Efficient Resource Allocation and User Association in NOMA-Enabled Vehicular-Aided HetNets with High Altitude Platforms

COMPUTER COMMUNIA CTION 1 Ef ficient Resource Allocation and User Association in NOMA-Enabled V ehicular -Aided HetNets with High Altitude Platforms Ali Nauman, Mashael Maashi, Hend K. Alkahtani, Fahd N. Al-W esabi, Nojood O Aljehane, Mohammed Assiri , Sara Saadeldeen Ibrahim, W ali Ullah Khan Abstract —The increasing demand f or massive connectivity and high data rates has made the efficient use of existing spectrum resour ces an increasingly challenging pr oblem. Non- orthogonal multiple access (NOMA) is a potential solution for future heterogeneous networks (HetNets) due to its high capacity and spectrum efficiency . In this study , we analyze an uplink NOMA-enabled vehicular-aided HetNet, where multiple v ehic- ular user equipment (VUEs) share the access link spectrum, and a high-altitude platf orm (HAP) communicates with roadside units (RSUs) through a backhaul communication link. W e propose an impro ved algorithm for user association that selects VUEs for HAPs based on channel coefficient ratios and terrestrial VUEs based on a caching-state backhaul communication link. The joint optimization problems aim to maximize a utility function that considers VUE transmission rates and cross-tier interfer ence while meeting the constraints of backhaul transmission rates and QoS requirements of each VUE. The joint resource allocation optimization problem consists of three sub-problems: bandwidth allocation, user association, and transmission power allocation. W e derive a closed-f orm solution for bandwidth allocation and solve the transmission power allocation sub-pr oblem iteratively using T aylor expansion to transform a non-conv ex term into a con vex one. Our proposed three-stage iterative algorithm f or resour ce allocation integrates all three sub-problems and is shown to be effective through simulation r esults. Specifically , the r esults demonstrate that our solution achiev es perf ormance impro vements over existing approaches. Index T erms —Non-orthogonal multiple access (NOMA) Het- erogeneous networks (HetNets) V ehicular user equipment (VUE) High altitude platform (HAP) r oadside units (RSUs). Ali Nauman is with the Department of Information and Communication Engineering, Y eungnam University , Republic of Korea. Mashael Maashi is with the Department of Software Engineering, College of Computer and Information Sciences, King Saud University , Po box 103786, Riyadh 11543, Saudi Arabia. Hend K. Alkahtani is with the Department of Information Systems, College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman Univ ersity , P .O. Box 84428, Riyadh 11671, Saudi Arabia. Fahd N. Al-W esabi is with the Department of Computer Science, College of Science & Art at Mahayil, King Khalid University , Saudi Arabia. Nojood O Aljehane is with the Department of Computer Science, Faculty of Computers and Information T echnology , University of T abuk, T abuk, Saudi Arabia. Mohammed Assiri is with the Department of Computer Science, College of Sciences and Humanities- Aflaj, Prince Sattam bin Abdulaziz Univ ersity , Aflaj 16273, Saudi Arabia. Sara Saadeldeen Ibrahim is with the Department of Computer and Self Dev elopment, Preparatory Y ear Deanship, Prince Sattam bin Abdulaziz Uni- versity , AlKharj, Saudi Arabia. W ali Ullah Kahn is with the Interdisciplinary Centre for Security , Reliability and Trust (SnT), University of Luxembour g, Luxembourg. Corresponding Author: Fahd N. Al-W esabi, email: falwesabi@kku.edu.sa I . I N T R O D UC T I O N Integrated terrestrial non-terrestrial networks refer to telecommunications systems that effecti vely merge terrestrial and non-terrestrial communication infrastructure, hence offer - ing complete and robust connectivity [1]. Integrating terrestrial networks, which possess advantageous characteristics such as high capacity and low latency , with non-terrestrial networks like satellite constellations and high-altitude platforms (HAPS) presents a versatile solution for addressing global commu- nication requirements [2]. This solution proves particularly beneficial in areas with limited terrestrial co verage, as well as in scenarios in v olving disaster recovery , Internet of Things (IoT) applications, aerospace, aviation, military operations, and emergency response [3]. By combining the strengths of both types of netw orks, this integrated approach ensures redundancy , high-speed data transmission, and dependable connectivity across diverse en vironments. W ith the increas- ing demand for connected devices and autonomous systems, future communication systems will rely heavily on integrating terrestrial and non-terrestrial networks to facilitate the deliv- ery of cutting-edge services [4]. T ogether , they can provide more comprehensi ve and reliable communications, of fering improv ed cov erage, high spectrum efficienc y , cost savings, and better use of resources [5]. Therefore, the conv ergence of terrestrial networks and satellites has garnered significant attention among researchers and professionals worldwide [6]. HAPs have receiv ed significant attention due to their abil- ity to stay aloft at high altitudes for extended periods and provide various services such as communication, surveillance, and remote sensing [7]. HAPs are unmanned aerial vehicles (U A Vs) operating at high altitudes ranging from 12 to 22 kilometers (7.5 to 14 miles), above most weather patterns and commercial air traffic [8]. HAPs have dev eloped for several decades, but recent advances in materials, propulsion, and communication technologies have renewed interest in their potential applications. Additionally , HAPs can be deployed quickly and at a lower cost than satellites, making them an attractiv e alternativ e for certain applications [9]. Heterogeneous networks (also called HetNets), which com- bine terrestrial base stations (BSs) and non-terrestrial, i.e., U A Vs, HAPs, and satellites networks, offer a range of benefits, including high-speed links for UEs pro vided by terrestrial BSs, global co verage and ample backhaul capacity by the non- terrestrial network [10]. In terrestrial non-terrestrial HetNets, COMPUTER COMMUNIA CTION 2 the challenges posed by the limited spectrum resource and the growing number of users who generate more interference must be addressed to ensure efficient transmission [11]. Despite its ability to mitigate interference, orthogonal multiple access technology is subject to certain limitations, including restricted improv ements in the utilization of spectrum and enhanced capacity because only one user is capable of connecting over a resource block (orthogonal) in a unit of time [12]. Moreov er , integrating NOMA technologies [13] can sig- nificantly enhance the effecti ve utilization of communication network spectrum resources [14]. Researchers widely recog- nize NOMA as a promising approach to allocate spectrum resources effecti vely and utilize them rationally in next- generation multiple access technologies [15]. NOMA facili- tates the sharing of the same spectrum of resources among multiple users, leading to improved performance, particularly in non-terrestrial communication with extensi ve coverage [16]. Similarly , by incorporating NOMA into terrestrial satellite HetNets is gaining prominence as a promising approach to boost communication performance and optimize spectrum resource utilization [17]. A. Resent Academic Advances Numerous research efforts hav e aimed to enhance NOMA system performance, covering areas such as interference man- agement, sum rate optimization, energy efficienc y maximiza- tion, power and bandwidth allocation, and distributed cluster allocation. For example, in [18] and [19], authors proposed optimization algorithms to tackle the challenge of interference management in NOMA-based communication systems. In [20], researchers e xplored a scheme for optimizing po wer and bandwidth distrib ution in NOMA-based downlink het- erogeneous networks, where each base station organizes into clusters and independently manages power and bandwidth allocation. Additionally , in reference [21], the authors delved into maximizing energy ef ficiency in NOMA-assisted net- works through bandwidth and power allocation. Furthermore, they derived a mathematical closed-form equation and an optimization algorithm for power and bandwidth using the generalized Dinkelbach method. Sev eral other research works hav e also in vestigated optimization problems in NOMA- assisted wireless networks [22]–[24]. In recent years, integrat- ing NOMA with satellite communication