A Survey on STAR-RIS: Use Cases, Recent Advances, and Future Research Challenges

The recent development of metasurfaces, which may enable several use cases by modifying the propagation environment, is anticipated to substantially affect the performance of sixth-generation (6G) wireless communications. Metasurface elements can produce passive subwavelength scattering to enable a smart radio environment. Simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS), which refers to reconfigurable intelligent surfaces (RISs) that can transmit and reflect concurrently (STAR), is gaining popularity. In contrast to the widely studied RIS, which can only reflect the wireless signal and serve users on the same side as the transmitter, the STAR-RIS can reflect and refract (transmit), enabling 360° wireless coverage, thus serving users on both sides of the transmitter. This article presents a comprehensive review of the STAR-RIS, focusing on the most recent schemes for diverse use cases in 6G networks, resource allocation, and performance evaluation. We begin by laying the foundation for RIS (passive, active, and STAR-RIS), and then discuss the STAR-RIS protocols, advantages, and applications. In addition, we categorize the approaches within the domain of use scenarios, which include increasing coverage, enhancing physical-layer security (PLS), maximizing sum rate, improving energy efficiency (EE), and reducing interference. Next, we will discuss the various strategies for resource allocation and measures for performance evaluation. We aimed to elaborate, compare, and evaluate the literature regarding setup, channel characteristics, methodology, and objectives. In conclusion, we examine this field’s open research problems and potential future prospects.

Industry and academia have done considerable theoretical and field testing, and regional standards developing bodies have started standardization. At the October 2020 radio communication sector summit hosted by the International Telecommunication Union (ITU), it was emphasized how important RIS is to the physical layer of future generation networks [9] The 3rd generation partnership project (3GPP) began receiving RIS suggestions during the conference in March 2021 [10], and more industrial stakeholders joined to promote the RIS as it became a crucial part of future wireless networks [11]. European Telecommunications Standards Institute (ETSI), in June 2021, also set up a new RIS industry specification group [12]. Moreover, the China Communications Standards Association (CCSA) is developing normative standards and has authorized a request to establish a research item on RIS [13], with a technical report expected in 2022.
However, the consensus amongst published research is that RIS can, at most, only reflect the incident signal. Therefore, when the transmitter and receiver are on opposite sides of the RIS, the communication system cannot use the RIS. Researchers have proposed employing STAR-RIS, which stands for simultaneously transmitting and reflecting RIS (STAR-RIS), to overcome this limitation [14]. Unlike traditional passive RIS, each STAR-RIS element can simultaneously refract and reflect the incident signal, eliminating the need to limit deployment to certain geographic areas and allowing for full-space coverage, hence also known as intelligent omni-surfaces (IOSs) as shown in Fig. 1. Regarding improving the efficiency of wireless networks, STAR-RIS can provide SRE with more versatility thanks to the ability to tune transmission and reflection coefficients (TARCs).

A. Motivation and Contribution
Previous research articles have provided in-depth tutorials or surveys relevant to RIS from distinct viewpoints, as shown in Table II. However, the subject of the present paper is quite distinct. Particularly, Liang et al. [15] examined how RISs can be utilized as reflectors and compared them to backscatter communications and reflecting relays. Renzo et al. [16] presented the concept of RIS meta-surfaces as a means to facilitate SRE, delved into the many features of RIS, and surveyed the state of the art of meta-surfaces in SRE studies. In contrast, the focus of the paper [17] was on the emerging use cases of RISs. Smart cities' idea is investigated through RIS in [18] and emphasizes RIS implementation's possible benefits and implementation along with exciting research prospects. ElMossallamy et al. [19] and Wu et al. [20] researched channel characteristics and addressed the main challenges for RISassisted wireless communications. Long et al. [21] provided useful scenarios and important performance measures along with a new signal model, hardware architecture, and competitive advantages while in [22] possible future use cases, deployment techniques, and design considerations for RIS devices in underground IoT, underwater IoT, Industry 4.0 were examined and explored future research challenges. When analyzing the difficulty of incorporating RISs into wireless networks, Yuan et al. [23] mentioned three issues: 1) channel state information (CSI) acquisition; 2) low-complexity phase shift optimization; and 3) passive information transfer. Liu et al. [24] analyzed machine learning (ML) and resource allocation algorithms for RIS-assisted wireless communications. In [25], they gave a tutorial on RIS-enhanced wireless sensing and localization. Björnson et al. [26] investigated RIS-enhanced wireless communications from a signal processing perspective, while [27] examined channel estimation approaches. Aboagye et al. [28] focused primarily on the applications of RISs in visible light communication (VLC) systems. ML-based algorithms are discussed in [29] for RIS, along with an overview of the spectrum allocation intelligently in IoT systems. Also presented the integration of ML methods and discussed promising applications of existing integrated ML solutions to technical issues.
Latest papers [30], [31] have primarily focused on IOS, as opposed to the aforementioned works, which focus on reflecting surfaces exclusively. Zhang et al. [30] introduced the notion of IOS, which enables full-dimensional communications by servicing users on both sides of the surface. A new hybrid beamforming approach is proposed for wireless communications based on IOS. Zhang and Di [31] conducted a comprehensive IOS review considering future cellular network applications and design principles, which include beamforming, channel modeling, experimental implementation, and measurements. This study presents the first full overview of a more general category of RIS, STAR-RIS, and their applications in wireless networking.
Following are our review's key contributions. 1) We built on the STAR-RIS existing literature and systematized an up-to-date, complete survey of state-of-theart schemes enabling an SRE. For deeper comprehension, we highlight the fundamentals of several RIS kinds, such as passive, active, and STAR-RIS. Additionally, we examine the STAR-RIS operating protocols before reviewing its various B5G and 6G applications and their benefits.
2) We organize all schemes that harness STAR-RIS based on use cases, resource allocation, and performance assessment. Additionally, we provide an overview and further divide the STAR-RIS use cases category into subcategories based on coverage, physical-layer security (PLS), sum rate, energy efficiency (EE), and interference. Then, we investigate the techniques used for STAR-RIS resource allocation and performance evaluation. 3) Furthermore, at the conclusion of each subcategory, summary tables are provided so that the reader can gain a better understanding of the rational association between the various schemes, taking into account crucial factors, such as system model in different scenarios, direct, reflective, and refractive channel characteristics, and proposed solutions. 4) We list some open problems and interesting new directions for future study in this dynamic field. This survey can benefit researchers of all levels, from novice to seasoned. As a result, new opportunities for significant advancements in the underlying field can be explored. The remaining sections of the survey are arranged as follows. Section II provides an explanation of RIS and its various forms, such as passive RIS, active RIS, and STAR-RIS. The advantages, applications, and operational procedures of STAR-RIS are also covered. We categorize the approaches in Section III in accordance with the use case scenarios, which include expanding coverage more effectively, bolstering PLS, maximizing sum rate, improving EE, and reducing interference. We will then discuss various methods for allocating resources and measuring performance. We list some open issues and suggested future research opportunities in Section IV. Section V concludes the STAR-RIS review. The taxonomy of our survey is shown in Fig. 2.

II. STAR-RIS PRELIMINARIES
RISs are also known as intelligent reflecting surfaces (IRSs) and large intelligent metasurfaces (LIMs) [32], [33]. By reconfiguring the propagation environment of electromagnetic waves, RIS is expected to reduce wireless network power consumption and improve system efficiency. An RIS can be regarded as a programmable cluster with many scatterers, each comprising many subwavelength and conductive parts. Like a signal processing system, an RIS has inputs, outputs, and many parallel subsystems with reconfigurable transfer functions. According to their research, there are three types of RIS: 1) passive RIS; 2) active RIS; and 3) simultaneous transmission and reflection (STAR) RIS as shown in Fig. 3.

