narrow bandwidth signals, UWB technology uses narrow pulse modulation signals at the microsecond or even nanosecond level for transmission, as shown in Figure 1. The bandwidth of UWB (ultra-wideband) signals is very large, and the pulse duration is very short (usually at the nanosecond level); so, there will be a significant time difference between UWB signals on each path. As long as the time difference between these signals is greater than the width of a pulse, they will not overlap at the receiving end. Therefore, this design of UWB signals minimizes power consumption and maximizes bandwidth. In this way, the uniqueness of each path’s signal can be retained, allowing the receiver to distinguish each multipath signal and effectively combat signal distortion and noise [6]. Due to its short pulse duration, UWB signals can effectively penetrate obstacles and achieve high-precision ranging, thus establishing UWB’s unique advantage in indoor positioning.

The stringent radiated power restrictions imposed by the Federal Communications Commission in 2002 facilitated UWB’s transition from military to civilian uses. A comparative analysis of various indoor positioning technologies with UWB was conducted to provide a detailed historical perspective on UWB’s development [7]. The integration of UWB with other indoor positioning technologies was further compared in [8], which highlighted the suitability of UWB for widespread applications due to its robust resilience to multipath effects, although UWB technology faces challenges such as high deployment costs [9]. Additionally, the evolution of UWB industry standards over the past two decades has been extensively reviewed [10], and UWB technology is widely utilized for asset tracking [11,12] and positioning [13].

Wi-Fi technology offers a range of measurement options through different protocols for indoor positioning, including Received Signal Strength Indicator (RSSI), Channel State Information (CSI), and RTT [14]. Due to their low frequency and long wavelength, Wi-Fi signals face challenges in penetrating obstacles and experience significant attenuation in NLOS environments. RSSI and CSI measurements are commonly utilized in fingerprint-based positioning, and the positioning accuracy remains within several meters.

The IEEE 802.11mc protocol, introduced in 2016, developed the RTT ranging technology, marking a significant advancement in Wi-Fi-based positioning. Unlike RSSI and CSI, RTT-based ranging relies solely on the flight time between transmitting and receiving devices, alleviating constraints related to beam width and antenna configurations and offering superior ranging accuracy. Major hardware manufacturers such as Google and Compulab have developed mobile phones and routers with RTT ranging capabilities, and software ecosystems like Android have supported the RTT protocol. Standardization efforts in both software and hardware are driving advancements in Wi-Fi RTT technology, paving the way for rapid development of indoor positioning applications [15].

The IEEE 802.11mc protocol, introduced in 2016, developed RTT ranging technology, marking a significant advancement in Wi-Fi-based positioning. Later standards, such as IEEE 802.11be (Wi-Fi 7) and IEEE 802.11az, further enhanced RTT to improve the accuracy and efficiency of location-based services in Wi-Fi networks. In Wi-Fi 7 (802.11be), the focus has been on improving multi-link operation, allowing devices to use multiple radio interfaces for data transmission and reception. This feature helps maintain a stable connection in high-traffic or interference-prone environments. Additionally, new features in Wi-Fi 7, such as enhanced power saving and low-latency performance, further optimize the system’s ability to support real-time location tracking with minimal delay. On the other hand, IEEE 802.11az extends RTT by refining the underlying protocols to improve distance measurement accuracy, making Wi-Fi-based positioning more reliable, particularly in dense or dynamic environments. Unlike RSSI and CSI, RTT-based ranging relies solely on the flight time between the transmitting and receiving devices, overcoming limitations related to beamwidth and antenna configurations, offering superior ranging precision. Major hardware manufacturers, such as Google and Compulab, have developed smartphones and routers with RTT capability, and software ecosystems like Android also support the RTT protocol. The ongoing standardization efforts in both hardware and software are driving the advancement of Wi-Fi RTT technology, paving the way for the rapid development of indoor positioning applications.

UWB and Wi-Fi RTT share similarities in the smartphone-centric approach, positioning principles, and algorithms, yet they face challenges related to NLOS ranging errors. In order to achieve high positioning accuracy and continuity, both UWB and Wi-Fi RTT systems must address ranging measurement errors and integrate with complementary technologies such as visual odometry and Inertial Measurement Units (IMUs). However, several challenges remain in enhancing positioning accuracy and continuity, and it is necessary to have a systematic survey on state-of-the-art methodologies and prospective development in UWB and Wi-Fi RTT high-precision positioning. In the past decade, review work has mainly focused on an overview of all available indoor positioning technologies, or focused on a basic UWB positioning review, or a review of fingerprint-based Wi-Fi positioning, but the coverage of Wi-Fi RTT positioning is limited, and there is almost no work that compares UWB positioning and Wi-Fi RTT positioning in detail, as shown in

Table 1. Previous survey about UWB/Wi-Fi RTT positioning.

Reference	Year	Technology
[16]	2015	Overview of the Wi-Fi Fine time measurement-based positioning principle
[7]	2016	Overview of UWB positioning principle, algorithms and challenges
[6]	2017	Overview of fundamental UWB positioning principle and algorithms
[17]	2018	Overview of indoor positioning technology
[18]	2021	Overview of UWB self-calibration and collaborative localization
[8]	2022	Overview of the main positioning algorithms in general
[19]	2022	Overview of UWB-based smart logistics
[20]	2022	Overview of indoor high-precision positioning technology
[21]	2023	Overview of indoor high-precision positioning technology
[22]	2023	Overview of indoor positioning technology
[23]	2023	Overview of the classification of NLOS error and its impact on UWB positioning accuracy
[24]	2023	Overview of fundamental UWB positioning principle and algorithms and their recent development.
[25]	2023	Overview of NLOS identification and error mitigation for UWB indoor positioning
[26]	2023	Overview of Wi-Fi assisted indoor positioning on different principles
[27]	2024	Overview of deep learning-based Wi-Fi indoor positioning

. This paper compares the strengths and weaknesses of UWB and Wi-Fi RTT technologies, and provides a comprehensive survey regarding both UWB and Wi-Fi RTT positioning, from the signal characteristics of both signals, an overview of fundamental positioning principles, NLOS signal identification and error mitigation, to advanced positioning methodologies concerning enhanced positioning accuracy, continuity, and practical field deployment considerations. Finally, this paper outlines the prospective development of UWB- and Wi-Fi RTT-based high-precision positioning technologies.