systems has attracted considerable attention from researchers aiming to le verage the unique adv antages of these technologies for optimal resource utilization. This trend has led to many studies, including those presented in [12], [25], exploring various optimization tech- niques for NOMA-assisted satellite networks. For example, in [12], a nov el iterativ e method for User equipment association, subchannel allocation, and power allocation was proposed, resulting in improved system throughput compared to existing methods. In [26], researchers focused on beamforming opti- mization in NOMA non-terrestrial IoT networks with a multi- beam architecture. Recent literature highlights the gro wing interest of researchers in optimizing resource utilization by integrating HAP communication systems with NOMA. Similarly , the authors of the referenced works [27], [28] concentrate on designing beamforming vectors and efficient resource allocation in NOMA terrestrial and non-terrestrial systems, respecti vely . Additionally , in [28], authors addressed user association and power optimization by presenting a math- ematical closed-form expression to find the power solution, which was then incorporated into user association schemes to achiev e globally optimal results. Furthermore, reference [25] tackled the optimal allocation of resources in IoT -based NOMA-enabled terrestrial networks integrated with satellite communications. The authors proposed a heuristic algorithm, such as particle swarm optimization, for both power and bandwidth optimization, utilizing a L yapunov frame work to find the optimal solutions. This work underscores the potential of integrating NOMA with non-terrestrial networks to enhance communication performance in IoT terrestrial networks. The integration of wireless caching and NOMA has gar- nered substantial attention in recent years as researchers aim to enhance the performance and efficienc y of wireless com- munication systems [18], [29]- [30]. Wireless caching, storing frequently accessed content at edge nodes in the network, can alle viate netw ork congestion and enhance low-latenc y communication performance [31], [32]. This technology can also enhance the performance of terrestrial non-terrestrial HetNets by reducing the demand on the backhaul link [33], [34]. Numerous studies hav e explored the integration of wire- less caching and NOMA, examining various aspects of the technology and its potential advantages. In [29], researchers delved into optimizing energy-ef ficient resource allocation in NOMA networks incorporating terahertz communication and caching, intending to improve energy efficienc y by judi- ciously utilizing resources in NOMA networks. Additionally , in [35] and [36], researchers in vestigated the application of deep learning algorithms for optimizing resource allocation in NOMA networks incorporating caching. In [37], two NOMA- based caching strategies were presented to mitigate latency in content deli very . The challenge of user association and power allocation in NOMA networks incorporating caching was addressed in [18] and [30], with the authors proposing joint algorithms to optimize these aspects. In conclusion, the integration of wireless caching and NOMA continues to be a subject of ongoing research and in vestigation. Recently , some researchers hav e also considered terrestrial- HAPS networks and ev aluate various aspects of system perfor- mance. Alidadi et al. [38] hav e in vestigated a fairness problem in terrestrial-HAPS network to maximize the minimum spec- tral efficienc y by optimizing the system resources. Ren et al. [2] hav e provided an adaptiv e delay minimization problem and optimized task splitting, power control, spectrum assignment, and computational resource allocation. Moreover , the authors of [39] have proposed a cell-switching methods in terrestrial- HAPS integrated networks to improve the energy consumption of the system. Alfattani et al. [40] have proposed a new optimization framew ork to enhance the users connecti vity and minimize the ener gy consumption in terrestrial-HAPS networks. Further, Zheng et al. [41] hav e studied the posi- tioning performance of terrestrial-HAPS integrated networks to improv e 3D positioning accuracy , the horizontal dilution of precision, and the vertical dilution of precision. Besides the abov e studies, Erdogan et al. [42] have proposed different COMPUTER COMMUNIA CTION 3 use-cases of terrestrial-HAPS network and inv estigated the physical layer security of the system. In addition, Shafie et al. [43] ha ve optimized the power allocation in terrestrial- HAPS network to maximize the spectral efficienc y under spe- cial correlation of channel gain and imperfect NOMA signal decoding. Then, the work in [44] has provided user association and codebook design scheme for the spectral efficiency of terrestrial-HAPS integrated networks. Of late, authors hav e also studied interference issues [45], mobile edge computing based task of floading [46], and terahertz communication in terrestrial-HAPS networks [47]. B. Recent Industry Advances The process of standardization for terrestrial non-terrestrial networks within the 3rd Generation Partnership Project (3GPP) commenced in the year 2017 [48]. The standardization en- deav or can be classified into tw o main domains: improve- ments for networks operating in non-terrestrial environments and enhancements for networks operating in terrestrial en- vironments. The primary objective of this initiative is to dev elop a univ ersally accepted benchmark for non-terrestrial communications, hence fostering substantial expansion within the satellite, HAPS, and UA V industry . The activities in the aforementioned domain hav e a twofold objecti ve, which is to guarantee that mobile standards are in line with the connection needs for secure functioning on platforms that are not on Earth. T able II provides a comprehensiv e summary of the objectiv es and results achiev ed by 3GPP in its endeav ors encompassing Rel-15 to Rel-17, together with the ongoing inv estigations for Rel-18. In the context of 3GPP , terrestrial non-terrestrial networks pertain to the application of satellites or HAPS to provide connectivity services, specifically in geographically isolated regions where conv entional cellular cov erage is insufficient. The core set of features added by 3GPP in Rel-17 aims to enable next-generation spectrum operation over terrestrial- satellite networks within the frequency range of FR1, encom- passing frequencies up to 7.125 GHz. The forthcoming Rel-18 of the 3GPP is focused on advancing the capabilities of next- generation operations inside terrestrial-satellite en vironments. The proposed enhancement aims to enhance the cov erage of handheld devices, in vestigate the feasibility of deploying networks in frequency bands abov e 10 GHz, tackle challenges related to mobility , ensure uninterrupted service transition be- tween terrestrial and non-terrestrial networks, and ev aluate the regulatory obligations associated with verifying user locations within the network [49]. The inclusion of non-terrestrial platforms in the preceding network generation was initially included in 3GPP’ s Rel- 15. This in volv ed the incorporation of signaling protocols to identify extraterrestrial users using subscription-based tech- niques. Furthermore, protocols were implemented to facilitate the reporting of essential characteristics pertaining to non- terrestrial platforms, including height, position, speed, and flight trajectory . In order to efficiently address non-terrestrial interference, especially in situations where there is a specific concentration of low-altitude non-terrestrial platforms, nov el measurement reports hav e been implemented for effecti ve management. In follo wing iterations, the 3GPP e xpanded its scope to cater to the requirements of linked non-terrestrial systems at the application layer , with a significant emphasis on security con- siderations. These releases hav e also established the ground- work for establishing the protocols by which non-terrestrial platforms engage with the T raf fic Management system, facil- itating synchronized and secure operations of non-terrestrial platforms inside the network. The upcoming release of 3GPP’ s Rel-18 aims to provide specialized next-generation spectrum support that is specifically designed for devices running on aerial vehicles. This dev elopment is in response to the e volving use cases of next-generation technologies. According to [49], the forthcoming progress will encompass the in vestigation of supplementary factors that can initiate conditional handov er . Additionally , it will incorporate the utilization of base station uptilting approaches to increase communication, as well as the integration of signaling mechanisms to indicate the beamform- ing capabilities of non-terrestrial platforms, alongside various other improvements. C. Motivation and Contributions Resource allocation in wireless networks has been a topic of significant research interest. The focus has been improving the performance and efficienc y of wireless caching networks, NOMA networks, and terrestrial-HAPs networks. Despite these efforts, most studies have only considered one network type in isolation rather than an integrated approach. Currently , there is a gap in the literature that addresses the comprehensive problem of efficient allocation of resources in caching-based NOMA for HAPs-terrestrial communication networks, partic- ularly concerning bandwidth, power , and user association. The complexities of resource allocation in these networks arise from the requirement for reliable wireless backhaul access networks and the added challenge of interference in multi-cell NOMA terrestrial-satellite heterogeneous networks. The objecti ve of this study is to address the challenges presented by NOMA-enabled vehicular-aided HetNet, which in volves RSU and HAPs VUEs sharing the same radio re- source. The HAPs are responsible for back-haul commu- nication links to the terrestrial RSU. T o achie ve this, the study begins by formulating a joint optimization problem considering the achie vable rate for cellular VUEs and the interference from RSU-HAPs. An optimization problem is dev eloped to consider factors such as user association, back- haul link limitations, and QoS requirements for each VUE. T o further enhance the system performance, the joint optimization formulation is broken down into sub-problems, including user association, efficient allocation of bandwidth, and transmission power . Subsequently , algorithms are proposed to optimize the system performance. This paper provides a comprehensi ve analysis of the resource allocation process, including a time complexity analysis of the proposed optimization algorithms. The ef fecti veness of the proposed scheme is demonstrated through simulation results. The main contributions of this study include: COMPUTER COMMUNIA CTION 4 T ABLE I: 3GPP standardization works on integrated terrestrial non-terrestrial networks. [48], [49] Release Advance in integrated terrestrial non-terrestrial networks Rel-15 The focus of Rel-15 is on New Radio (NR), a technology proposed for the purpose of supporting terrestrial non-terrestrial networks as outlined in the technical report [TR 38.811]. Additionally , this study provides pertinent use case possibilities for integrating terrestrial non-terrestrial networks and spectrum, explicitly focusing on the S-band and Ka-band frequencies. In addition, it delineates the dimensions of the footprint, the angle of assessment, the configuration of the beam, and the design of the antenna. Additionally , this release provides specific details regarding the channel propagation model as outlined in the technical report [TR 38.901]. Rel-16 The present release presents potential resolutions for the integration of new technologies inside terrestrial non-terrestrial integrated networks, as outlined in the technical report [TR 38.821]. The primary emphasis is placed on utilizing FR1 bands inside terrestrial non-terrestrial networks to facilitate the seamless integration and functioning of the Internet of Things. Additionally , this approach facilitates the identification of necessary modifications in the physical layer and other layers, considering the assumptions made during system-lev el simulations. In addition to this, the present study also examines the influence of resource optimization on the performance of terrestrial non-terrestrial networks. Moreov er, it integrates the utilization of terrestrial non-terrestrial networks in the context of next-generation communications, as indicated in the technical report [TR 22.822], to facilitate the provision of div erse services. Rel-17 The topic of Rel-17 pertains to the inclusion of narrowband IoT and machine-type communication in integrated terrestrial non- terrestrial scenarios, as referenced in the technical report [TR 36.763]. The technology is primarily designed to meet the unique requirements of IoT applications. Considerable focus has been dev oted to the architectural concerns for satellite access within the framew ork of 6G, as outlined in the technical report [TR 23.737]. This endeavor in volv es improvements in various aspects, such as advances in radio frequency and physical layer characteristics, optimizations in protocols, and more efficient management of radio resources. In addition, this process entails the selection of a suitable architectural framew ork, addressing challenges related to integrated terrestrial non-terrestrial roaming, and enhancing conditional handover procedures. Rel-18 The advancements pertaining to terrestrial non-terrestrial communication will in vestigate the extent of system cov erage for handheld devices in practical scenarios, as well as explore access capabilities beyond the 10 GHz frequency range for both stationary and mobile platforms. The study aims to investig ate the necessary conditions for the network-validated user location and address challenges pertaining to user mobility and the uninterrupted provision of services during transitions between terrestrial and satellite networks, as well as various non-terrestrial networks. 1) In this study , we present an up-link communication scenario in the HetNet model incorporating caching strategies for spectrum sharing between RSU and HAPs VUEs is presented. T o ensure optimal system perfor- mance, the interplay of inter-cell, cross-tier , as well as intra-cell interference is analyzed. The optimal decoding order for successiv e interference cancellation (SIC) is then calculated. T o assess system performance, a utility function is created that considers the achiev able rate of cellular UEs and the cross-tier interference generated by RSUs tow ards the satellite. The resource allocation problem is then posed as a system utility maximization problem, with joint user associations, back-haul con- straints, and VUE QoS requirements all considered. 2) The resource allocation optimization problem in this system model is highly non-linear and computation- ally complex, with Non-deterministic Polynomial (NP) complexity . T o tackle this challenge, the problem is decomposed into three sub-problems, which are solved independently . The first sub-problem focuses on the VUE-AP association and proposes an advanced prefer- ence relation, caching, and swapping-based algorithm. The algorithm’ s implementation in volves prioritizing satellite VUEs based on the channel coefficient ratios and considering the caching state and av ailability of the back-haul communication link in the association of ter- restrial VUEs. The second sub-problem yields a closed- form expression for bandwidth allocation. In contrast, the third sub-problem in volves transforming the non- con vex objectiv e function optimization into a con vex form through successive con vex approximation, which is then solved iterati vely through a po wer allocation algorithm. 3) T o optimize the resource allocation in the system, a three-stage iterati ve algorithm is proposed. The proposed approach for resolving the optimal allocation of re- sources in the up-link up-link communication scenario in HetNet communication in volves the iterative solution of three sub-problems, specifically user association, band- width allocation, and power allocation. A detailed anal- ysis of the time complexity of the proposed algorithms is carried out to ev aluate their computational efficienc y . The validity of the algorithms’ efficac y in optimizing resource allocation is established through simulation results, which also demonstrate the con ver gence of the proposed algorithms. I I . N E T W O R K M O D E L & M AT H E M A T I C A L F O R M U L AT I O N This section presents the caching-based NOMA-enabled vehicular -aided HetNet, which aims to enhance communica- tion network performance by efficiently using network re- sources. T o achieve this goal, a utility function is constructed, and an optimization problem is formulated, considering three essential elements of the network: bandwidth assignment for front-haul and back-haul communication, user association, and power allocation. This approach leverages the advantages of NOMA as compared to OMA, such as improved spectrum efficienc y and