A. Passive RIS
EM materials are used to construct the passive RIS. Due to their low cost, RIS can be integrated into several compositions, including building facades, reconfigurable walls, high-altitude platforms, roadside billboards, highway polls, glasses, and pedestrian clothing [24]. The RIS can alter the environment for wireless transmission by accounting for long-distance power losses. BS and mobile user (MU) can generate virtual LoS lines by passively reflecting the received signals. Unlike typical relay systems, such as amplify and forward and decode and forward [34], instead of needing a power amplifier, passive  RIS can modify the input signal by controlling the phase shift of each reflector. This means deploying RIS is much more eco-friendly and energy-efficient than deploying conventional relay systems. Due to the fact that electromagnetic waves are only reflected, RIS also allows full-duplex (FD) and full-band transmission. In terms of complexity, passive RIS is generally less complex than active RIS since it does not require any active components or control signals. However, the actual complexity of an RIS system depends on various factors, such as the number of elements, the design of the elements, and the algorithms used to optimize the reflected signal.
Although passive RIS provides a reliable reflection link for signal transmission alongside the direct link, this reflection link always has a double fading effect, which means that signals received over this link have twice as much large-scale fading. In many situations where the direct link is robust, passive RISs can only contribute a small amount of capacity [35]. In contrast to a direct link, a transmitter-RIS-receiver link's equivalent path loss is the product (as opposed to the sum of path losses) [35]. Furthermore, signals from the comparatively long reflection link affect more than those from the shorter direct link in terms of power loss if the fading coefficient is high. This means that a system with RIS is only slightly better than one without it. To avoid "double fading" or "multiplicative fading," [36], [37], [38] have developed active RIS.

B. Active RIS
The multiplicative fading effect of passive RIS is a major performance bottleneck, and the concept of active RIS was introduced as a possible solution [36]. Active RIS, similar to passive RIS, can reflect incident signals with adjustable phase shifts. Active RIS can further amplify the reflected signals, unlike passive RIS, which merely reflects incident signals. Active RIS, in contrast to passive RIS, has a different hardware architecture [39]. Phase shift circuits and reflection-type amplifiers boost the signal strength in an active RIS. It is impossible to ignore the power requirements of active RIS because they may be comparable to those of the BS amplifiers. Since the overall power consumption of active RIS-assisted systems may be much higher than that of passive RIS-assisted systems [40].
Thus, active RIS changes the multiplicative channel loss of passive RISs into an additive form and adds an amplification gain, making it more effective than specular reflection alone. Using active and passive RIS components in a hybrid architecture could improve analysis and optimization. The power consumption of active RIS components could be an issue for RIS that is not connected to the power grid. Future research should carefully examine these challenges. An active RIS is a promising research topic because we expect it to surpass amplify and forward relay in cost and efficiency [21].

C. Simultaneously Transmitting and Reflecting RIS
For a traditional RIS to function, its transmitter and receiver must always be on the same side, as it can only reflect incident wireless signals. This results in the exploitation of a halfspace SRE [41], [42], [43]. Frequently, users are placed on both sides of an RIS, which drastically limits its adaptability and effectiveness. To overcome this issue, Liu et al. [14] and Zhang et al. [30] suggested innovative ways of simultaneously reflecting and transmitting signals leveraging STAR-RIS. The surface of STAR-RIS separates incoming signals into two distinct components. To achieve 360 • coverage, a portion of the signal is reflected in the reflection region, and the rest is transmitted. RIS was originally conceived as wall-mounted or building-front-mounted devices. RIS can be positioned inside a wall or in the center of a communication area to receive and transmit signals using various techniques. From a signal processing standpoint, this RIS fits the profile of a single-input, dual-output system.
In contrast, the system communication protocol must consider various factors, including energy, mode, and time. Other considerations include the sophisticated hardware composition and perhaps the unique appearance of such RISs, which may be the cost of obtaining 360 • wireless coverage. Moreover, these surfaces can enhance coverage from one room to another and from the outdoors to the inside of a building. Coverage on both sides of the RIS could be reduced due to the multiplicative path-loss effect, which is why STAR-RIS design is still a concern. Developing an active STAR-RIS might be crucial to attaining comprehensive coverage in this setting. 1) A major advantage of STAR-RIS is its ability to transmit and reflect incident signals, allowing them to cover the entire space and service both sides with a single RIS. 2) Since STAR-RISs provide additional Degrees of Freedom (DoF) for modifying signal propagation, the design's flexibility is increased, allowing it to meet even the most rigorous communication requirements. 3) With its optical transparency, STAR-RIS may be used in windows and has a pleasant aesthetic, both of which are important for real-world applications.

3) Application of STAR RIS:
After reviewing STAR-RIS benefits, we examine various prospective wireless communication network applications [30]. STAR-RIS can improve wireless network coverage and quality by overcoming obstacles, such as buildings, and trees along roads, automobiles, etc.

III. STAR-RIS EMPOWERED USES CASES, RESOURCE ALLOCATION, AND PERFORMANCE ANALYSIS
This section will first focus on how the STAR-RIS influences communication performance based on different use cases. Specifically, we categorize the available literature for different use cases, which include enhancing coverage, improving PLS, enhancing sum rate, improving EE, and mitigating interference. We will review the available literature on STAR-RIS resource allocation and performance analysis.

A. Enhancing Coverage Leveraging STAR-RIS
As mentioned before, a STAR-RIS can serve users on both sides. Therefore, expanding the range of cellular network coverage is one of the most important applications for the STAR-RIS, as shown in Fig. 4. For instance, a STAR-RIS deployed near the cell edge can enhance users' performance within cell coverage and provide service to users outside of cell coverage. This section investigates the state-of-theart strategies for increasing wireless communication coverage leveraging STAR-RIS for different scenarios. It is supplemented by summary Table IV to enable readers to comprehend the primary principles and have a more thorough intuition about the associated technology.
Reflecting and transmissive signals may have distinct channel models and power, making IOS-assisted communication challenging. In such a situation, IOS-assisted communications cannot easily be applied to reflective IRS-assisted communications studies. Furthermore, the BS-user link and IOS-assisted communication system's reflecting and transmissive links can coexist. Therefore, the IOS phase shift design should consider multiple communication connection superposition. In this article, Zhang et al. [44] examined an IOS-assisted downlink communication system that improves the MU link quality by introducing an IOS phase shift configuration. Unlike IRS in most existing systems, IOS can transmit or reflect signals to the MU, boosting wireless coverage. To maximize MU downlink spectral efficiency (SE), an IOS phase shift optimization problem and a branch-and-bound technique for constructing the best IOS phase shift within a finite set are provided. Simulations have proven that the IOS-assisted system can transmit signal across a wider area than the IRS-assisted system.
Papazafeiropoulos et al. [45] investigated STAR-RIS as a means of enabling mMIMO. The coverage capability of a STAR-RIS-mMIMO system was characterized by closedform expressions that consider phase-shift errors and correlated fading. Intriguingly, phase setting occurs by enhancing the coverage probability for every few synchronization gaps because it requires statistical CSI in massive data sets. Consequently, in the case of STAR-RIS networks with rapid CSI, they could construct passive beamforming with fewer complications but at a higher cost. The numerical results highlight characteristics, such as the effect of RIS element influence and phase errors and confirm the superiority of STAR-RIS over RIS. Future research could focus on Ricean channels and possibly mmWave transmission.
To ensure complete coverage in all directions, Zhang et al. [46] analyzed a downlink NOMA network that included STAR-RIS support and randomly deployed users. The authors designed two STAR-RIS-assisted channel models, one for situations with many RIS elements (i.e., the Central limit) and one for settings with multiple cells (i.e., the curve fitting). A Gamma distribution can be used to approximate the curve-fitting model closely, and closed-form expressions can be used to construct the error functions of the Central limit model. Additionally, closed-form outage probability (OP) expressions have been obtained for NOMA users. Numerical results show that the Central limit model is an upper bound, and the curve-fitting model is a lower bound for the N-pairs with no error bound. Moreover, the two-channel models reach boundaries in regions with a high signal-to-noise ratio (SNR) and match simulation findings well in parts with low SNR.
When it comes to SRE, STAR-RIS can span full-space coverage. Xie et al. [47] considered downlink NOMA multicell network with STAR-RIS, where incident signals at STAR-RISs are separated into two halves for transmitting and reflecting. The authors leverage the controllable Gamma distribution to approximate composite small-scale fading power. The location of RIS, BSs, and user equipments (UEs) is then provided using a unified mathematical method based on stochastic geometry. Moreover, this method calculates the typical UE and connected UE coverage probability and Ergodic rate (ER). For the coverage probability under interference-limited situations, closed-form equations are derived and also generated theoretical expressions in typical RIS-aided networks for comparison. Analysis indicates that there are optimal STAR-RIS ES coefficient values that maximize both system coverage and ER. The numerical findings show that STAR-RIS can manage UEs on both sides and provide higher coverage and throughput than typical RIS with proper ES coefficients.
Wu et al. [48] investigated STAR-RIS for OMA and NOMA and modeled a sum coverage range maximization problem. Individually, the resource allocation at the AP and the TARCs at STAR-RIS were optimized to meet the communication requirements of consumers. NOMA, a nonconvex decoding order constraint, was transformed into a linear constraint, thereby transforming the problem into a convex one that can be solved in the most efficient manner. Initially, it was demonstrated that the OMA optimization problem was convex for a specified frequency/time resource allocation. Then, for the best results, the fundamental search-based method was chosen. In comparison to conventional RISs, STAR-RIS can significantly increase coverage, according to the research.
Zhai et al. [49] proposed a novel STAR-RIS to improve computation accuracy in wireless devices across a wide coverage area. The authors proposed a joint beamforming design for optimizing the transmit power at the wireless devices, the passive reflect and transmit beamforming matrices at the STAR-RIS, and the receive beamforming vector at the fusion center to minimize the computation mean-squarederror (MSE). The authors derived closed-form solutions for the updates of the passive reflect and transmit beamforming matrices by introducing an auxiliary variable and exploiting the coupled binary phase-shift conditions. This article also provided theoretical and numerical evidence to support the effectiveness of the proposed beamforming design in improving computation accuracy in wireless devices.