In this paper, we comprehensively delve into various aspects of high-precision indoor positioning technologies, with a specific focus on UWB and Wi-Fi RTT positioning. The key areas of exploration include:

(1)

Comparative analysis of UWB and Wi-Fi RTT-based positioning: we first comparatively analyze the signal characteristics of UWB and Wi-Fi signals, which are the root cause of differences in performance such as bandwidth, multipath effects.

(2)

NLOS identification and ranging error mitigation methods: a critical aspect of improving ranging and positioning accuracy involves the detection, identification, and correction of NLOS errors in UWB and Wi-Fi RTT ranging measurements.

(3)

Advanced positioning methodologies: A variety of advanced positioning methodologies are reviewed, including methods for enhancing positioning accuracy and continuity, and practical considerations for efficient field deployment. Multi-sensor fusion is highlighted as a fundamental approach in this context.

(4)

Advantages and disadvantages: We thoroughly discuss the strengths and limitations of UWB and Wi-Fi RTT high-precision indoor positioning technologies, providing a balanced perspective on their applicability, potential challenges, and future development.

By addressing these key areas, this survey aims to contribute significantly to the understanding, development, and optimization of UWB- and Wi-Fi RTT-based indoor positioning, serving as a guiding reference for young researchers, engineers, and system designers exploring novel research directions and solutions within these domains. The structural framework is illustrated in Figure 2.

2. Overview of Positioning Principles Using UWB and Wi-Fi RTT Radio Signals

UWB and Wi-Fi RTT are two positioning technologies that have gained attention due to their high accuracy and broad applications. In this section, we will explore the fundamental principles of positioning using these two types of radio signals, focusing on two different timing measurement methods: the Time of Flight (TOF)-based positioning method and the Time of Arrival (TOA) and Time Difference of Arrival (TDOA)-based positioning methods.

2.1. TOF-Based Positioning Principle

UWB and Wi-Fi RTT technologies measure the distance between a request initiator (e.g., a tag or smartphone) and a responder (e.g., a base station) using the TOF approach, which calculates the distance by measuring the signal’s flight time between transmitting and receiving devices, as shown in Figure 3.

However, Wi-Fi RTT technology differs in that it measures the round-trip time of the signal travel multiple times and then averages these values to estimate the distance between the signal’s transmitting and receiving ends. The Wi-Fi RTT ranging principle is outlined as follows.

$$\left\{\begin{array}{l}d={\displaystyle \frac{1}{2}}\cdot T_{RTT}\cdot c-d_{trans}+d_{com}-d_{NLOS}-d_{MPE}+a+\epsilon \hfill \\ T_{RTT}={\displaystyle \frac{1}{N}}({\displaystyle {\sum }_{i=1}^N{t}_{4\_i}-{\displaystyle {\sum }_{i=1}^N{t}_{1\_i})-\frac{1}{N}({\displaystyle {\sum }_{i=1}^N{t}_{3\_i}-{\displaystyle {\sum }_{i=1}^N{t}_{2\_i})}}}}\hfill \end{array}\right.$$

(1)

where $d$ is the actual distance between the router and the mobile device; $T_{RTT}$ is the calculated average flight time; $d_{trans}$ is the error caused by the signal propagation time delay in the hardware; $d_{com}$ is the error introduced by the communication protocol; $d_{NLOS}$ and $d_{MPE}$ are errors caused by NLOS propagation and multipath effect, respectively, as shown in Figure 4; $a$ is the phase distortion error; $\epsilon$ represents the noise effect, $t_{1\_i}$, $t_{2\_i}$, $t_{3\_i}$, $t_{4\_i}$ are the times when the signal arrives and leaves the transmitting and receiving devices, respectively; $c$ is the speed of light in vacuum; $N$ is the number of successful ranging measurements.

The TOF approach directly measures the flight time of the signal traveling between a mobile carrier and the base station. Consequently, TOF only requires recording the time when the target transmits the signal and receives it back, although more accurate methods may also account for the reaction time of the base station. UWB and Wi-Fi RTT technology, both based on the TOF approach, are less susceptible to the inconsistencies in time reference between the transmitting and receiving ends. This is achieved through the calculation of the difference between the transmitting and receiving signal times at both ends, followed by the calculation of the difference in the obtained time differences and averaging them. Importantly, this processing occurs separately at each end, utilizing their respective time references. Consequently, the final calculation involves time differences rather than absolute time, eliminating the need to synchronize time references between the transmitting and receiving ends.