flexible interference management, to optimize resource allocation in the context of vehicular-aided HetNets [50], [51]. A. Network Model As shown in Figure 1, this study focuses on the up-link communication scenario in HetNet, which consists of M roadside units (RSUs) and one high altitude platform (HAP) denoted by l , serving N vehicular user equipment (VUEs). RSUs play a crucial role in the network by pro viding a communication infrastructure for vehicles, enabling efficient communication between vehicles and the HAP . W ithout RSUs, COMPUTER COMMUNIA CTION 5 HAPS VUE to RSU link RSU to HAPS link Inter-cell interference Cross-tier interference VUE to HAPS link RSU VUE V UE VUE VUE RSU RSU Fig. 1: System Model the vehicles would hav e to rely solely on their own com- munication capabilities, leading to potential communication breakdowns and de graded network performance. The total system bandwidth, B , is divided into front-haul and back-haul links, with (1 − η ) B allocated to front-haul communication between vehicles and RSUs, and η B allocated to back-haul communication between RSUs and the HAP . The system incorporates wireless caching technology for data of floading in RSUs to meet network requirements. The NOMA technology enables multi-user transmission in this system, while com- munication between RSU-HAP links occurs over the C-band. The system model also considers several types of co-channel interference, including cross-tier HAPs interference, inter-cell interference, and intra-cell interference caused by NOMA. A comprehensiv e analysis of the system model is presented in subsequent sections, highlighting the importance of the proposed system design for efficient utilization of network resources and improved network performance. 1) T errestrial Communication Model: In a terrestrial com- munication network, NOMA technology allo ws an RSU to serve multiple VUEs [52]. The set of VUEs served by a particular RSU, denoted as m , determines the signal received by these VUEs at their associated RSU. Mathematically , this can be represented as: y n,m = √ p n,m h n,m s n,m + [ Φ I 1 n,m + [ Φ I 2 n,m + [ Φ I 3 n,m + ς n (1) Follo wing that, the equation (1) describes the impact of different types of co-channel interference, including intra and inter-cell interference as well as cross-tier interference denoted by [ Φ I 1 n,m , [ Φ I 2 n,m , and [ Φ I 3 n,m , respectively . These sources of interference are mathematically represented as follows: [ Φ I 1 n,m = X n ′  = n ∈ L m √ p n ′ ,m h n,m s n ′ ,m . (2a) [ Φ I 2 n,m = X j ∈ M /m X i ∈ L j √ p i,j h i,m s i,j . (2b) [ Φ I 3 n,m = X i ∈ L l √ p i,l h i,m s i,l . (2c) In a NOMA-enabled HetNet, the transmission power from the n th VUEs to wards the m th RSU and the HAPs is represented as p n,m and p n,l , respectiv ely . The channel coefficients be- tween the n th VUE and m th RSU are denoted as h n,m , while the signal transmitted from the n th VUE to the m th RSU and from the j th VUE to the HAPs are represented by s n,m and s j,l , respectiv ely . The channel between VUE-RSU/HAPs is modelled as additiv e white Gaussian noise (A WGN) and assumed to follo w a complex Gaussian distribution, ς ∼ CN(0 , σ 2 ) , where σ 2 is the variance. Giv en these conditions, the signal-to-interference-plus-noise ratio (SINR) between the n th VUE and m th RSU can be calculated as follows: Γ n,m = | h n,m | 2 p n,m Φ I 1 n,m + Φ I 2 n,m + Φ I 3 n,m + σ 2 , (3) Where the intra, inter-cell, as well as cross-tier interference, are represented by Φ I 1 n,m = X n ′  = n ∈ L m p n ′ ,m | h n,m | 2 . (4a) Φ I 2 n,m = X j ∈ M /m X i ∈ L j p i,j | h i,m | 2 (4b) Φ I 3 n,m = X i ∈ L l p i,l | h i,m | 2 . (4c) respectiv ely . Follo wing that, the allocation coefficient, η , determines the portion of bandwidth allocated to the backhaul network. The user association between VUEs, RSUs, and HAP is denoted by COMPUTER COMMUNIA CTION 6 the matrix U [ u n , m ] N × M + l . This binary matrix indicates the connection between VUEs and RSU/HAPs, where u n,m = 1 if VUE n is associated with RSU/HAPs, and 0 otherwise. By utilizing the user association information and the allocation coefficient, the transmission rate from VUE n to RSU m can be calculated based on the av ailable bandwidth, channel quality , and other rele vant factors. This rate represents the amount of data that can be transmitted over a certain period. R n,m = (1 − η ) B log 2 (1 + Γ n,m ) , (5) Meanwhile, the overall achiev able rate at the m th RSU can be expressed as follows: R m = X n ∈ N u n,m R n,m . (6) 2) VUE-HAPs Communication: The signal receiv ed at the HAPs from the n th VUE, denoted by y n,l , can be expressed mathematically as: y n,l = √ p n,l h n,l s n,l + d Φ I 1 n,l + d Φ I 2 n,l + ς l (7) In this expression, d Φ I 1 n,l and d Φ I 2 n,l represent the intra-cell interference and the interference from other RSUs at the HAP , respectiv ely . Mathematically , can be expressed as: d Φ I 1 n,l = X n ′ ∈ L l /n √ p n ′ ,l h n,l s n ′ ,l (8a) d Φ I 2 n,l = X j ∈ M X i ∈ L j √ p i,j h i,l s i,l (8b) Similarly , the channel coefficient between a VUE and the HAP is denoted by h u,l . The SINR that the HAP receiv es from the VUE can be expressed as a function of this coefficient as follows: Γ n,l = | h n,l | 2 p n,l Φ I 1 n,l + Φ I 2 n,l + σ 2 (9) Where, Φ I 1 n,l = X n ′ ∈ L l /n p n ′ ,l | h n,l | 2 (10a) Φ I 2 n,l = X j ∈ M X i ∈ L j p i,j | h i,l | 2 . (10b) respectiv ely . Similarly , VUEs associated with HAPs are subject to both intra-HAPs and RSU-originated interference, which can im- pact their SINR. Furthermore, since the transmission power of HAPs’ VUEs is constant, the SINR at the HAPs is also affected by cross-tier interference from RSUs’ VUEs. 3) Caching Model: Local caching is a popular feature implemented in RSUs to reduce network traf fic and alleviate backhaul pressure by storing frequently requested data at the network’ s edge. This allows VUEs to access content directly from the local storage of RSUs instead of relying on the limited capacity of backhaul links. The effecti veness of the caching strategy is influenced by the power allocation v alues. The caching index, denoted as X [ x n,m ] N × M , reflects the success of caching UE n ’ s content at RSU m during the caching phase, where x n,m = 1 indicates successful caching and x n,m = 0 indicates otherwise. It is worth noting that each VUE’ s content can only be cached at a single RSU, and all RSUs must hav e the same buf fer capacity . The constraints enforced by this approach can be expressed as P m x n,m ≤ 1 and P i ∈{ L m } x i,m ≤ x max . 4) Back-haul Link Capacity Model: In the proposed sys- tem, as illustrated in Figure 1, VUEs access content via HAPs, and front-haul communication is constrained by the back- haul communication link between the RSUs and the HAPs. Caching ( x n,m = 1 ) allows BSs to serve cached content to VUEs, reducing the back-haul burden and link ov erhead. The Orthogonal Frequency Division Multiple Access (OFDMA) technique is used for communication between RSUs and HAPs, as reported in [53]. The system also specifies the Signal-to-Interference-plus-Noise Ratio (SINR) of the back- haul link from RSUs m to the HAPs. Γ m,l = p m,l | h m,l | 2 P m ′ ∈ M \ m p m ′ ,l | h m,l | 2 + σ 2 . (11) In addition, the rate equation for the backhaul communication link from the m th RSU to the HAPs can be expressed as: R m,l = Ψ m log 2 (1 + Γ m,l ) , (12) where Ψ m = | L m |− P i ∈ L m x i,m | L m | η B M and | L m | = P i ∈ c u i,l . The achiev able back-haul transmission rate is R l = X m ∈ M R m,l . (13) 5) Mechanism of SIC Decoding: The proposed system utilizes the NOMA technique to address the challenges of multiplexing multiple signals in the same frequency band [54]. The technique relies on the SIC approach to prioritize the decoding process based on the receiv ed signal strength, with the user recei ving the strongest signal being decoded first. This method is employed for both uplink and downlink transmissions, with the user with the best channel conditions receiving the highest priority for decoding in an uplink NOMA scenario with equal transmitter capabilities [52]. This approach enables the system to allocate resources more efficiently , leading to improv ed communication performance. In the pro- posed system model, various factors such as the inter-cell interference Φ I 1 n,m , interference from outside the cell Φ I 2 n,m , and interference from the HAPs to RSU Φ I 3 n,m are considered in determining the channel state. The system’ s approach is formalized in a theorem presented in the paper as follows. Theorem I : For NOMA networks featuring VUE-RSU