B. Improving Physical-Layer Security Through STAR-RIS
Without resorting to higher-layer encryption, PLS is a reliable means of transmitting secret messages over a wireless channel. At the same time, Eves are present [50]. The basic idea behind this is to limit the amount of data an unauthorized receiver can extract by taking advantage of the randomness of noise and fading channels [51]. STAR-RIS can help boost PLS performance, e.g., a user on one side of the STAR-RIS can pick up signals from a transmitter on the other side, but this user's secret transmissions cannot be directed toward Eves on the same side as the user [52]. Therefore, the STAR-RIS channel enables secure communications, as shown in Fig. 5.
Despite this, STAR-RIS-enhanced PLS poses new research challenges. When a STAR-RIS is present, wireless channels become configurable. This generally necessitates a reevaluation of the PLS paradigm in light of the new design constraints that permit the input distribution of a system to be altered in response to the system's own states [53], [54]. Prior knowledge of the Eves in the context of PLS is typically difficult. Therefore, optimizing the STAR-RIS to guarantee the required level of security is challenging. Multiple users may share a single STAR-RIS in a multiuser system, and optimizing user competitiveness is a significant research problem. This section investigates the state-of-the-art strategies for improving PLS utilizing STAR-RIS considering single and multiple Eves. It is also supported by the summary Table V for enabling readers to grasp the primary techniques and get better intuition about the STAR-RIS-based PLS.

1) STAR-RIS Assisted PLS for Single Eve Scenario:
Combining NOMA with STAR-RIS, as Han et al. [55] claims, is a win-win strategy that can greatly improve coverage performance. A secure communication technique with artificial noise (AN) support was proposed to address the issue and improve the secrecy rate (SR). An alternating optimization (AO)-based technique was presented to find the best AN model and RIS settings. Methods of semi-definite relaxation (SDR) and successive convex approximation (SCA) are used in this algorithm. The proposed scheme performs better than the benchmark approaches regarding secrecy while requiring less AN power. Increasing the number of RIS elements allows the AN's power to be further lessened. Furthermore, increasing the number of transmit antennas decreases the AN power when the Eve is close to the transmitter but increases it when it is far away. While Han et al. [55] concentrated on the downlink scenario, Zhang et al. [56] looked at secure transmission in the uplink NOMA system with STAR-RIS assistance, wherein authorized maximize users proactively change the EM propagation environment to transmit confidential messages to the BS. The accessibility of the eavesdropping CSI was attributed to both the statistical CSI and the full CSI. Using an adaptive-rate wiretap code, they can maximize minimum secrecy capacity under successive interference cancelation (SIC) decoding order for the full eavesdropping CSI scenario. The authors introduced the alternating hybrid beamforming (AHB) algorithm to optimize transmit power, reflection/transmission coefficients, and receive beamforming. In the statistical eavesdropping CSI state, constant-rate wiretap code was used to minimize the maximum secrecy OP (SOP), subject to the Quality of Service (QoS) limits of legitimate users. Next, develop a better AHB algorithm for the joint secrecy beamforming design and use constant-rate coding to derive a precise SOP expression. The simulation results demonstrate that the proposed strategy is effective.
In contrast to NOMA-based works in [55] and [56], Fang et al. [57] investigated the performance of IOS in the context of PLS across MIMO-based communication networks.
In the presence of a multiantenna Eve, the IOS focuses on a specific case to improve the receiver's secrecy performance. Using AN-assisted beamforming increased additional security robustness even further. The block coordinate descent (BCD) optimization technique and the Lagrangian dual method were chosen to reduce the complexity of the AN-assisted beamforming design. Using quadratically constrained quadratic programming (QCQP), the problem of frequency change efficiency was resolved. Simulation validates IOS's superiority over RIS and validates the method's efficacy.
2) STAR-RIS Assisted PLS for Multi-Eve Scenario: STAR-RIS-assisted communication, according to Xu et al. [58], is an exciting topic of study since it can satisfy the strict requirements for increased spectrum, efficiency, and coverage quality. The study's main focus was the downlink STAR-RIS-NOMA system, which improves transmission quality between users and a multiple-antenna BS. Two NOMA users on either side of the projected STAR-RIS are served via the ES protocol. Each reconfigurable element can function simultaneously in transmission and reflection modes based on the ES protocol. Initially, the closed form of SOP was established to evaluate the STAR-RIS-NOMA system's secrecy performance. The performance of the generated SOP was then evaluated asymptotically. Thanks to the secrecy diversity order, additional insights might be extracted, created by the asymptotic approximation in the high SNR and main-to-eavesdropper ratio regimes. A subsequent advancement was the optimization of the system's parameters to lower the SOP. The analysis showed that the multiple-antenna BS did not affect the secrecy diversity order for the NOMA system supported by STAR-RIS. Theoretical findings and simulation results closely complement each other, and the SOP of the STAR-RIS-enhanced NOMA system was lower than that of the OMA system, according to simulations. Future research, however, will take the channel estimation error into account.
Different from the NOMA-based work in [58], Niu et al. [59] employed a STAR-RIS to increase security in a multiple-input single-output (MISO) network by studying the three protocols, i.e., TS, ES, and MS. By jointly designing TARCs and beamforming coefficients, weighted sum SR (WSSR) is maximized. At first, a path-following technique was created to transform the nonconvex problem into a convex one, and then the TARCs and beamforming were constructed in a way that allowed for flexibility. The TARCs for the ES scheme were then solved using the penalty concave-convex procedure. In addition, resolving a mixed-integer problem for MS and proposing a two-layer optimization strategy for the TS. Finally, the results of the simulation confirm STAR superiority.
Wang et al. [60], using the STAR-RIS system, looked into the problems of transferring e-health data quickly and safely via the Internet of medical things network. Patients' telemedicine transmissions were secured using the STAR-RIS to prevent eavesdropping. A joint active and passive beamforming approach was created to maximize secrecy EE (SEE) while considering the imperfect CSI of all channels. To estimate the semi-infinite inequality constraints, the reformulated problem was solved using an AO framework that used the S-procedure and general sign certainty. In places with low downlink power, the TS mode of STAR-RIS was favored, whereas the ES mode provided the highest performance in regions with sufficient downlink power. Without CSI evaluation's accuracy and bit resolution power usage, STAR-RIS's aids could not be acquired. The simulation findings show that STAR-RIS can increase SEE for the Internet of medical thing networks far more than RIS. In another work, Wang et al. [61] proposed a novel IOS-enhanced air secure unloading method to prevent security breaches, improve authentic receiving quality, and increase the safety installation area of unmanned aerial vehicles (UAVs). By assigning computing frequency, determining offloading strategy, controlling transmit power, designing phase shifts, and organizing maximize UAV locations, a nonconvex resource allocation problem was constructed to maximize the SEE of the system. The interconnectedness of the variables made it hard to find a simple solution to the problem. Therefore, the original problem was divided into subproblems, and an iterative technique with low complexity was used to optimize the computation and communication settings. The findings proved that IOS could fully use UAVs' deployability flexibility, which enabled secure offloading. The SEE of the IOS-enhanced air secure offloading method significantly outperforms that of conventional systems. Unexplored were a few intriguing aspects, including multifunctional metasurface, CSI acquisition, resilient design, and the optimization and deployment of various IOSs.