2.2. TOA/TDOA-Based Positioning Principle

TOA and TDOA are alternative positioning techniques applicable to UWB and Wi-Fi RTT ranging measurements, both requiring precise time synchronization. As illustrated in Figure 5a,b, both methods involve recording the response time of each base station upon receiving a signal from the mobile device to determine the distance. Successful positioning necessitates at least three base stations. The key difference between TOA and TDOA is that TOA also requires knowledge of the signal’s transmission time to directly calculate the distance between the mobile device and each base station, necessitating precise time synchronization between the transmitter and all base stations. Since both methods heavily rely on accurate time acquisition between the base station and the mobile device, any clock instability can significantly affect the precision of time measurements, thereby impacting overall positioning accuracy.

Utilizing multiple ranging measurements, several observation equations are formulated to resolve the position of a mobile carrier (e.g., a tag or smartphone). In a simplified two-dimensional scenario, assuming the carrier transmits the signal at time $t_0$, and the coordinates of three base stations are represented as $(x_i,y_i)$ with corresponding response times $t_i(i=1,2,3)$ and considering the speed of light as $c$, the TOA observation equation is as follows:

$$\sqrt{(x-x_i{)}^2+(y-y_i{)}^2}=c(t_i-t_0),i=1,2,3$$

(2)

where $(x,y)$ is the coordinate the mobile carrier to be resolved.

The method described above is known as the trilateration algorithm, which essentially involves finding the intersection point of three circles, as shown in Figure 5a. However, due to the inherent errors in measurements, the least squares solution of the equation is commonly employed for more accurate results.

As shown in Figure 5b, TDOA does not require the synchronization of the clocks between the mobile carrier and base stations, and it requires only the synchronization between several fixed base stations, which significantly reduces the technical complexities compared to the TOA approach. In a simplified two-dimensional scenario, the observation equation for the TDOA algorithm is as follows:

$$\left\{\begin{array}{c}\sqrt{(x-x_1{)}^2+(y-y_1{)}^2}-\sqrt{(x-x_2{)}^2+(y-y_2{)}^2}=c(t_1-t_2)\\ \sqrt{(x-x_1{)}^2+(y-y_1{)}^2}-\sqrt{(x-x_3{)}^2+(y-y_3{)}^2}=c(t_1-t_3)\end{array}\right.$$

(3)

where all symbols are same as those in Equation (2).

2.3. Ranging Error Optimization

The TOF, TOA, and TDOA ranging methods can all result in significant ranging inaccuracies due to hardware, environmental factors, or signal delay errors and communication errors during signal transmission. To reduce the impact of system errors on ranging measurements, it can be determined by averaging multiple statistical measurement results [28,29]. In addition, there are generally two methods to correct these system errors, including hardware compensation and data compensation. Hardware compensation involves extending the signal propagation time in the hardware to offset system errors, which can reduce ranging errors due to relative position changes between the router and mobile device to a certain degree. However, this method may introduce errors over time with hardware aging and poses challenges related to device storage and usage [30]. On the other hand, data compensation corrects the systematic errors during data processing, and it provides stable correction performance without additional costs. Although data compensation may struggle to compensate for errors caused by relative position changes, it provides consistent and reliable correction performance.

The geometric positioning principles of both UWB and Wi-Fi RTT require multiple base stations. The geometric layout of these base stations may also affect the ranging errors. For complex NLOS situations, when optimizing the layout of UWB base stations, researchers often use indicators such as dilution of precision (DOP) or geometric DOP (GDOP) to evaluate positioning accuracy and enhance the performance of UWB systems [31]. Using DOP as an evaluation indicator for base station placement, a significant correlation between DOP and positioning accuracy was confirmed [32]. The effect of antenna direction on positioning accuracy was further studied and demonstrated through multi-base station simulation [33]. A well-designed layout can significantly improve accuracy, especially when the distance between the target area and each base station is uniformly distributed. Refs. [34,35] performed GDOP simulations using small UWB antennas in various typical layouts. The algorithm based on position DOP (PDOP) optimizes the deployment of UWB base stations, breaking through the limitation of only studying typical layout methods [36], and proposes a three-stage UWB base station layout design for underground parking lots, providing a scientific process reference for underground parking lots [37]. Mathematical modeling and genetic algorithms are used to optimize PDOP and base station sharing value as the objective function, providing an innovative method for deploying UWB positioning networks. It improves positioning accuracy and is also applicable to different environments.

2.4. Comparison of UWB and Wi-Fi RTT Positioning Technologies

UWB and Wi-Fi RTT represent two prospective radio positioning technologies for indoor environments. Although both technologies are based on the same geometric positioning principles, they have their respective advantages due to their own signal characteristics and application ecosystem. UWB indoor positioning has the following advantages:

High positioning accuracy: UWB signals have a time resolution of the picosecond level, enabling centimeter-level positioning accuracy.

Low positioning latency: UWB signals have a low latency of millisecond or even microsecond level.

High robustness: UWB signals demonstrate high resilience to multipath effects, obstacles, and interference, offering more stable positioning performance compared to other wireless technologies. However, it is important to note that when significant obstacles, reflections, or strong multipath effects are present in the environment, UWB signals may still experience distortion or delays, particularly in dense urban or indoor environments with complex infrastructure.

On the other hand, Wi-Fi RTT positioning, based on the IEEE 802.11mc protocol, has several advantages:

High compatibility: Wi-Fi RTT is integrated with Wi-Fi network devices, including smartphones, which are used as mobile carrier; hence, Wi-Fi RTT positioning has high accessibility and applicability.

High security: Wi-Fi RTT protocol leverages Wi-Fi network encryption and authentication mechanisms to guarantee positioning security, effectively preventing from spoofing or replay attacks.