uplink and multiple cells, it is necessary for the channel coefficients of users n 1 and n 2 in cell m (where m ∈ M ) to meet a specific requirement in order for the SIC technique to be applied successfully and eliminate n 1 ’ s signal from n 2 ’ s signal. | h n 1 ,l | 2 ≥ | h n 2 ,m | 2 . (14) Proof: The power receiv ed by RSU m from VUES n 1 and n 2 are as follows: n r 1 = | h n 1 ,m | 2 ( p n 1 ,l + p n 2 ,m ) + Φ I 2 n 1 ,m + Φ I 3 n 1 ,m + σ 2 n r 2 = | h n 2 ,m | 2 ( p n 2 ,m + p n 2 ,m ) + Φ I 2 n 2 ,m + Φ I 3 n 2 ,m + σ 2 (15) COMPUTER COMMUNIA CTION 7 The condition for successful implementation of the SIC tech- nique in uplink RSU-HAPs multiple-cell NOMA networks, used to decode and eliminate user n 1 ’ s signal from user n 2 ’ s signal, can be expressed as follo ws: Φ I 2 n 1 ,m = Φ I 2 n 2 ,m = P j ∈ M /m P i ∈ L j p i,j | h i,m | 2 and Φ I 3 n 1 ,m = Φ I 3 n 2 ,m == P i ∈ L l p i,l | h i,m | 2 . Analysis shows that for user n 1 ’ s signal to be decoded successfully , its recei ved signal must be greater than or equal to that of user n 2 , i.e., | h n 1 ,m | 2 ≥ | h n 2 ,m | 2 . This completes the proof. Arranging the channel coefficients of the VUEs connected to the RSU m in ascending order facilitates the systematic implementation of the SIC technique. By decoding the signal from the VUEs with the superior channel condition first, communication performance can be improv ed. W ( L m ) ≜ | h 1 ,m | 2 ≥ | h 2 ,m | 2 ≥ · · · ≥ | h N ,k | 2 , ∀ m. (16) Therefore, based on the above assumption, the SINR of RSU from VUEs n can be expressed as J n , where J n represents the set { 1 , 2 , · · · , n } . Γ n,m = | h n,m | 2 p n,m P i ∈ L m \ J n p i,m + Φ I 2 n,m + Φ I 3 n,m + σ 2 (17) B. Pr oblem F ormulation In a NOMA-enabled vehicular-aided HetNet, the primary objectiv e is to achieve high transmission rates while limiting the impact of cross-tier interference on the QoS requirements of VUEs connected to a HAP . NOMA technology enables multiple users to share the same frequency and time resources using power -domain multiplexing, but cross-tier interference can still occur and af fect the QoS of VUEs. T o address this challenge, the utility function for VUEs must consider both the achie vable transmission rate and the impact of cross- tier interference while taking into account the specific QoS requirements of the VUEs and the methods used to calculate the utility function. Therefore to address this, W e formulate a weighted sum of the achie vable transmission rate and a penalty function for the interference level. By adjusting the weight factors, the system can balance the two objectiv es based on the system’ s priorities. This balance is critical for providing reliable and efficient communication services in vehicular networks, as demonstrated mathematically as follows. F n,m = X n ∈ L m  R n,m − Ω | h n,l | 2 p n,m  , (18) The interference pricing factor Ω plays a crucial role in balancing the trade-of f between transmission rate and cross- tier interference. While a higher transmission rate increases the utility , cross-tier interference can significantly reduce it. In our simulations, we adopt a dynamic approach for setting the parameter Ω to adapt to real-time network conditions and user requirements. Therefore, the total utility of the system is the sum of the utilities of each RSU, represented by F m . F m = X n ∈ L m F n,m . (19) The problem of resource allocation in the uplink NOMA- enabled vehicular-aided HetNet can be mathematically for- mulated as an optimization problem. The objectiv e is to maximize the system utility , subject to constraints such as power budget, QoS requirements of users, and interference limits. Specifically , let F m denote the utility of the RSU, the problem can be expressed as: max { U ,η , P } X n ∈ L m  R n,m − Ω t | h n,l | 2 p n,m  (20a) C 1 : R n,m > R n , ∀ m ∈ M , ∀ n ∈ N , (20b) C 2 : X n ′ ∈ L m (1 − g n ′ ,m ) R n ′ ,m < R m,l , ∀ m ∈ M . (20c) C 3 : W ( L m ) , ∀ m ∈ M + l, (20d) C 4 : u n,m ∈ { 0 , 1 } , ∀ m ∈ M + 1 , ∀ n ∈ N , (20e) C 5 : X M + l j =1 u n,j = 1 , ∀ n ∈ N , (20f) C 6 : p n,m ∈ [0 , P max ] , ∀ n ∈ N , ∀ m ∈ M + l, (20g) C 7 : η ∈ (0 , 1) , (20h) The resource allocation problem in the uplink NOMA-enabled vehicular -aided HetNet network is a complex optimization task that aims to jointly optimize the allocation of resources, including the VUE-RSU/HAPs association v ariable U , the indicator for bandwidth assignment η , and the transmission power vector P . The problem is subject to a set of constraints that must be satisfied to ensure optimal resource allocation. 1) Constraint (20b) is imposed to guarantee that each VUEs quality of service (QoS) requirement is met. 2) Constraint (20c) is imposed to ensure that the total rate achiev ed by each cell m does not exceed its available backhaul link rate. 3) Constraint (20d) ensures that the decoding order of users in each cell m is maintained through the use of the access point-user association matrix U [ u n,m ] N × l . 4) Constraint (20e) restricts each VUEs only to be associ- ated with one access point, either an RSU m or HAPs l , with un, m = 1 indicating association and u n,m = 0 indicating no association. 5) Constraint (20f) enforces that each VUE can only be served by one AP at a time. 6) The power constraint is defined in (20g). 7) Constraint (20h) represented the bounds for the back- haul bandwidth allocation factor . These constraints ensure that the optimal allocation of re- sources in the up-link NOMA-enabled vehicular -aided HetNet network is optimized and meets the necessary requirements. The optimization problem described in equation (20) presents a challenging mixed-integer nonlinear programming problem that is difficult to solve and optimize. The problem is characterized by non-con ve xity and NP-hard complexity , making it challenging to obtain a globally optimal solution. Additionally , the user association strategy dynamically af fects the channel conditions, further complicating the optimization problem. T o overcome these challenges, the joint optimization problem is decomposed into sev eral distinct subproblems, re- sulting in improved solution efficienc y . This approach enables COMPUTER COMMUNIA CTION 8 the optimization of each subproblem separately , leading to bet- ter con ver gence and reducing the computational complexity of the ov erall problem. By breaking down the joint optimization problem, it becomes possible to solve for the optimal resource allocation with reduced computational resources and time. I I I . F R A M E W O R K F O R O P T I M A L A L L O C A T I O N O F R E S O U R C E S The optimization problem specified in equation (20) is a challenging mixed-inte ger nonlinear programming issue with computational difficulties and dif ficulties in optimizing the solution. Due to non-con vexity and NP-hard complexity , obtaining a globally optimal solution is challenging. Addi- tionally , the user association strategy dynamically influences channel conditions, adding further complexities. T o address these issues, the joint optimization problem is di vided into three sub-problems to improve solution efficienc y . The first sub-problem de velops a methodology for UE association based on the current caching state and user preferences with fixed power -to-bandwidth ratios. The second sub-problem focuses on bandwidth allocation. Finally , the third sub-problem deals with po wer allocation. An iterative algorithm is proposed to find the optimal po wer allocation, using information from previous sub-problems on UE association and allocation of bandwidth. The methodology for each sub-problem is detailed in the following section. A. User Association Method The sub-problem for UE association can be expressed as follows: max { U } X n ∈ L m  R n,m − Ω | h n,l | 2 p n,m  (21a) C 1 : W ( L m ) , ∀ m ∈ M + l, (21b) C 2 : u n,m ∈ { 0 , 1 } , ∀ m ∈ M + 1 , ∀ n ∈ N , (21c) C 3 : X M + l j =1 u n,j = 1 , ∀ n ∈ N , (21d) T o address the UE association optimization sub-problem stated in equation (21), an algorithm is utilized that considers caching, preference relations, and swapping. Notably , the algo- rithm doesn’t account for the QoS requirements of the UE and the back-haul link, which provides a broader user swapping range and facilitates obtaining globally optimal solutions. 