C. Enhancing Sum-Rate Using STAR-RIS
NOMA is capable of satisfying the huge connection demands and high SE of the future generation networks [62]. NOMA permits several users to share a single resource block (RB) in the code or power domain instead of OMA, which serves users via orthogonal time/frequency RBs [63], [64]. At the transmitter, user signals are combined using power-domain NOMA and recovered using SIC. When users have distinct channels, NOMA outperforms OMA, as shown in Fig. 6. STAR-RIS-assisted NOMA systems can further maximize NOMA gain by reconfiguring wireless channels [65].
The state-of-the-art methods for increasing the sum rate using STAR-RIS are examined in this section. These methods are further classified into OMA and NOMA multiantenna/FD/vehicular domains. Additionally, the summary in Table VI is included to help readers understand the key concepts and gain a better insight into the STAR-RIS technology.
1) STAR-RIS-OMA-Based Sum-Rate Maximization: Wu et al. [66] stated that IOS as a component of RIS gained more attention due to its ability to serve UE on both sides of the metasurface continuously. Therefore, reflection and refraction signals must share the IOS's phase shift, resulting in the inevitable pairing of refraction and reflection beams. They suggested that bilayer-IOS (BIOS) offers flexible reflection and refraction beamforming to solve this issue. BIOs can completely control the refraction and reflection beam for UEs on both sides due to the proximity of two IOSs. With MISO, a BIOS was implemented in a multiuser system. The authors jointly optimize BS precoding and BIOS passive beamforming to maximize SE while utilizing AO. Future research into the capabilities of BIOS in more complex situations, such as MIMO, wideband, and multilayer IOS, would be quite intriguing. In another work, Niu and Liang [67] simultaneously designed base frequencies and TARCs to maximize the weighted sum rate (WSR) for many users under the discrete coefficient restriction. An approach based on irregular optimization was created to address the nonconvex objective function. When TARCs are fixed through element-wise optimization, the beamforming can be derived in closed form through the bisection approach. Similarly, Mohamed et al. [68] presented an algorithm for simultaneously optimizing the covariance matrix at the BS, TARCs matrix, and power level that is reflected and refracted by the IOS. The unique aspect of this study is that it investigates the interdependence of an IOS's transmission and reflection capabilities. Simulation results were used to show the convergence of the suggested strategy and the advantages of employing surfaces with continuous reflection and refraction capabilities.
Different from the works in [66], [67], and [68], Liu et al. [69] investigated the issue of improvement for downlink UAV-assisted IOS (UAV-IOS). Despite RIS, IOS can transmit and reflect signals instantaneously, enhancing the rate in all dimensions. Combining the UAV trajectory with IOS's phase shift helped identify the rate enhancement problem. Although its nonconvexity makes it difficult, they devised a method for generating a superior suboptimal solution. Comparing UAV-IOS communications to other standard methods, computational results indicate that UAV-IOS communications can achieve higher rates. This demonstrates IOS's capacity to provide comprehensive telecom coverage in forthcoming 6G networks.
Zhang et al. [70] implemented IOS-assisted networking to expand wireless communication system coverage and provide MUs with reflective and refractive service. A multiantenna SBS and an IOS collaborate to execute beamforming via various reflective/refractive paths to boost the received power of multiple MUs on either side of the IOS. The authors defined an optimization problem for SBS beamforming and IOS phase shift and presented an iterative solution for sum rate optimization. According to theoretical analysis, IOS expands SBS coverage and rate compared to IRS. In another work, Zhang et al. [71] proposed IOS to achieve full-dimensional communications and determine the optimal beamforming technique for IOS because obtaining the ideal CSI presented challenges. Then beamforming was designed at the BS and IOS, employing beam training with codebooks. It was proposed that the cross-beam training system conduct user beam training concurrently, thus decreasing training costs. The simulation findings show that their proposed strategy achieves a greater data rate than most advanced beam training methods and performs nearly as well as the ideal CSI scenario. Similarly, Perera et al. [74] studied the FD communication system assisted by STAR-RIS to maximize the system's WSR by enhancing STAR-RIS elements for ES and MS protocols. The authors utilize SCA and propose a suboptimal solution. The maximum average WSR and related factors were quantified at the STAR-RIS. Then, the performance of the proposed system design was evaluated through simulations compared to halfduplex (HD) equivalents and conventional RISs. STAR-RIS has been found to enhance the performance of FD systems.
IRS's capacity to dynamically regulate the phase shift of replicated EM waves to create an optimal broadcast environment led to its widespread adoption. The revolutionary concept of IOS allows for modifying signal reflection and transmission, whereas IRS focuses solely on signal reflection. Consequently, IOS represents a novel paradigm for establishing ubiquitous wireless technology. Cai et al. [72] presented IOS with an MU-MISO scheme that makes use of IOS's reflecting and transmissive properties to enhance MU-MISO broadcast. Using the Riemannian manifold, weighted minimum mean square error (WMMSE), second-order cone programming, and BCD algorithms, both the power minimization and sum-rate maximization problems were solved. Simulation results validated the utility of the associated joint beamforming design technique and the improvements of the anticipated IOS-based wireless networks. In addition, several IOS-based system issues, such as hardware flaws, quick channel estimation, realistic transmission protocols, the creation of more complex procedures, learning-based methods, etc., should be examined in future research. Unlike the MU-MISO-based work in [72], Niu et al. [73] investigated MIMO using STAR-RIS support. Utilizing the ES scheme predominantly, the authors seek to optimize the WSR of the system under consideration. A suboptimal BCD method was developed to construct the precoding matrix and the TARCs to maximize the weighted sum-based primarily on ES to solve this optimization problem. Precoding matrice was solved using the Lagrange dual method, and TARCs were produced using the constrained concave-convex process. The results demonstrate that STAR-RIS is superior to RIS. Furthermore, the TS strategy outperforms the MS and ES strategies in single-hop interactions. In contrast, the ES strategy in broadcast communication is superior to the TS and MS strategies.
2) STAR-RIS-NOMA-Based Sum-Rate Maximization: Zuo et al. [75] proposed a STAR-RIS-NOMA system in which active beamforming, power allocation coefficients, transmission beamforming, and reflection beamforming enhance the possible sum rate by improving each other's decoding order. The authors framed a nonconvex problem with interconnected variables. To address this issue, a suboptimal two-layer iterative technique was presented. In the inner layer iteration, for instance, the power allocation coefficients, active beamforming, transmission beamforming, and reflection beamforming were alternately tuned for a specific decoding order. The solutions were disclosed during the outer layer iteration, while the inner layer iteration was used to modify the decoding order of NOMA users for each cluster. Simulation results indicated that the proposed system outperforms traditional RIS-assisted systems. In another work, Liu et al. [76] investigated the performance of the NOMA networks that STAR-RIS supported for ultrareliable low-latency communications. Effective capacity (EC) was employed to assess the delay requirements of NOMA customers. Specific analytical expressions of the EC were derived for the network with two NOMA users on different STAR-RIS edges. The asymptotic analysis of the ECs at high SNR was also provided using the high SNR slope and power balance. As a result, STAR-RIS-NOMA networks had superior EC compared to STAR-RIS-OMA and conventional RIS networks. Furthermore, tight delay constraints result in inferior ECs. Additionally, as the number of elements in STAR-RIS increases, so does the geographic variety and ECs. Therefore, the given analytical technique will help the STAR-RIS-NOMA networks theoretically.
Zhao et al. [77] assessed the ERs of a STAR-RIS-NOMA system in which STAR-RIS provides LoS links to these celledge users. Due to obstructions, direct links between the BS and these cell-edge users were not LoS. The merged channel power gain distribution was fitted to a gamma distribution to obtain closed-form expressions of ERs and high SNR slopes for cell edge users. According to numerical results, the ERs of the proposed systems are greater than those of conventional RIS-NOMA systems, and the slopes of the high SNR are constant.
Zhang et al. [78] examined the system stability of the STAR-RIS NOMA system with queue awareness. The continuing stability-oriented queue-weighted issue was rewritten as a queue-WSR (QWSR) maximization problem in respective time slots, leveraging the Lyapunov drift theory to address the case of the indefinite periods required for stability. The BS maintained the data queue length, and transmission to each user was awaited. Active beamforming coefficients (ABCs) at the BS, NOMA decoding order, and passive TARCs at STAR-RIS were jointly optimized to maximize the QWSR. ES, MS, and TS three STAR-RIS operating protocols were considered. For ES, the BCD and SCA were used to iteratively and alternatively optimize problems to manage the highly coupled and nonconvex problems. The defined issue was divided into two subissues for TS. B both of them can be handled in the same way that ES is. To solve the binary amplitude limited problem for MS, the proposed iterative method was expanded into a penalty-based two-loop procedure. The simulation results indicated that the queue would remain stable under the revised QWSR maximization problem. In addition, the proposed STAR-RIS NOMA communication system performance surpasses conventional systems. Simulations demonstrated that, regarding QWSR and average queue length, the TS protocol was superior to the other two protocols.
Wu et al. [79] stated that STAR-RISs have emerged as a potential technique for modifying the radio propagation environment throughout the entire universe. Prior research on STAR-RISs has concentrated primarily on the ES operation protocol, which has high hardware complexity. In addition, the passive beamforming design of STAR-RIS always assumes the availability of complete and rapid CSI, which is practically difficult due to many STAR-RIS componentsobserving the STAR-RIS MS design and STAR-RIS-NOMA communication system issues. Effective two-timescale (TTS) broadcast methods have been developed to increase the average sum rate for each channel configuration. Specifically for LoS-dominant channels, short-term power provision at the BS was designed based on all users' evaluated effective fading channels. Simultaneously, the long-term transmission and reflection factors for STAR-RIS were optimized using only the statistical CSI. For dense scattering environments, partition-the-estimate (PTE) is recommended. BS determines the long-term STAR-RIS surface-partitioning strategy-based solely on path-loss information, assigning each subsurface to a single user. Accordingly, BS designs its power allocation and STAR-RIS phase shift. In addition, efficient algorithms for solving short-term and long-term optimization issues were proposed. Similarly, both proposed transmission protocols reduce the channel estimation overhead significantly.
Zuo et al. [80] proposed a STAR-RIS-NOMA that optimized the sum rate of an NOMA system. Furthermore, the authors proposed a two-layer iterative algorithm to tackle the nonconvex problem with coupled variables that alternately optimized the power allocation coefficients, active beamforming, transmission and reflection beamforming, and decoding order. Simulation results showed that the proposed STAR-RIS-NOMA system outperforms conventional RIS-NOMA and RIS-OMA systems.
Gao et al. [81] investigation of the STAR-RIS-assisted NOMA communication systems. The system involves the deployment of a STAR-RIS within a predefined region to establish communication links for users. The deployment location of the STAR-RIS and the beamforming at the BS and the STAR-RIS are optimized to maximize users' WSR. Furthermore, an AO algorithm is proposed to solve the nonconvex problem. The numerical results showed that optimizing the deployment location of the STAR-RIS can significantly enhance the system's performance, and both beamformerbased NOMA and cluster-based NOMA prefer asymmetric STAR-RIS deployment.
Abrar et al. [82] proposed MU-MISO mmWave hybrid (H-NOMA) in a wireless network to mitigate the half-space coverage limitation of conventional RIS operating on mmWave. The authors proposed an optimization framework based on AO that iteratively solves active and passive beamforming subproblems. The authors also proposed channel correlations and strength-based techniques for a case of two-user optimal clustering and decoding order assignment. The simulation results showed that the proposed framework leveraging H-NOMA outperforms conventional OMA and NOMA to maximize the achievable sum rate. Furthermore, using STAR-RIS allows the number of elements to be significantly reduced while ensuring a similar QoS.