Low positioning cost: Wi-Fi RTT leverages existing Wi-Fi network infrastructure, such as routers and mobile phones; thus, it requires no additional purchases or dedicated deployment of hardware. However, it is worth noting that the number of devices that support the IEEE 802.11mc Wi-Fi protocol is limited, but it is gradually increasing. A few manufacturers provide Access Points (APs), phones and other devices that optimize support for Wi-Fi RTT, but it may take several years to achieve full popularization.

Table 2. Summary of devices supporting Wi-Fi RTT.

Category	Manufacturer	Product	Band/Version Supporting FTM RTT
Access Points	Google	Nest Wifi Pro (Wi-Fi 6E)	up to 6 GHz
		Google Wi-Fi	up to 5 GHz
		Google Nest Wi-Fi Router	up to 5 GHz
		Google Nest Wi-Fi Point	up to 5 GHz
	Aruba	AP504, AP505, AP514, AP515, AP518, AP503H, AP505H, AP534/AP535, AP555, AP565, AP575	up to 5 GHz
		AP61x	up to 6 GHz
	Cisco	9130, 9136, 9164, 9166	up to 6 GHz
	Compulab	Compulab WILD AP	up to 5 GHz
Phones	Google Pixel	6, 6 pro, 5, 5a, 5a 5G, 4 XL, 4, 4a, 3 XL, 3, 3a XL, 3a, 2 XL, 2, 1XL, 1	Android 9.0+
	Xiaomi Mi	10 Pro, 10, 9T, 9, Note 10, Note 10 Lite, CC9 Pro	Android 9.0+
	Xiaomi Redmi	Mi 9T Pro, Note 9S, Note 9 Pro, Note 8T, Note 8, K30 Pro, K20 Pro, K20, Note 5 Pro	Android 9.0+
	LG	G8X ThinQ, V50S ThinQ, V60 ThinQ, V30	Android 9.0+
	Samsung Galaxy	Note 10+ 5G, S20+ 5G, S20+, S20 5G, S20 Ultra 5G, S20, Note 10+, Note 10 5G, Note 10, A9 Pro	Android 9.0+
	POCO	X2	Android 9.0+
	Sharp Aquos	R3 SH-04L	Android 9.0+
Retail, Warehousing and Distribution Center Devices	Zebra Handhelds	TC52, TC52x, TC72, TC57, TC57x, TC77, TC83. EC30, EC50, EC55, MC3300x, MC9300, PS20 TC52ax, MC3300ax, TC52-HC, TC52x-HC TC52ax-HC	up to 5 GHz, Android 10.0+
	Zebra Wearables	WT6300	up to 5 GHz, Android 10.0+
	Zebra Tablets	ET51, ET56, L10A *ET40, *ET45 *ET40-HC, *ET45-HC	up to 5 GHz, Android 10.0+
	Zebra Vehicle Mounted and Concierge	VC8300, CC600, CC6000	up to 5 GHz, Android 10.0+
	Zebra Handhelds	TC53, TC53e, TC53e-RFID, TC73, TC58, TC58e, TC78, MC3400, MC9400, MC9450, PS30 *TC22, *TC27, HC50, *HC20	up to 6 GHz, Android 10.0+
	Zebra Wearables	WT6400, WT5400	up to 6 GHz, Android 10.0+
	Zebra Tablets	ET60, ET65	up to 6 GHz, Android 10.0+
	Skorpio	X5	Android 10.0+

lists the existing devices supporting Wi-Fi RTT [38,39].

In addition to hardware support, researchers have developed several applications to explore Wi-Fi RTT capabilities using Android smartphones. Google’s initial Wi-Fi RTT-based application, WiFiRttScan, enables the discovery of nearby RTT-enabled APs and estimates distances through the Wi-Fi RTT Application Programming Interface (API) [40]. This application can display and log a variety of details about the discovered APs. Based on WiFiRttScan, the authors of [28] developed a new application with additional features, such as sampling and displaying the RTT of nearby APs. However, a major limitation of both applications is their inability to simultaneously handle readings from multiple nearby RTT-enabled APs. Compulab addressed this with its WILD Minimal application [41], which can read information from multiple APs at once. Then, Google released WiFiRttLocator in 2021, a new application to demonstrate indoor positioning based on Wi-Fi RTT [42]. This application not only provides distance estimates but also calculates location information using at least three APs. However, it requires a more complex setup before starting a Wi-Fi scan, including mapping the floor plan and obtaining precise global coordinates for each AP. In 2022, Horn introduced a new Android application called FTMRTT [43]. In FTM RTT, the location of APs can be set using any convenient Cartesian coordinate system, making configuration easier than in WiFiRttLocator. Additionally, the application supports more advanced features, such as displaying measurement errors and working with non-cooperative APs in unilateral RTT mode [44]. Recently, WifiRttScanX was developed in 2023 based on the WiFiRttScan and FTMRTT, which can estimate the unknown offset in FTM RTT results [45].

3. NLOS Ranging Error and Correction Methods of UWB and Wi-Fi RTT Signals

In typical indoor positioning environments, UWB and Wi-Fi both suffer largely from NLOS and multipath effects. NLOS signals may cause significant ranging errors, and NLOS identification and correction have been widely recognized as research topics.

3.1. NLOS Signal and Identification

As wireless radio signals, both UWB and Wi-Fi RTT signals are susceptible to NLOS errors, a common challenge encountered in all wireless signal transmission [46]. NLOS errors arise due to signal diffraction caused by obstructive elements in the environment, resulting in signal propagation distances that exceed the actual straight-line distance. Despite UWB signals possessing robust anti-interference capabilities, UWB positioning outcomes are still impacted by NLOS errors. These errors can lead to deviations of several tens of centimeters in UWB ranging measurements, thereby degrading the accuracy of UWB positioning to the decimeter level.