1) Pr eparation pr ocess: In the proposed resource alloca- tion method for the NOMA-enabled vehicular-aided HetNet network, the selection of VUEs connected to the HAPS is based on the ratio of channel coefficients, κ n = | h n,l | 2 | h n,m | 2 . This ratio is computed for each VUE and sorted in descending order . The top L l VUEs with the highest κ n values are then designated as the HAP users, while the remaining VUEs connect to the RSUs. The selection of the RSU is determined by the VUEs’ preference list, which is sorted according to the channel coefficients, with the position of RSU m in the preference list indicated by l m,n . The proposed method pro- vides a systematic approach to allocate resources in vehicular- aided HetNet networks while considering VUE preferences and channel conditions. This approach can help optimize the performance of the network while ensuring fair allocation of resources among the VUEs. 2) Sense and Action: The VUEs connected to the RSU transmit request signals in accordance with their preference list. The decision rule for any m 1 , m 2 ∈ M where m 1  = m 2 is stated as follows: L m 1 ,n ≻ L m 2 ⇔ N B n,m 2 ( { L m 1 , L m 2 } ) < N n,m 1 ( { L k 1 ∪ { n } , L m − 2 \{ n }} ) . (22) Equation (22) presents the preference of VUE n between RSU m 1 and m 2 , which is based on the RSU that can offer higher utility . T o make the best use of caching resources, VUEs prefer to connect to RSUs that hav e their desired content cached. Howe ver , the long communication distance can lead to a high path loss and reduce the benefits of caching. T o address this, the utility function is weighted by a factor α . Thus, when x n,m = 1 , the ev aluation function can be expressed as: ω n,v = x n,m (1 + Γ n,m ) α (1 − η ) Ξ v , ∀ v ∈ V (23a) Ξ v = (1 + Γ n,v ) α (1 − η ) (1 + Γ v ,l ) (1 − α ) ∥ L z /m ! (23b) The proposed method in this study inv olves assigning a weighting factor , denoted by α , to the front-haul link. The set V consists of all RSUs that are closer to VUE n than RSU m . The decision of VUE n to connect to a particular RSU depends on the v alue of ω n,v . If the v alue of ω n,v is greater than or equal to 1 for all RSUs in V , it implies that the rev enue generated by caching compensates for the increased path loss due to long-distance communication, and VUE n connects to RSU v . On the other hand, if the value of ω n,v lies between 0 and 1 for some RSUs in V , VUE n selects the RSU with the lowest ω n,v as it generates more rev enue than RSU m . The action of VUE n is denoted by Λ n , and the abov e conditions can be expressed as follows: Λ n =      u n,m = 1 , Case 1 , P v ∈ V u n,v = 1 , Case 2 , judge and swap by (19), Case 3 . (24) Similarly , the cased can be represented as follows Case 1: if x n,m = 1 , ω n,v ≥ 1 , ∀ v ∈ V , ∀ m ∈ M Case 2: if x n,m = 1 , 0 < ω n,v < 1 , ∃ v ∈ V , ∀ m ∈ M Case 3: if x n,m = 0 , ∀ m ∈ M (25) 3) Swap Matching pr ocess: For any pair of RSU m 1 , m 2 ∈ M , where m 1  = m 2 , and any pair of VUEs n 1 , n 2 ∈ N , where n 1  = n 2 such that u n 1 ,m 1 = 1 and u n 2 ,m 2 = 1 , the swapping matching process is explained as follows.. { L } n 2 n 1 = { L }\ { L m 1 , L m 2 } ∪ { L m 1 \{ n 1 } ∪ { n 2 }} ∪ { L m 2 \{ n 2 } ∪ { n 1 }} . (26) COMPUTER COMMUNIA CTION 9 Similarly , the rules for the swapping can be expressed as follows: { L } u, 2 u1( n 1 ,n 2 ) ≻ { L } ⇔ X m ∈ M F m ( { L } ) < X k ∈ M F m  { L } n 2 n 1  . (27) In Equation 27, the system utility function F m is defined as the sum of the revenue generated by all VUEs connected to RSU m , subtracted by the product of the channel coefficient and transmit power of each VUE. The matching L is only updated to L n 2 n 1 if the new matching results in a higher system utility compared to the previous matching. 4) End of the Algorithm: The algorithm consists of two processes: the sense and action process and the swap matching process. The sense and action process in volves continuously optimizing the VUE association matrix by calculating the user utility using Equation (22) and making changes until no VUE wants to switch access points. In the swap matching process, two VUEs are randomly selected, and the utility function is e valuated using Equation (26) to find the optimal AP- UE association. This process continues until con vergence is reached. B. Bandwidth Assignment Decoupling the original optimization problem (20) into a sub-problem for bandwidth allocation yields the following mathematical expression: max { η } X m ∈ M X n ∈ L m  R n,m − Ω | h n,l | 2 p n,m  (28a) C 1 : X n ′ ∈ L m (1 − g n ′ ,m ) R n ′ ,m < R m,l , ∀ m ∈ M . (28b) C 2 : W ( L ∗ m ) , ∀ m ∈ M + l, (28c) C 3 : η ∈ (0 , 1) , (28d) The term W ( L ∗ m ) in the equation refers to the decoding sequence of cell m with the optimal VUE association U [ u ∗ n,m ] . Using this, the closed-form solution for optimizing the band- width as giv en in Equation (28) can be obtained. The solution in volves maximizing the system throughput by allocating bandwidth to each cell based on the optimal VUE association. The optimization problem is decoupled into a sub-problem for bandwidth allocation, and the solution is obtained by iterativ ely adjusting the bandwidth allocations until the optimal solution is achieved. The mathematical details of this solution are provided below . Definition 1 : In the optimization sub-problem (28) for bandwidth allocation, the optimal value of η can be found by maximizing J B m ov er all m ∈ M . The expression for J B m is given as follows: J B m = P i ∈ L ∗ m x ′ i,m log 2 (1 + Γ i,m ) Ψ m Θ m + P i ∈ L ∗ m x ′ i,m log 2 (1 + Γ i,m ) , ∀ m ∈ M . (29) Where, Ψ m = | L ∗ m |− P i ∈ L ∗ m x i,m | L ∗ m | 1 M , x ′ i,m = 1 − x i,m and Θ m = log 2 (1 + Γ o m ) . Similarly , X i ∈ L ∗ m x ′ i,m (1 − η ) B log 2 (1 + Γ i,m ) ≤ Ψ m η B log 2 (1 + Γ k,s ) ⇒ η ≥ P i ∈ L ∗ m x ′ i,m log 2 (1 + Γ i,m ) Ψ m log 2 (1 + Γ m,l ) + P i ∈ L ∗ m x ′ i,m log 2 (1 + Γ i,m ) = J B m ⇒ η ≥ max m J B m . (30) Proof : In the optimization subproblem represented by equation (28), the optimal value of the allocation parameter η can be obtained as the maximum of J B m for all m ∈ M . For each cell m ∈ M , the v alue of J B m is defined based on the conditions (28c) and (28d). Using condition (28d), the expression in equation (30) can be deriv ed for each m ∈ M . Consequently , the optimization subproblem (28) can be transformed and formulated based on the deriv ed expression in equation (30) as follows. max { η } X m ∈ M X n ∈ L m  R n,m − Ω | h n,l | 2 p n,m  s.t. max m J B m ≤ η < 1 , ∀ n ∈ N , m ∈ M . (31) The solution to the bandwidth allocation optimization sub- problem, represented by equation (31), can be obtained by computing the optimal value of η ∗ that corresponds to the lower bound of the monotonically decreasing utility function. This can be achiev ed using bisection search or other efficient numerical optimization techniques. Once the optimal v alue of η ∗ is obtained, the optimal bandwidth allocation can be obtained using equation (30). C. P ower Allocation The mathematical expression for the sub-problem to find the transmission power is as follows: max { P } X m ∈ M X n ∈ L m  R n,m − Ω | h n,l | 2 p n,m  (32a) C 1 : R n,m > R n , ∀ m ∈ M , ∀ n ∈ N , (32b) C 2 : W ( L ∗ m ) , ∀ m ∈ M + l, , (32c) C 3 : p n,m ∈ [0 , P max ] , ∀ n ∈ N , ∀ m ∈ M + l. (32d) The objectiv e function of the optimization prob- lem is non-con vex and can be represented as P m ∈ M P n ∈ L m  R n,m − ω | h n,l | 2 p n,m  with respect to p n,m . T o address this issue, the objecti ve function is reformulated as follows: max { P }    (1 − η ∗ ) B X m ∈ M X n ∈ L ∗ m log 2  Υ 1 n,m  − (1 − η ∗ ) B X m ∈ M X n ∈ L ∗ m log 2  Υ 2 n,m  − ω X n ∈ M X n ∈ L ∗ m | h n,l | 2 p n,m    . (33) Where, Υ 1 n,m = | h n,m | 2 p n,m +   h B n,m   2 P i ∈ L ∗ m \ J n p i,m + Φ I 2 n,m + Φ I 3 n,m + σ 2 =   h B n,m   2 P i ∈ L ∗ m \ J n p i,m + Φ I 2 n,m + COMPUTER COMMUNIA CTION 10 Φ I 3 n,m + σ 2 Subsequently , the second term of the objectiv e function, denoted as (33), exhibits non-con vexity and requires transformation into a con ve x form. This non-con ve x problem can be effecti vely addressed using the successive con vex ap- proximation approach, which has been shown to con verge well and adhere to the Karush-Kuhn-T ucker (KKT) condition [29], [55]. The inequality mentioned abov e can be approximated to the upper bound of the logarithmic function [56], which con verges to t i = t i 0 . − γ = − X m ∈ M X n ∈ L ∗ m log 2   | h n,m | 2 X i ∈ L m \ J n p i,m + Φ I 2 n,m + + Φ I 3 n,m + σ 2  . (34) W e define it as follows. Definition 2 : Lower bounds at − ¯ γ can be obtained via the non-con vex term − γ of (34), which con ver ges at the local point p n,m = p n,m [ t ] . ˜ γ = X n ∈ L ∗ m log 2  Φ I 1 n,m [ t ] + Φ I 2 n,m + Φ I 3 n,m + σ 2  + X j ∈ M \{ m } X i ∈ L ∗ j log 2  Φ I 1 j,i + Φ I 2 j,i [ t ] + Φ I 3 j,i + σ 2  + 1 ln 2 X n ∈ L X i ∈ J n − 1 | h i,m | 2 ( p n,m − p n,m [ t ]) Φ I 1 i,m [ t ] + Φ I 