D. Improving Energy Efficiency Leveraging STAR-RIS
A smart controller in STAR-RIS can segregate the incident signal into the transmission and reflection sectors to give 360 • coverage. The top features of STAR-RIS, besides optical transparency and attractiveness and suitability for usage with windows, are that it can transmit and reflect incident signals simultaneously, enables the full space SRE, and sets dynamic power ratios for STAR-RIS elements. One part of the incident signal is reflected in the reflection space, while the other is transmitted in the opposite direction (i.e., the transmission space). Controlling a STAR-RIS element's electric and magnetic currents allows the TARCs to reconfigure signals. By designing the STAR-RIS TARCs, system performance can be increased. This section examines the latest methods for enabling energy-efficient solutions using STAR-RIS, supported by the summary in Table VII to assist readers in comprehending the innovative approaches.
Mu et al. [83] considered OMA network and investigated an innovative STAR-RIS concept where a power consumption minimization issue was developed to optimize active beamforming at the BS and passive broadcast and reflection beamforming at the STAR-RIS under user-imposed communication frequency constraints. The resulting strongly connected nonconvex optimization problem was addressed using an iterative method involving the penalized method and sequential convex approximation. Calculated results revealed that, compared to RIS, STAR-RIS could drastically reduce BS power consumption. Moreover, it was revealed that element-wise amplitude control performed better than group-wise amplitude control for STAR-RIS. Similarly, Mu et al. [84] conducted research on the unique concept of STAR-RIS. Three realistic working protocols for STAR-RIS, MS, TS, and ES were proposed based on the STAR signal model development.
A STAR-RIS-assisted downlink communication system that transmits information to users on both sides of the STAR-RIS was also considered. A challenge was presented for the best allocation of active beamforming at the BS and passive transmission and reflection beamforming at the STAR-RIS, subject to user-imposed communication rate limits for each proposed operational protocol. Using an iterative strategy based on SCA and the penalty method, ES's resulting highly linked nonconvex optimization problem was resolved. The optimization problem for MS involving mixed-integer nonconvexity was solved by extending the proposed penalty-based iterative technique. Using convex optimization approaches and cutting-edge algorithms, the optimization issue for TS was divided into two manageable subproblems. In addition, numerical findings reveal that the proposed approach substantially reduces the required power and that the TS and ES operating protocols are frequently selected for unicast and multicast transmission, respectively.
Zhong et al. [85] investigated the STAR-RIS MISO scenario and considered a coupled phase-shift model. By defining a joint passive and active beamforming optimization problem, the power consumption for long-term broadcasting was minimized under the coupled phase-shift restriction and the least data rate constraint. The formulated issue was handled by employing a hybrid continuous and discrete phase-shift control policy, despite the coupled nature of the phase-shift model. This realization led to the creation of two hybrid RL algorithms: 1) the joint deep deterministic policy gradient (DDPG) and 2) deep-Q network-based algorithm and the hybrid DDPG algorithm. In accordance with the hybrid action mapping, the hybrid DDPG technique manages the associated highdimensional continuous and discrete actions. The combined DDPG-DQN algorithm provides a joint hybrid control by generating two Markov decision processes (MDPs) dependent on the outer and inner environment. The simulation results demonstrate that the STAR-RIS uses less energy than other RISs. Moreover, the joint DDPG-DQN algorithm achieves enhanced performance despite having a higher computational complexity than any of the presented approaches, surpassing the original DDPG algorithm.
Wang et al. [86] claimed that RIS was a technique with the potential to enhance the functionality of future wireless networks. Then, an innovative STAR-RIS has been proposed to facilitate communication. Using a sophisticated controller to alter the EM properties of the STAR-RIS components, the incident signal can be split into reflected and transmitted signals, allowing for 360 • coverage. In their work, the authors demonstrated the efficacy of an FD communication system with STAR-RIS support, in which an FD-BS simultaneously communicates with a UL and DL user across the same timefrequency minimum domain. The objective was maintaining a minimum data rate while decreasing the total transmit power. The original issue was subdivided into power optimization and STAR-RIS passive beamforming issues using the AO paradigm. Each iteration obtained the closed-form expression for the ideal power plan using the SDP and SCA method and solved the passive beamforming optimization subproblem. The simulation findings demonstrated that STAR-RIS was more successful than conventional RIS. STAR-RIS-assisted FD schemes would also perform better than HD options in cases with higher data rates and lower self-interference (SI).
In another work, Zuo et al. [87] recommended using the uplink NOMA communication technology with STAR-RIS, which is different from the downlink OMA scenario in [83], [84], [85], and [86]. By optimizing user transmit power concurrently, the BS receives beamforming vectors, STAR-RIS, STAR beamforming vectors, and time slots, hence posing the problem of minimizing total power usage. Here, transmission and reflection beamforming of STAR beams via STAR-RIS are introduced. To reduce the total power consumption problem, a penalty-based AO strategy was proposed. The effectiveness of the proposed plan has been proven by simulation results, which also demonstrated that alternative system configurations affect total energy consumption and the system has already been tested by simulation results, which also revealed the many system configurations that influence total energy consumption.
In contrast to uplink NOMA-based access work in [87], Guo et al. [88] examined the NOMA-assisted STAR-RIS downlink network's EE maximization issue. Due to the EE's fractional nature, it was challenging to maximize it using conventional convex optimization methods. To maximize the EE, transmission beamforming vectors at the BS and the coefficient matrices at the STAR-RIS were optimized using a DDPGbased technique. The simulation findings suggested that the recommended approach can successfully maximize system EE while considering channels that vary over time.
To improve EE in an MIMO-enabled NOMA system, Fang et al. [89] proposed an algorithm to optimize transmit beamforming and the phases of the passive elements on the STAR-RIS to maximize system EE. The algorithm decomposes the EE problem into beamforming and phase shift optimization problems and uses signal alignment and zeroforcing precoding methods to address the nonconvex beamforming optimization problem. The Dinkelbach approach and dual decomposition are then used to optimize the beamforming vectors. Finally, the nonconvex phase shift optimization problem is addressed using an SCA-based method. Simulation results demonstrated that the proposed algorithm with NOMA technology outperforms the OMA scheme and the random phase shift scheme in terms of EE performance. In another work, Wang et al. [90] focused on maximizing EE in an MISObased wireless network leveraging STAR-RIS and NOMA. The problem is nonconvex due to the coupling of beamforming vectors and phase shifts but is solved using fractional programming to transform the problem into a convex SDR problem with a rank-one constraint. Furthermore, a novel sequential rank-one constraint relaxation (SROCR) is proposed to convert the rank-one constraint into a convex one. The proposed method achieves superior performance in EE.
Guan et al. [91] proposed a new approach to enhancing the performance of wireless networks using a STAR-RISaided FD communication system. The proposed approach aimed to maximize EE by jointly optimizing the BS and the uplink user's transmit power and the passive beamforming at the STAR-RIS. The authors decoupled the problem into two subproblems and optimized them iteratively using different methods. Simulation results show that the proposed approach outperforms other baseline schemes, offering greater deployment flexibility and significantly improving wireless network performance.