Indoor environments are filled with obstacles, walls, people, and infrastructure, which disrupt electromagnetic wave distribution and pose challenges for wireless positioning due to the prevalence of NLOS signals. These signals arise from radio propagation phenomena such as multipath effects, obstructions, reflections, refractions, and scattering. For instance, UWB or Wi-Fi signals interacting with surfaces and objects lead to multiple signal paths, causing delays and distortions in time-of-flight measurements.

To address NLOS challenges, it is crucial to recognize UWB and Wi-Fi tags and base stations under such conditions while implementing measures to mitigate their effects. Advanced algorithms, such as machine learning for identifying NLOS signal patterns and data fusion techniques, enhance positioning robustness [47]. Moreover, high-precision sensors and augmented reality technologies can reduce positioning errors caused by NLOS, ensuring greater accuracy and reliability.

3.2. NLOS Ranging Error Mitigation

Given the physical attributes of NLOS errors and the impact of substantial sample collections on real-time positioning performance, current research focuses on NLOS error mitigation through real-time NLOS/Line-of-Sight (LOS) identification. The Wi-Fi RTT protocol, known for generating systematically biased ranging measurements, requires calibration for precise ranging and positioning outcomes. Studies indicate that Wi-Fi RTT exhibits larger ranging errors in NLOS scenarios compared to LOS conditions [48,49]. Various methods have been reported to enhance ranging accuracy using RTT measurements [50] or multiple RTT measurements from diverse frequency bands [51]. These identification and correction methods primarily include:

Threshold-based identification: The threshold method compares specific feature parameters with a predefined threshold to identify NLOS signals and make corresponding corrections. In [52], a threshold was implemented to identify the positioning error caused by NLOS. First, the initial tag position coordinates were obtained by using the Chan algorithm using the quadratic weighted least squares method. Then, a suitable threshold was set to compare with the difference in distance. The results greater than the threshold were iteratively optimized by the particle swarm algorithm and then output. The results lower than the threshold were directly output. Although these methods are easy to implement, their disadvantages are that the threshold selection depends on the environment and scenario and is sensitive to noise.

Statistical models: A variety of statistical models, which are based on the probability of NLOS propagation, have been reported for identifying NLOS instances and improving the ranging measurements. In [53], a constrained weighted least squares algorithm was developed for TDOA positioning in the presence of NLOS signals. A quadratic programming algorithm was reported for improving TOA positioning [54]. Another approach first constructs a scene recognition model based on Gaussian process regression to identify NLOS and LOS measurements, subsequently establishing an error calibration model for ranging correction. Owing to the frequency diversity, measurements from various frequency channels are considered unrelated and they are weighted and averaged to effectively mitigate NLOS error [51]. Notably, a kernel function was employed to reduce noise, significantly enhancing the robustness [55]. A ranging error model based on the Gaussian mixture model effectively addresses non-Gaussian ranging errors caused by multipath effects [56].

Bayesian filtering: As ranging measurements are time series, the Bayesian filtering methodology has been utilized for enhancing the accuracy and robustness of ranging measurements in the presence of NLOS conditions. The Extended Kalman Filter (EKF) was utilized to minimize the effect of NLOS errors in TDOA positioning solutions [57]. The authors of [58] introduced the optimization concept and fused TDOA data using a simulated annealing algorithm, while [59] reported a joint TOF and TDOA algorithm to improve the positioning accuracy. Several studies explored the data fusion using the Bayesian filtering approach, for example, fusing either UWB or Wi-Fi RTT with IMU data. In [60], the EKF was utilized for data fusion and CIR data were leveraged to identify NLOS measurements. The Bayesian filtering approach has been applied to locate mobile devices using a set of ranging measurements, and measurements of inertial sensors are integrated for improving the positioning accuracy [61,62,63,64,65].

Machine learning: Machine learning models have been proven to be an efficient approach for identifying NLOS signal patterns in received signals [66]. The authors of [9] developed a machine learning method that can identify NLOS signals of Wi-Fi using data from one second. This study highlighted that the proposed machine learning method segmented the samples for NLOS/LOS conditions, achieving over 96% discrimination accuracy with a sample size of 10. The authors of [67] summarized three machine learning methods, including random forest, least squares support vector machine, and deep neural network, which are effective for real-time NLOS/LOS identification leveraging Wi-Fi RSSI and RTT ranging measurements. Recent advancements include the utilization of a neural network for processing raw measurements of Wi-Fi RTT [68]. Ref. [69] reported a convolutional neural network (CNN) using CIR with NLOS mitigation, which achieved a mean error of 65 cm with three anchors under NLOS conditions. An adaptive NLOS mitigation method was reported in [70], which used and optimized machine learning models, including deep neural network (DNN), CNN and long short-term memory (LSTM), and achieved a mean error of 26.1 cm in complex NLOS indoor environments. Effectively addressing these NLOS challenges is imperative for improving the accuracy and reliability of indoor positioning systems based on UWB and Wi-Fi RTT technologies. Regardless of biased raw distance measurements or varying measurement patterns across multiple indoor sites or between LOS and NLOS conditions at the same site, the neural network adapts to these patterns flexibly and effectively enhances the ranging accuracy. These methods adapt effectively to complex environments and scenarios. However, the machine learning approach requires a large volume of labeled data and computational resources, which are the main challenges to be addressed when applying these methods.