2 i,m + Φ I 3 i,m + σ 2 + 1 ln 2 X n ∈ L ∗ k X j ∈\{ m } X i ∈ M ∗ j | h n,j | 2 ( p n,m − p n,m [ t ]]  Φ I 2 n,m [ t ] + Φ I 3 n,m + σ 2 . (35) As described by (35), the non-con vex and non-linear optimiza- tion problem (32) transformed effecti vely and expressed in the more trackable form (36) as indicated below . Subsequently , a method to find the optimal best solution for (36) is proposed through an iterativ e power allocation scheme. max { P }    (1 − η ∗ ) B X m ∈ M X n ∈ L ∗ m log 2  Υ 1 n,m  (36a) − (1 − η ∗ ) B ˜ γ − ω X n ∈ M X n ∈ L ∗ m | h n,l | 2 p n,m    . (36b) C 1 : R n,m > R n , ∀ m ∈ M , ∀ n ∈ N , (36c) C 2 : W ( L ∗ m ) , ∀ m ∈ M + l, , (36d) C 3 : p n,m ∈ [0 , P max ] , ∀ n ∈ N , ∀ m ∈ M + l. (36e) The optimization problem (36) is solved through an iterative power allocation scheme. In each iteration, an initial power value p n,m [ l ] is specified, whereas the solution for the trans- mission power is calculated using the standard optimization toolbox e.g. interior point method. The computed po wer at the current iteration is then treated as the initial po wer v alue for the next iteration, p n,m [ t +1] . The algorithm is executed repeatedly until con ver gence is achie ved. The successiv e con ve x approxi- mation approach used in this iterativ e po wer allocation scheme has been proven to provide a good conv ergence, satisfying the KKT conditions and yielding an effecti ve solution for the non- con vex optimization problem (36). Algorithm 1: Framework for UE Association 1 input: N ← UEs, M ← Base Station, l ← satellite, B ← Bandwidth ; 2 Initialization: W ( L m ) ← decoding order and UE priority list. ; 3 Execution: ; 4 while Until Con ver ge do 5 L l ← Calculate { κ n , ′ D escend ′ } / Satellite UEs. ; 6 Update the UE association matrix U ; 7 while Until Con ver ge do 8 for each n ∈ { 1 , · · · , N ′ } do 9 if x n,m = 1 then 10 Find ( V ) → claulate → Ω ; 11 end 12 Update U ← calculate (24); 13 Update W ( L m ) ← by U calculated in step 12 ; 14 end 15 end 16 while Until Con ver ge do 17 [ n 1 , n 2 ] ← Randomly selected users such that n m 1  = n m 2 ; 18 { M } n 2 n 1 ← solve (26) U ← solve (27) 19 end 20 end Algorithm 2: Dynamic T ransmission Po wer Alloca- tion with Ω Selection 1 Initialization: P o ← Transmission Power , t max , F o , Ω ← Initial V alue ; 2 Execution: ; 3 while t ≤ t max or er ror ≤ ϵ do 4 for each n ∈ { 1 , · · · , N ′ } do 5 if x n,m = 1 then 6 p n,m ← Calculate the power using (36) ; 7 end 8 end 9 Update Power p n,m [ t + l ] ← p n,m [ t ] , ; 10 Select Ω dynamically based on real-time network conditions and user requirements; 11 if Network is congested then 12 Ω ← Ω + δ ; 13 end 14 else if Network interfer ence is high then 15 Ω ← Ω − δ ; 16 end 17 else 18 Ω remains unchanged; 19 end 20 F[t] = solve the Utility Function with the current Ω ; 21 error = F[t] - F[t-1] ; 22 end I V . A L G O R I T H M D E S I G N F O R R E S O U R C E A L L O C A T I O N T o solve the optimization problem, the proposed algorithm adopts a two-stage approach. The first stage optimizes the user association, bandwidth allocation, and transmission po wer, while the second stage further improves the solution obtained in the first stage by performing AP switching. The algorithm COMPUTER COMMUNIA CTION 11 Fig. 2: Conver gence of Algorithm 1 is designed to iterativ ely perform the two stages until con ver- gence is achiev ed. The first stage of the algorithm is implemented using the alternating optimization technique, where each variable is optimized sequentially while keeping the others fixed. The user association and bandwidth allocation are optimized jointly , while the transmission power is optimized separately . The user association and bandwidth allocation optimization problem is solved using the subgradient method, while the transmission power optimization problem is solved using the successive con vex approximation approach. The second stage of the algorithm in volves randomly se- lecting a VUE and ev aluating the potential utility gain of switching to another AP . If the utility gain is positiv e, the VUE is switched to the new AP , and the optimization problem is solved again to update the user association and transmission power . This process continues until no further AP switching results in a positiv e utility gain. A. Complexity Analysis The computational complexity of the proposed algorithm depends on the number of VUEs and APs, as well as the con ver gence criteria. The user association and band- width allocation optimization subproblem have a complexity of O ( N UE N AP ) , while the transmission power optimiza- tion subproblem has a complexity of O  N UE N AP log 1 ϵ  , where ϵ is the accuracy of the solution. The complex- ity of the AP switching process is O  N 2 UE N AP  . There- fore, the o verall complexity of the proposed algorithm is O  T  N UE N AP + N UE N AP log 1 ϵ + N 2 UE N AP  , where T is the number of iterations required to achiev e con ver gence. V . R E S U L T A N D D I S C U S S I O N This section presents the simulation results that demonstrate the effecti veness of our proposed algorithms in mitigating interference and maximizing the sum rate in a NOMA-enabled vehicular -aided HetNet. Mitigating interference is crucial for improving the performance of wireless communication net- works, and the proposed algorithms aim to optimize resource allocation and interference management to achieve this goal. Fig. 3: Conver gence of Algorithm 2 Fig. 4: Comparison of Proposed W ith Others Benchmarks Schemes Fig. 5: Utility Function value across the number of BSs. T o mimic real-world conditions, we carefully chose the pa- rameters used in the experiments. The HAP’ s altitude is fixed at 1000 km, and the radius of each RSU is set to 50 m. Each user de vice and RSU has a maximum transmission power capacity of 23 dBm and 43 dBm, respectiv ely . The HetNet has a total system bandwidth of 20 MHz and an additiv e white Gaussian noise power of − 174 dBm/Hz. The weight factor in the proposed UE association approach is 0 . 99 , and there are 50 COMPUTER COMMUNIA CTION 12 Fig. 6: Utility Function value across the number of users Fig. 7: Utility Function value across users density Fig. 8: Utility Function value across Power V alues VUEs and 5 RSUs in the network. The Rayleigh and Rician fading models are used to model the terrestrial and satellite links, respectively . The results show that our proposed algorithms effecti vely mitigate interference and maximize the sum rate of the HetNet. Figure 2 compares the performance of the proposed scheme with fixed user association algorithms by plotting their con- ver gence. Each VUE’ s power is set to 23 dBm, and the weight factor is set to 0 . 99 . The VUE association algorithm and Fig. 9: Utility Function value across Power Spectral Density the fixed association algorithm begin with rapid increases in their utility curves, as shown in the figure. Howev er, the rate of increase slows as iterations continue until conv ergence. The proposed algorithm achie ves conv ergence in 500 to 600 iterations, which is f aster than the random swapping algorithm, which takes around 1000 iterations. This indicates that our proposed UE association algorithm has lo w computational complexity and can effecti vely mitigate interference in the network, leading to a higher sum rate. The con ver gence of Algorithm 2 is ev aluated with respect to various VUE association schemes, as shown in Figure 3. The results demonstrate that the utility function value stabilizes, indicating that the transmission power allocation using the suc- cessiv e conv ex approximation-based algorithm has conv erged. Furthermore, the proposed algorithm outperforms the others by considering the utility function value as a performance metric. Simulation results in Figure 4 sho w the impact of VUE’ s on the utility function for various schemes. Similarly , the proposed scheme is compared to three other schemes: ideal backhaul, random power allocation, and random power allo- cation with ideal backhaul. The results demonstrate that as the number of VUE’ s increases, so does the utility function for all four schemes. At the same time, the proposed scheme provides the same epsilon results as the ideal backhaul approach. When combined with ideal backhaul schemes, it outperforms ran- dom power allocation and random power allocation schemes. W ith the Ideal Backhaul scheme, the optimization problem’ s tractable re gion is expanded, which improves the system’ s value in hybrid networks that combine HAPS and ground infrastructure. Figure 5 depicts the relationship between the performance of the utility function