E. Mitigating Interference Leveraging STAR-RIS
In future cellular networks, it is projected that the coverage of densely deployed small cells would overlap, hence increasing the likelihood of multicell interference. In this context, one of the most important uses of the STAR-RIS technology is the decrease of interference in wireless networks caused by many cell scenarios. For instance, the STAR-RIS might be designed for signal refraction and reflection toward the desired edge users in a given cell, while concurrently suppressing the signals aimed at the undesired users as shown in Fig. 7.
Coordination among small cells is an intricate and ongoing research problem since multiple small cells may share the same STAR-RIS. While in practice, a BS in small cells only gets the CSI of its own linked users, the analog beamforming at the STAR-RIS has an effect on all users in the near vicinity. As a result, methods are required to coordinate the small-cell BSs efficiently. In this setting, ML can be utilized to coordinate multiple small cells in challenging wireless environments. This section examines cutting-edge methods for mitigating interference using STAR-RIS for various scenarios. It is complemented by a summary in Table VIII to help readers understand the fundamental ideas and develop a deeper understanding of the key technologies.
In the article, Hou et al. [92] presented a novel simultaneous-signal-enhancement-and-cancelation-based (SSE CB) design based on signal-cancelation-based (SCB) designs and signal enhancement-based (SEB) designs. This permits the intercell differences and preferred signals to be removed and improved in parallel. In addition, the simulation analysis revealed that numerous RIS elements may be used to perfectly eliminate intercell interference while optimizing the needed signals, and how the recommended SSECB design is superior to the conventional SCB and SEB schemes. In future investigations, the passive beamforming, active beamforming, and detection vectors should be designed cooperatively to increase RA and TA at the BS and for users. Combining stochastic geometry approaches with the evaluation of the consequences of unpredictability in user location is a potential future development. In another work, Zhang et al. [93] investigated an IOS that enables simultaneous signal transmission and reflection, thereby benefiting users on both sides. To eliminate intercell interference, they developed an IOS-enhanced indoor communication structure consisting of an IOS installed between two autonomous APs. To increase the total rate without sending CSI between APs, they devised a remote hybrid beamforming system that combines digital beamforming at the AP level with IOS-based analog beamforming. In terms of sum rate performance, simulation findings reveal that the proposed system surpasses existing schemes and closely matches the performance of an optimal centralized scheme.
FD transmission provides more bandwidth than HD transmission, according to Fang et al. [94]; nonetheless, SI is the largest issue. This study, however, differs from those in [92] and [93], which focused on intercell interference. Initially, IOS-FD-MISO was recommended to address the intensity problem, whereas ES-IOS and MS-IOS were presented to increase the data rate and lower SI power, respectively. These obstacles were difficult to address directly. Consequently, they designed a revolutionary optimization method. Specifically, the amplitudes and phase shifts of the ES-IOS and MS-IOS were optimized using QCQP. Due to the complexity of binary IOS variables, they apply SDR and the Gaussian randomization technique to solve the problem. The findings demonstrate that both IOSs effectively reduce SI compared to the lack of an IOS, proving the usefulness of the suggested techniques.

F. Resource Allocation
Unlike conventional reflecting-only RISs, STAR-RIS utilized in wireless systems can accomplish a full-space SRE. It is crucial to develop a resource allocation policy for STAR-RIS with the primary goal of optimizing system performance to reach its full potential. However, the optimization factors are highly interdependent on one another (such as transmit power, channel assignment, TARCs, and factor of time allocation/decoding order). Since the user-grouping problem appears in multicarrier transmission, allocating resources is not a simple task. In view of this, it is crucial to create effective algorithms for STAR-RIS-based wireless communication systems. This section adds to the STAR-RIS research by looking at how resources are allocated in OMA and NOMA communication networks. Moreover, a summary in Table IX of resource allocation-based schemes is also enclosed for comparative analysis. Two significant issues for RIS-aided vehicular communications are the cascaded links encountering double fading and the CSI acquisition imposed by high mobility. By simultaneously sending and amplifying the incident signals instead of only reflecting them, Chen et al. [95] presented RIOS as a novel type of RIS to address these issues. Active RIOS is installed on the automobile window to improve transmission for both passengers and nearby drivers. Their objective was to simultaneously optimize the broadcast precoding matrix at the BS and RIOS coefficient matrices to minimize the BS's transmit power despite having a limited comprehension of the large-scale CSI. With a limited understanding of the large-scale CSI, their goal was to concurrently optimize the broadcast precoding matrix at the BS and RIOS coefficient matrices to reduce the BS's transmit power. To decrease the frequency of network various trends, an effective transmission protocol was proposed to use high active RIOS beamforming gain with minimal network training overhead by skillfully adjusting the timing of CSI collection. Constrained stochastic sequential convex approximation (CSSCA) and AO techniques were used to approach the examined resource allocation problem. Simulation results Qin et al. [96] introduced a novel concept called STAR-RIS for wireless-powered mobile edge computing (MEC) systems. The aim is to improve energy transfer and task offloading efficiency by extending half-space coverage to full-space coverage and manipulating signal propagation. In this article, the authors proposed to maximize the total computation rate of all users by jointly optimizing energy transfer time, transmit power, CPU frequencies, and STAR-RIS configuration design. Three operating protocols are studied: 1) ES; 2) MS; and 3) TS. An iterative algorithm was proposed based on the penalty, SCA, and linear search methods to solve the nonconvex problem. Simulation results showed that STAR-RIS outperforms traditional reflecting/transmitting-only RIS and that the TS protocol achieves the most effective computation rate among the three operating protocols of STAR-RIS.
In contrast to the studies in [95] and [96], which only investigated vehicular and MEC OMA downlink network, respectively, Yang et al. [97], studied an NOMA system assisted by STAR-RIS. They presented discrete amplitude allocation and joint power techniques that can lessen the burden associated with channel approximation and the difficulty of the hardware. To certify the QoS requirements of the reflected user, the Beaulieu series over Nagakami-m fading was created to appraise the performance of the suggested method. Then, closed-from terminologies were discovered for the various order of the reflected user and OP. Additionally, a lower bound of the communicated user's outage possibility was examined due to the integrated network statistic of the communicated user. Numerical findings showed that the suggested technique significantly enhances the appearance of the affected user and produces a higher output.
Unlike the work in [97], which focused on downlink scenario, Ni et al. [98] considered uplink and combined over-the-air federated learning (AirFL) with NOMA through concurrent broadcasts in a scalable and unified model. The signal processing order was modified in a particular way to use STAR-RIS for effective interference reduction and omnidirectional service improvement. To observe the effects of nonideal wireless communication on AirFL, a closed-form equation was derived for the optimality gap over certain communication circles to examine the omnidirectional effects of nonideal wireless communication on AirFL. These findings showed that the allocating resource mechanism and channel interference significantly influence learning performance. To reduce the obtained optimality gap, an MINLP problem was framed by concurrently designing the communication power on the user's side and the configuration mode provided by STAR-RIS. Based on the results, using the STAR-RIS helps boost the training process in the sense of omnidirectional test accuracy and learning loss.
Wu et al. [99] studied the issue of resource distribution in multicarrier communication networks aided by STAR-RIS. To maximize the user's sum rate, a shared optimization problem involving power allocation, channel assignment, reflection, and transmission beamforming at the STAR-RIS for OMA was designed. They established a channel task scheme based on matching theory and iteratively improved the beamforming vectors and resource allocation approach using the AO-based method. The authors then investigated the optimization of the sum rate for NOMA with adjustable decoding orders. Initially, a location-based matching algorithm that groups a transmitted and reflected user on a subchannel was presented to effectively address the issue. Semidefinite programming, convex upper bound approximation, and geometry programming were proposed as a three-stage process for this reflection and transmission subchannel task plan. Numerical results show that same-side user pairing for channel assignment in a generic design is superior to OMA. In contrast, the projected reflection and transmission system performs the exhaustive search-based NOMA algorithm. In addition, the STAR-RIS-NOMA network outperforms typical RIS and OMA networks. Analytical modeling of the STAR-RIS coefficients' interactions was crucial. Considering the practical models, resource allocation will become significantly more difficult as numerous new limitations are added. So, the questions about how to deal with the constraints by simplifying the model and using more effective methods need to be looked at in more depth.