Fingerprinting: Fingerprinting positioning is an effective approach for indoor positioning. It first maps the environment of interest and constructs the radio map or what is named signal fingerprint, which reflects the spatial distribution of radio signals. The effects of NLOS and multipath signals have been represented implicitly. When the radio map is used for location resolution in the subsequent online positioning stage, the positioning observables include NLOS components similar to those in the radio map, which are mainly related to the environment and are assumed to remain statistically correlated at both stages of radio map construction and online positioning. Hence, NLOS conditions provide useful information for the fingerprinting approach. In [71], trilateration and fingerprinting were integrated to improve the positioning performance in NLOS conditions, and [72] reported an UWB indoor positioning method based on LOS/NLOS signal mapping, which utilized prior environmental maps to tackle severe NLOS interference in intricate indoor settings. In general, the fingerprinting approach yields relatively high localization accuracy under NLOS conditions, whereas it may be less effective in near-LOS scenarios unless the radio map includes sufficiently dense fingerprints, which, however, incurs a large offline labor cost.

Optimization: This approach employs optimization algorithms to mitigate the impact of NLOS signals, such as convex optimization, particle swarm optimization, genetic algorithms, etc., and the optimization outcome is utilized to assess the presence of NLOS signals and conduct corresponding correction. For instance, [73] reported the method of tightly integrating IMU and multiple UWB data using the graph optimization model. This kind of method excels in fully utilizing information from CIR data; however, these methods entail high computational complexity and may encounter the local optima problem [74,75]. It is particularly challenging to deal with multipath effect due to their intricate propagation modes, leading to varying ranging results even across different channels within the same environment. To address this issue, a non-convex optimization method considers the spatial constraints of virtual positioning clients to effectively solve the optimal estimation problem [76].

4. Advanced Positioning Methodologies Using UWB and Wi-Fi RTT Signals

4.1. Integrated Positioning of Data Fusion

Positioning methods that rely on a single measurement method have difficulty achieving optimal accuracy, reliability, and continuity in different scenarios. Factors such as signal blocking, attenuation, interference, and reflection may degrade positioning performance or even cause complete failure. Fusion positioning provides an effective solution to improve positioning performance by integrating multiple data types, usually from different sensors.

Data fusion methods can be divided into two types: loosely coupled and tightly coupled. Loosely coupled methods calculate positioning results based on each sensor independently and integrate these results into an optimal solution through weighted fusion techniques. Although loosely coupled methods are easy to implement and highly adaptable, they usually result in low accuracy and insufficient information utilization. Tightly coupled methods fuse information directly in the measurement domain to maximize information utilization and thus improve accuracy. However, this method is relatively complex and may be less robust, especially in complex NLOS environments. Several studies have explored fusion positioning solutions that combine UWB or Wi-Fi RTT with other data sources. Fusion methods include Kalman filtering, Bayesian filtering, particle filtering, least squares, maximum likelihood estimation, factor graph optimization, etc. These methods process data from different signal sources by weighting, optimization, estimation, and calibration to improve positioning accuracy and robustness.

Table 3. Summary of Wi-Fi RTT-based integrated positioning studies.

Year	Sensors	Fusion Category	Algorithm	Platform	Experiment	Accuracy (m)	Limitations
2019 [64]	IMU, Wi-Fi RTT	Loose coupling	Enhanced particle filter	Smartphone	An office room of 11 m × (12.4 m/10 m) × 3 m	<1 m in 86.7% of the cases	Large amount of calculation and poor stability
2019 [65]	IMU, Wi-Fi RTT	Loose coupling	PDR and Wi-Fi RTT are integrated by UKF	Smartphone	An office room of 12 m × 12 m; A shopping mall	<2 m	Long-term localization performance needs to be improved
2020 [77]	IMU, Wi-Fi RTT	Tight coupling	Weighted least squares method and EKF	Smartphone	An office room of 20 m × 6 m	0.68 m in 80% of the cases	The method is unstable when the Wi-Fi RTT measurements change drastically caused by multipath
2021 [78]	IMU, Wi-Fi RTT	Loose coupling	A fusion-tracking federated filter	Smartphone	The two areas of 10 m × 10 m and 6 m × 4 m; An office room of 32 m × 21 m	<1 m in about 80% of the cases	Poor performance in large scenes
2022 [79]	IMU, Wi-Fi RTT	Tight coupling	Error-state Kalman filter; Rauch–Tung–Striebel smoothing for localization optimization	UAV with IMU of the smartphone	An office room of 5 m × 4 m; A U-shaped public area with length of 13 m	1.36 m in office room; 0.92 m in public area	Unknown performance in large scenes
2023 [80]	IMU, Wi-Fi RTT	Tight coupling	Adaptive extended Kalman filter	Robot with IMU of the smartphone	An office building with LOS and NLOS experiments	0.8 m in 80% of LOS cases; 1.04 m in 80% of NLOS cases	The robot speed is unknown
2023 [81]	IMU, Wi-Fi RTT/RSS	Tight coupling	Factor graph optimization	Smartphone	Two real areas of 126 m2	0.39 m	Three pre-set phone modes
2024 [9]	Wi-Fi RTT, Wi-Fi RSS	Loose coupling	A nonparametric regression	Smartphone	A passage and a classroom environment of 112.5 m2	1.15 m in 90% of the cases	Further validation of the method across diverse experimental settings is needed

and

Table 4. Summary of UWB-based integrated positioning studies.