G. Performance Analysis
The use of STAR-RIS is a potential way to generate a flexible wireless broadcast environment. Incident signals can be transmitted and reflected back to users at different surface edges in the most recent STAR-RIS design. Thus, each STAR-RIS component required the setup of two coefficients. The first is set to change the phase shifts and amplitudes of the transmit signal, and the second is set for the reflected signal. In difficult locations with insufficient direct AP-user connectivity, in order to boost signal STAR-RIS can be used.
In this section, we will examine the studies conducted to assess STAR-RIS performance in relation to OMA and NOMA. Outage probabilities are calculated for various wireless network configurations, including perfect and imperfect SICs, varying channel conditions, and phase shift configuration strategies, all of which are relevant to evaluating the efficacy of STAR-RISs under different access technologies. Additionally, the STAR-RIS channel estimation and bit error rate (BER) performance are evaluated considering several users spread out on surface sides using an NOMA-based approach. Table X summarizing the performance evaluation-based works is given for a better understanding of each scheme's scenarios, channel characteristics, techniques, and evaluation strategies.

1) Performance Analysis for OMA-Based STAR-RIS:
In contrast to RIS, the coverage of STAR-RIS is increased to 360 • . Xu et al. [100] proposed a generic hardware model for STAR-RIS. The diversity gain of STAR-RIS was compared to that of conventional RISs by simulating channels and then projecting the results for both far-field and nearfield conditions. Numerical simulations corroborate analytical results, which show that full diversity order can be achieved across both sides of the STAR-RIS. Different from [100], Wu et al. [101] looked into a STAR-RIS-aided uplink channel estimation design for a two-user communication system. They first evaluated the STAR-RIS TS protocol and then devised a workable method for estimating the channels of two users using a transmission/reflection training pattern. Then, the authors considered the realistic coupled phase-shift model and devised a novel approach to simultaneously estimate the channels of both users when evaluating the ES protocol for STAR-RIS. The authors demonstrated an effective way to generate a high-quality solution by simultaneously creating the pilot sequences, training patterns, and power-splitting ratio. However, reducing the channel estimation error by using the ES protocol was a significant obstacle to overcome. Estimates of the uplink channel show that TS is more cost-effective than ES. Future research may focus on downlink channel estimations and resilient beamforming in imperfect CSI.

2) Performance Analysis for NOMA-Based STAR-RIS:
Wang et al. [102] evaluated the OP in a STAR-RIS-assisted NOMA-wide network by analyzing the spatial correlations between channels. The authors use a moment-matching technique to initially approximate the distribution of the composite channel gain as a gamma random variable to study the effects of channel correlations on system performance. The authors then presented closed-form expressions of OP for two NOMA users. The theoretical method is validated by numerical results, which also exhibit performance loss due to channel correlations. In another work, Yue et al. [103] examined the OP and ER of networks over Rician channels and provided an in-depth analysis of STAR-RIS-NOMA. The corresponding probability of outages for users "n" and "m." The diversity rankings of the users' "n" and "m" are determined using asymptotic results. It has been shown that STAR-RIS-NOMA is more likely to have outages than STAR-RIS-OMA. In addition, the theoretical formulations of the ER for users "n" and "m" with pSIC were meticulously reported along with their corresponding high SNR slopes. According to numerical results, the ER of user "n" with pSIC outperforms orthogonal users at high SNRs. The NOMA-based STAR-RIS system's throughput was also measured in delay-limited and delay-tolerant configurations. STAR-RIS-NOMA has the potential to achieve stricter QoS demands in practical use cases, where "n" and "m" users may be high and low-bitrate video streaming clients, respectively.
Perfect CSI configuration may result in overstated performance promises for NOMA-based STAR-RIS networks. Future studies may examine the effects of inaccurate CSI and look for efficient channel estimate techniques.
STAR-IOS regulates NOMA-based, SIC by dividing energy or altering active elements. Zhang et al. [104] examined STAR-IOS-assisted NOMA's benefits with randomly distributed users and proposed three tractable channel models, including Central limit, curve fitting, and M-fold convolution. Curve fitting analyses multicell networks, whereas central limit modeling fits massive STAR-IOS scenarios. Both categories cannot organize variation. M-fold convolution organizes diversity. STAR-IOS uses ES, TS, and MS. NOMA users get analytical closed-form OP from ES protocol-based central limit and curve fitting models. M-fold convolution and three protocols calculate NOMA user diversity gains. NOMA users diversified like STAR-IOS. The central limit model gives an upper bound, and the curve fitting model provides a lower bound in regions of high SNR ratio; the TS protocol performs best but demands more time blocks than other protocols; and the ES protocol outperforms the MS protocol due to its larger diversity gains. Similarly, the BER performance of STAR-RIS in NOMA networks was studied by Aldababsa et al. [105]. A STAR-RIS-adopting MS protocol serves multiple NOMA users in the analyzed behavior network. The BER equations for both the perfect and imperfect SIC instances are derived. An asymptotic analysis is also conducted to further investigate the BER's behavior at high SNR. Monte Carlo simulations support the authors' theoretical investigation. Compared to traditional NOMA, STAR-RIS-NOMA delivers superior BER performance, implying it could be an NOMA 2.0 option.
In contrast to [102], [103], [104], and [105], which focused on downlink NOMA-based systems, Sheng [106] examined STAR-RIS for uplink that enables several users to communicate in a power-domain NOMA environment while sharing the same time-frequency resources. Though users have nearly the same spread power and distance from the AP as they approach the cell boundary, system performance degrades at this point. To actively address this issue, only elements that can enhance the channel for the targeted users were selected. Therefore, the entire cascade channel is adjusted to provide an optimum environment for NOMA transmission. Simulation findings reveal that the improved performance of SIC at the receiver can reduce BER.
3) Performance Analysis for Both OMA and NOMA-Based STAR-RIS: Xu et al. [107] investigated STAR-RIS and concentrated on transmission and reflection phase shifts that were linked. By presenting the design for diversity preservation, the authors demonstrate how to acquire complete diversity on both sides. For OMA and NOMA, a STAR-RIS-assisted twouser downlink communication system is explored. Upper and lower performance limitations are compared with OP, diversity order, and power scaling rules. It has been established through simulations that the proposed diversity-preserving phase-shift strategy for the STAR-RIS provides the same diversity order as the independent phase shift of STAR-RIS and obtains a comparable power scaling law with only a 4-dB power drop.
Xu et al. [108] examined OMA and NOMA utilizing a two-user downlink communication system aided by STAR-RIS in their article. To evaluate the impact of the connected broadcast and reflection phase-shift model on communication performance, the diversity-preserving phase shift configuration (DP-PSC), the primary-secondary phase shift configuration (PS-PSC), and the T/R-group phase-shift configuration (TR-PSC) were designed. According to the findings, the proposed DP-PSC technique simultaneously satisfies all guidelines for users on both sides of STAR-RIS. In addition, scaling power rules were developed for the random phase-shift configuration and the three proposed techniques. Using NOMA rather than OMA on each side of the STAR-RIS improved performance, according to numerical simulations. In addition, it was demonstrated that the proposed DP-PSC technique achieves the same variety order as STAR-RIS under the independent phase shift model and a similar power scaling law with just a 4 dB decrease in received power.