Year	Sensors	Fusion Category	Algorithm	Platform	Experiment	Accuracy (m)	Limitations
2018 [82]	Monocular camera, UWB	Tight coupling	Joint nonlinear optimization on Lie-Manifold	Vehicle	The EuRoC MAV datasets	0.036	The cost of generating global maps is high
2021 [83]	IMU, UWB	Loose coupling	Deep learning-based speed estimation; KF-based position integration	Smartphone	An area of 5 m × 31 m	0.21 m	The design trajectory is simple
2021 [84]	IMU, UWB	Tight coupling	A complex Kalman filter for integration	Smartphone	A closed area of 9 m × 18 m	/	No NLOS impact
2021 [85]	Camera, IMU, UWB	Tight coupling	Mutual-information-based residual optimization	Robot	An indoor environment with the dimension about 2.70 m × 1.80 m using a mobile robot	0.16	The environment is relatively simple with fixed UWB anchors
2021 [86]	mono-VIO, UWB	Tight coupling	Residual-based optimization	UAV	Simulates the UAV flights using Gazebo and RotorS	<0.2	Pursuing real-world experiments
2021 [87]	VIO, UWB	Tight coupling	Residual-based optimization	Robot	“Loop” tests tin a 6 m × 6 m indoor area; “Open” tests in a 30 m × 10 m outdoor area	0.06~0.14	Only-one-robot scenarios, and multi-robots scenarios need to be extended
2021 [88]	VIO, UWB	Tight coupling	Graph optimization	Robot	An indoor environment with the dimensions of about 2.70 m × 1.80 m using a mobile robot	<0.2	The method considers an ideal environment considering only a situation where the UWB signal is a LOS
2023 [89]	UWB, Visual-inertial odometry (VIO)	Loose coupling	Distributed pose graph optimization	UAV	An indoor environment with dimensions of 5 m × 4 m × 2.5 m using two UAVs	0.2	The cost of generating global maps online is unacceptable. The bandwidth for communication is high
2024 [90]	IMU, UWB	Loose coupling	An Error-State-Kalman-Filter-based real-time fusion	Electric vehicle	A tunnel of 140 m × 5 m × 4 m	<0.4 m in LOS cases with a moving speed of 36 km/h	No experiment in the NLOS cases
2024 [91]	Vision SLAM, UWB	Tight coupling	EKF with threshold detection and adaptive measurement noise estimator	Robot	Obstacle-free and obstacle-rich environments	0.076~0.082	The method overlooks the small spatial offset generated using the UWB label device on the mobile platform. The experiments are conducted in a small field, leading to fewer accumulated errors

list the research on the fusion of Wi-Fi RTT with other sensors and the research on the fusion of UWB with other sensors in recent years, respectively. IMU is robust in the relative positioning of vehicle odometers or pedestrian odometers, and can effectively supplement the absolute positioning of UWB or Wi-Fi RTT signals. In addition, vision-based odometer technology can also be combined with UWB or Wi-Fi RTT, which is particularly suitable for indoor applications of unmanned aerial vehicles (UAVs) and robots.

Among these studies, the mainstream positioning fusion methods are Kalman filtering and factor graph optimization. Kalman filtering is a recursive method based on current observation data and previous state estimation, which can efficiently realize real-time state estimation. EKF and unscented Kalman filtering (UKF) are commonly used nonlinear filtering methods in fusion positioning systems. EKF approximates by linearizing the Jacobian matrix, while UKF uses unscented transformation to improve accuracy under strong nonlinear conditions. Adaptive Kalman filtering can dynamically adjust noise parameters to enhance robustness, but the computational resource requirements are high. However, Kalman filtering has strict assumptions on the noise model and usually requires the error distribution to be Gaussian, which makes it less robust in abnormal situations such as NLOS and multipath effects. At the same time, when dealing with nonlinear systems, the linearization of EKF may introduce errors, and although UKF has improvements, it increases the computational complexity. In this case, the accuracy can be improved by replacing the Gaussian assumption with a more flexible model, such as a Gaussian mixture model or a robust error model. In addition, adaptive filtering techniques that dynamically adjust noise parameters based on environmental conditions can also help improve performance in dynamic and unpredictable environments [92,93,94].

Factor graph optimization makes full use of time correlation information by constructing and optimizing a factor graph containing current and historical data. Factor graph optimization gradually approaches the optimal solution through multiple iterations, especially when the measurement distribution assumption is not fully met, as it can significantly improve positioning accuracy. Due to its flexible error modeling capabilities, it can handle non-Gaussian distributions and complex environments, and effectively reduce linearization errors through multiple iterative optimizations in nonlinear scenarios [95]. However, factor graph optimization has high computational complexity, and batch processing limits its real-time performance. It requires technical optimization such as sliding windows or marginalization to reduce the amount of data processed each time, thereby improving the efficiency of real-time applications. In addition, factor graph optimization is sensitive to initial values and may fall into local optimality. Through smarter initialization strategies, such as machine-learning-based methods or heuristic algorithms, the risk of falling into local optimal solutions can be reduced and robustness can be improved. In addition, integrating real-time adaptive error models into factor graph optimization can also improve the adaptability of factor graph optimization in response to environmental changes.

4.2. Autonomous Calibration of Base Station Coordinates

With the TOA and TDOA principles, UWB- and Wi-Fi RTT-based geometric positioning requires known coordinates of base stations, as shown in Equations (1)–(3). Hence, these systems need to calibrate the coordinates of base stations of either UWB or Wi-Fi RTT when they are deployed in the field. The calibration process is often cumbersome and labor-consuming. Furthermore, the requirement of calibrating the base stations limits the positioning systems only working within the coverage area of these calibrated base stations. Methods have been reported to estimate the unknown positions of base stations online by jointly integrating complementary data of multiple sensors [82,96,97,98,99]. These methods not only alleviate the tedious tasks of measuring and calibrating base station coordinates in advance, but also enable new applications such as dynamic expansion and deployment of UWB or Wi-Fi RTT base stations, for example, in unknown complex environments where it is difficult to calibrate the coordinates of base stations manually.

The complementarity of different sensor data enables the inverse estimation and calibration of the coordinates of base station positions at the initialization stage. Alternatively, the unknown coordinates of base stations are directly incorporated as state variables to be estimated together with the state parameters of mobile tags’ positions during the online positioning process. Refs. [96,97] reported successive joint estimation methods for tightly coupling vision–UWB combinations, effectively handling unknown positions of three base stations as additional state variables. The first method [96] and second method [97] involved joint estimation of the scale factor and positions of UWB base stations, requiring initial values for the additional states. The former method eliminated prior error considerations but enhanced the Levenberg–Marquardt algorithm by constructing an explicit square error function. It then added the UWB measurement’s residual term to the reprojection error for tight coupling, as illustrated in Figure 6. The latter constructed a ranging error and prior error equation, where the scale factor and positions of UWB base stations were both included in the error equation, solving it through minimization using the Levenberg–Marquardt method. VIO was utilized for precise positioning during the initialization, and the trajectory in the world coordinate frame was established. The distances between base stations and the trajectory were determined by UWB ranging information, forming an optimization problem on the data window. The optimization process was terminated when the position uncertainty fell below a threshold, determining the positions of base stations. It should be noted that it increases the number of estimated states by introducing the positions of base stations as additional states, and it also generates geometric constraints on motion trajectories, which enhances the solution robustness.

Alternatively, the authors of [98] reported a two-layer sliding window algorithm based on a single base station. The study constructed a cost function introducing positions of UWB base stations in the estimated state vector during the initialization process, and the positions of base stations were fixed upon completion of the estimation and were considered invariant known variables in the subsequent navigation process. Using Wi-Fi RTT ranging measurements, the authors of [99] reported the method of jointly estimating the positions of Wi-Fi routers and user carriers together. The authors of [82] created a map consisting of visual features and anchor points using a monocular camera and UWB base stations. As a vehicle explored the unknown space of interest, new base stations were continuously deployed, and their positions were estimated online together with the vehicle position. This approach formed a UWB base station constellation to fulfill the task of exploring unknown environments and constructing maps. It is apparent that the reported methods above can work with either UWB or Wi-Fi RTT in principle, even though the original works employed only one type of measurements.

5. Discussion and Conclusions

This work provides a comprehensive review of UWB and Wi-Fi RTT indoor positioning technologies, with particular attention on three aspects: positioning principles, NLOS signal identification and ranging error correction, and advanced positioning methodologies. The discussion encompasses the strengths and weaknesses of UWB and Wi-Fi RTT positioning, with a particular emphasis on analyzing NLOS errors, their impact, and correction techniques. Moreover, it sheds light on recent research trends and future directions, aiming to guide students, young researchers and engineers in exploring innovative ideas in this field.

UWB and Wi-Fi RTT signals suffer from intricate propagation challenges, as environmental factors such as physical structures and materials influence signal propagation, especially in NLOS conditions characterized by reflections, diffractions, and the multipath effect. It is crucial to address these complexities for accurate ranging measurements. The variability in NLOS conditions causes another layer of complexity, making the development of universally applicable correction models a daunting task. Many existing correction methods may pose computational challenges, particularly with large datasets or real-time applications. Keeping a balance between accuracy and computational efficiency remains an ongoing endeavor. For instance, although machine learning approaches can extract valuable insights and classify data effectively, the data-processing efficiency needs to be optimized by determining the optimal training dataset size for specific machine learning algorithms, and it remains a task for future exploration.

Due to the physical limitation of a single type of sensors, positioning systems with either UWB or Wi-Fi RTT signals suffer from challenges regarding accuracy, reliability, continuity, etc. Advanced integrated positioning through data fusion is essential for improving the positioning performance. The fusion of relative and absolute positioning information derived from diverse sensors provides synergistic advantages, and it will be the dominant solution for prospective development. The convergence and integration of different sensors may result in novel integration models, such as hybrid neural network architectures incorporated with Kalman filtering.

In practical applications, both UWB and Wi-Fi RTT positioning systems face significant challenges in handling computational load and managing latency while keeping affordable system costs. UWB and Wi-Fi RTT positioning technologies typically require substantial processing power to handle the dense data from multiple sensors in real time. This can become a bottleneck, particularly in fast-paced environments such as automated guided vehicle (AGV) tracking in factories, where even a few milliseconds of delay can disrupt operations. The computational demands can exacerbate latency issues, impacting applications that require near-instantaneous location updates. To address these challenges, various strategies can be implemented to achieve low latency and high accuracy. For instance, edge computing can significantly reduce data transmission delays, while distributed computing architectures enable parallel processing across multiple nodes, enhancing efficiency. Additionally, optimizing positioning algorithms—such as simplifying Kalman filters or employing lightweight neural network models—can further lower latency and computational demands. Hardware acceleration is also crucial, with technologies like Field-Programmable Gate Array (FPGA), Graphics Processing Unit (GPU), or Application-Specific Integrated Circuit (ASIC) providing significant improvements in real-time processing performance. Moreover, optimizing wireless communication protocols, such as adopting low-latency Wi-Fi 6 or 5G networks, and dynamically adjusting data update frequencies based on system operational states allow for flexible resource allocation and improved overall performance. Furthermore, real-time positioning systems must carefully balance accuracy, computational efficiency and system cost, tailoring solutions based on the specific requirements of use cases, such as AGV tracking or indoor navigation, where low latency and high scalability are critical.