IV. FUTURE RESEARCH DIRECTIONS AND CHALLENGES
QoS has improved with each generation of wireless networks. Concurrently with the deployment of 5G networks, 6G network research is being conducted. The RISs are among the competitive 6G components [19], [24], [109]. They improve mmWave communications [110], energy-efficient communication [6], and propagation for edge users. RISs can do more than just signal to boost and can perform functions, such as estimating channels [111], users' localization [112], and integrated sensing and reflecting [113].
In this section, we examine where the field of STAR-RIS-assisted wireless networks is headed and what future challenges it may face. We briefly discuss STAR-RIS challenges and potential future directions of research. Additional intriguing applications of STAR-RISs in 6G networks include STAR-RIS-assisted simultaneous wireless information and power transfer (SWIPT), STAR-RIS-assisted VLC, and STAR-RIS-enhanced robotic communications. These applications hold promise for future research.

A. Open Issues and Future Research Direction
Moreover, the implementation of the STAR-RIS presents a number of novel challenges and complications, including the following.
1) To expand coverage, the analog beamforming and digital beamforming at the RIS and BS, respectively, must be designed jointly. Practical constraints, such as discrete phase changes and necessitate efficient algorithms. How to achieve a balance between performance on the two sides of STAR-RIS is a daunting challenge to investigate. 2) At STAR-RIS, the effectiveness of analog beamforming is heavily reliant on the accuracy of the CSI [4]. Since these channels of STAR-RIS' sides are interconnected, the CSI of users on both sides of the surface must be jointly estimated.
3) Evaluating the efficacy of multiuser NOMA networks aided by STAR-RIS with defective SIC and CSI is an additional challenging future research topic. 4) A STAR-RIS performance depends on its proximity to the transmitter and users. The optimal deployment of a STAR-RIS by balancing its reflecting and transmitter capabilities is an unanswered research question. 5) While STAR-RISs have certain advantages, developing their corresponding TARCs can be difficult. When it comes to the STAR-RIS, first, transmission-reflection beamforming is a far more advanced option than reflection-only beamforming. Because electric and magnetic impedance depends on the EM properties of the STAR components, STAR-RISs cannot independently affect TARCs. Coupled transmit and reflect coefficients necessitate a hybrid continuous and discrete control scheme for phase-shift design. Given the aforementioned obstacles and the fact that current convex optimization and ML techniques only permit continuous or discrete control, it is challenging to solve the transmission and reflection beamforming problem for STAR-RISs. In such cases, attention should be given to hybrid algorithms for small action dimensions. 6) The performance of wireless communication systems depends on efficient resource use. Phase shift and beamforming of passive elements are optimized in STAR-RIS-enhanced communication to increase coverage and boost PLS-based security over traditional communication systems. RIS and resource allocation-for example, subcarrier, power distribution, and trajectory design in the case of UAV integration-are frequently related, making design optimization difficult and resulting in suboptimal designs. Differences between optimal and suboptimal performance, however, are not apparent. Therefore, to optimize RIS-enhanced communication in a wide range of applications, appropriate strategies must balance computational complexity and system performance. 7) The resource allocation for wireless communication systems involving STAR-RIS for OMA and NOMA system resource allocation, including power allocation, channel assignment, reflection, and transmission beamforming at the STAR-RIS for OMA. It is difficult to optimize large-scale STAR-RIS-enhanced wireless communications, particularly when UAVs are involved and placed in an environment that is partially unknown. Due to nonlinear models, it is particularly difficult to create an optimal UAV trajectory, RIS reflecting elements, and network resource optimization. Thus, it is complicated to design approaches with low complexity and efficient system performance. Approaches based on AI and ML are formidable tools for developing and optimizing such networks. These techniques are rapidly evolving and offer powerful and promising tools for planning and optimizing complex situations. Moreover, the complex system can be analyzed using hybrid models, data-driven approaches, and hybrid offline and online methods to improve system performance.
8) There are three main obstacles to overcome while developing and deploying active STAR-RIS for high-mobility applications like vehicular communication networks. To begin with, the active STAR-RIS will always increase the noises, leading to even more noises at the receiver, which could reduce the system's efficiency. As a result, optimally setting the coefficients of the active STAR-RIS elements is essential for striking a balance between the competing demands of maximizing the received signal strength while simultaneously reducing the noise impact. Second, beamforming precision is highly dependent on CSI acquisition precision. As the predicted CSI quickly becomes out of date in cases with highly dynamic channels, there will always be noticeable CSI discrepancies due to Doppler shifts. This means that in a dynamic environment, the beamformed transmission calls for strong beamforming techniques that are specially designed to the unique requirements of the transmission. Third, it is challenging to keep track of the instantaneous CSI in practice due to continuously changing mobile channels. Moreover, significant signaling overheads are associated with the frequent feedback of fading information. These challenges are made worse by the large number of cascaded channel coefficients introduced by active STAR-RIS. Therefore, there is a need for further research in highly mobile scenarios considering the aforementioned open challenges in active STAR-RIS. 9) The STAR RIS technology has the potential to revolutionize wireless communication networks. However, hardware challenges are associated with its implementation, such as designing and implementing highly efficient and reconfigurable surface materials, developing low-cost and scalable fabrication techniques, and integrating with existing wireless communication networks. Additionally, the control and optimization of signals require the development of sophisticated algorithms. Overall, multidisciplinary research efforts are required to overcome these challenges. However, the potential benefits of this technology make it an exciting area of research for the future of wireless communication networks. 10) For sensing and localization, RF signals are increasingly used because they are inexpensive and maintain confidentiality. RF sensing and localization are reliant on exploiting the environment-dependent features of wireless signals. To attain a high level of precision, it is imperative that the signals captured at two distinct sites are as diverse as possible. STAR-RIS is deemed effective for RF sensing and localization in this scenario because it can alter propagation channels to make them more distinct from one another. Additionally, the STAR-RIS's ability to communicate in all dimensions can reduce the number of uncovered areas effectively. 11) Nonetheless, the implementation of sensing and localization through STAR-RIS must overcome a number of challenges. Optimizing the STAR-RIS analog beamforming to reduce errors while sensing and localization are one of them. Compressed sensing techniques can be used in a variety of situations when signals are limited in specific areas. The signals can also be sorted according to the presence of things and users' locations using ML techniques. 12) Future IoT networks may feature SWIPT. EH makes it possible for IoT devices to draw power from ambient EM sources or from purpose-built EH sources, such as those that emit EM waves simultaneously [114]. Low EH efficiency makes it hard to use SWIPT systems in the real world [115]. Implementing a STAR-RIS for the SWIPT system is one approach that shows promise for addressing this issue. To be more precise, the STAR-RIS is able to concentrate EM waves, resulting in an increase in the efficiency of energy harvesting.

V. CONCLUSION
We offered an innovative approach to exploring the potential of the STAR-RIS technology in 6G networks. We made several valuable contributions and provided a comprehensive and up-to-date survey of the state-of-the-art schemes that enable an SRE. Starting with a distinctive viewpoint, we introduced the fundamentals of different RIS types, including passive, active, and STAR-RIS, creating a solid foundation for the remainder of this article. We then examined the STAR-RIS operating protocols, applications, and benefits, building on existing literature and providing a comprehensive technology overview. We continued our approach by organizing STAR-RIS schemes based on use cases, resource allocation, and performance evaluation. We provided a logical structure to the discussion, offering an overview of the various use cases and subcategories, such as coverage, PLS, sum rate, EE, and interference. We investigated various STAR-RIS resource allocation and performance evaluation techniques, providing a comprehensive and innovative understanding of the technology. Finally, we identified several open problems and new research directions in the field, offering opportunities for significant advancements. Our approach sets this article apart from others in the field and makes it a valuable resource for all levels of researchers seeking to explore and advance the potential of the STAR-RIS technology in 6G networks.