and the VUE’ s for different RSU config- urations. The graph re veals a positiv e correlation between the number of RSUs and the utility function performance. This can be attributed to the increased a vailability of candidate RSUs for each VUE when the number of VUE’ s is constant. The improved selection of candidate RSUs results in an enhancement in the overall system performance, as indicated by the utility metric. Similarly , the results in Figure 6 analyze cross-tier in- terference impact on satellite networks. The proposed algo- COMPUTER COMMUNIA CTION 13 rithm’ s performance is compared to the Random Power Al- location approach, which utilizes the UE association scheme. Moreov er , the results demonstrate that the system’ s overall performance improves with the UE increase. The proposed algorithm outperforms the random Po wer Allocation scheme by effecti vely reducing the cross-tier interference from BSs in the satellite network. In comparison, the same effect is achiev ed by considering the negativ e impact of cross-tier interference and regulating UE transmission power , ensuring optimal quality of service for satellite UEs. Similarly , to re veal the ef fectiv eness of the proposed schemes, results are compared with the Random Power Al- location algorithm through an examination of the utility as a function of the UE density per RSU, as shown in Figure 7. The graph is based on a fixed number of 5 RSUs. The results sho w that as the UE density per RSU increases, the utility function increases, thanks to the higher transmission power of UEs. Furthermore, the proposed algorithm achieves a significantly larger system utility value than the Random Power Allocation approach. This difference becomes more apparent with more UEs. Similarly , results in Figure 8 show the relationship between the number of terrestrial RSUs and the efficiency of the system analyzed. The graph is based on an environment with 50 UEs and varying maximum transmission po wer le vels per UE: [10 , 25 , 20] dBm. Results demonstrate that a rise in the number of RSUs leads to improved system efficiency , aligning with the trend seen when ev aluating the UE density per RSU. Furthermore, elev ating the maximum allowable transmission power for each UE enhances system efficienc y by extending the feasible range of the power optimization problem. Figure 9 depicts the effect of the A WGN power spectral density on system efficienc y . This graph shows the impact of dif ferent le vels of A WGN on system performance for cases with 10 and 6 terrestrial RSUs and a fixed number of 100 UEs. The graph shows that increasing the A WGN power spectral density reduces system efficienc y . When the A WGN power spectral density is constant, the graph shows that increasing terrestrial RSUs improves system efficiency . This demonstrates the significant impact of the number of terrestrial RSUs on system efficiency in the presence of A WGN. V I . C O N C L U S I O N A N D F U T U R E W O R K In this study , we proposed a three-stage iterati ve resource optimization algorithm for a NOMA-based uplink caching heterogeneous network with an RSU-HAPs configuration. The proposed algorithm optimizes the resource allocation to im- prov e the utility performance of the network while considering the QoS constraints of terrestrial VUEs and the backhaul constraints of backhaul HAPS. In the first stage of the al- gorithm, we developed an improved caching and swapping algorithm that incorporated preference relations to optimize the RSU-VUE association sub-problem. In the second stage, we deriv ed a closed-form expression for the bandwidth allo- cation coefficient. Finally , the third stage utilized the succes- siv e conv ex approximation method to solve the non-con ve x power allocation sub-problem iterativ ely . The simulation re- sults demonstrated that the proposed algorithm significantly improv ed the network’ s utility performance. The results also showed that increasing the number of terrestrial RSUs and the maximum allow able transmission power for each VUE led to improv ed system efficienc y . Moreov er , increasing the number of RSUs and decreasing the A WGN power spectral density improv ed system efficiency in the presence of A WGN. Overall, our proposed algorithm successfully optimized resource allo- cation and enhanced system ef ficiency while considering the constraints of a NOMA-based uplink caching heterogeneous network with an RSU-HAPs configuration. For future research, we intend to expand this study to include more complex networks and explore other optimiza- tion aspects of such networks. Specifically , our focus will extend to areas such as minimizing latency and optimizing energy consumption. W e aim to contribute to the ongoing dev elopment of advanced communication networks for smart cities and vehicular applications by addressing these vital aspects in our future work. A C K N O W L E D G M E N T The authors extend their appreciation to the Deanship of Scientific Research at King Khalid Uni versity for fund- ing this work through lar ge group Research Project un- der grant number (RGP2/10/44). Princess Nourah bint Ab- dulrahman Univ ersity Researchers Supporting Project num- ber (PNURSP2024R384), Princess Nourah bint Abdulrah- man Univ ersity , Riyadh, Saudi Arabia. Research Support- ing Project number (RSPD2024R787), King Saud Univ ersity , Riyadh, Saudi Arabia. This study is supported via funding from Prince Sattam bin Abdulaziz Univ ersity project number (PSA U/2023/R/1444). R E F E R E N C E S [1] W . U. Khan, Z. Ali, E. Lagunas, A. Mahmood, M. Asif, A. Ihsan, S. Chatzinotas, B. Ottersten, O. A. Dobre, Rate splitting multiple access for next generation cognitive radio enabled leo satellite networks, IEEE T ransactions on Wireless Communications. [2] Q. Ren, O. Abbasi, G. K. Kurt, H. Y anikomeroglu, J. Chen, Handoff- aware distributed computing in high altitude platform station (haps)– assisted vehicular networks, IEEE Transactions on Wireless Communi- cations. [3] M. Ahmed, A. W ahid, S. S. Laique, W . U. Khan, A. Ihsan, F . Xu, S. Chatzinotas, Z. Han, A survey on star-ris: Use cases, recent adv ances, and future research challenges, IEEE Internet of Things Journal. [4] B. Al Homssi, A. Al-Hourani, K. W ang, P . Conder , S. Kandeepan, J. Choi, B. Allen, B. Moores, Next generation mega satellite networks for access equality: Opportunities, challenges, and performance, IEEE Communications Magazine 60 (4) (2022) 18–24. [5] H. Dong, C. Hua, L. Liu, W . Xu, R. T afazolli, Intelligent reflecting surface-aided integrated terrestrial-satellite networks, IEEE Transactions on W ireless Communications. [6] W . U. Khan, E. Lagunas, Z. Ali, M. A. Javed, M. Ahmed, S. Chatzinotas, B. Ottersten, P . Popovski, Opportunities for physical layer security in uav communication enhanced with intelligent reflective surfaces, IEEE W ireless Communications 29 (6) (2022) 22–28. [7] G. K. Kurt, M. G. Khoshkholgh, S. Alfattani, A. Ibrahim, T . S. Darwish, M. S. Alam, H. Y anikomeroglu, A. Y ongacoglu, A vision and frame work for the high altitude platform station (HAPS) networks of the future, IEEE Communications Surveys & T utorials 23 (2) (2021) 729–779. [8] D. Renga, M. Meo, Can high altitude platform stations make 6G sustainable?, IEEE Communications Magazine 60 (9) (2022) 75–80. [9] W . Liu, X. Hou, L. Chen, Y . Hokazono, J. Zhao, Interference coordi- nation method for integrated HAPS-terrestrial networks, in: 2022 IEEE 95th V ehicular T echnology Conference:(VTC2022-Spring), IEEE, 2022, pp. 1–6. COMPUTER COMMUNIA CTION 14 [10] B. Feng, G. Li, G. Li, Y . Zhang, H. Zhou, S. Y u, Enabling efficient service function chains at terrestrial-satellite hybrid cloud networks, IEEE Network 33 (6) (2019) 94–99. [11] S. Ji, M. Sheng, D. Zhou, W . Bai, Q. Cao, J. Li, Fle xible and distributed mobility management for integrated terrestrial-satellite networks: chal- lenges, architectures, and approaches, IEEE Network 35 (4) (2021) 73– 81. [12] L. W ang, Y . W u, H. Zhang, S. Choi, V . C. Leung, Resource allocation for NOMA based space-terrestrial satellite networks, IEEE Transactions on W ireless Communications 20 (2) (2020) 1065–1075. [13] W . U. Khan, M. A. Jamshed, E. Lagunas, S. Chatzinotas, X. Li, B. Ottersten, Energy efficienc y optimization for backscatter enhanced NOMA cooperativ e V2X communications under imperfect CSI, IEEE T ransactions on Intelligent Transportation Systems. [14] W . U. Khan, N. Imtiaz, I. Ullah, Joint optimization of noma-enabled backscatter communications for beyond 5g iot networks, Internet T ech- nology Letters 4 (2) (2021) e265. [15] A. Ihsan, W . Chen, W . U. Khan, Q. W u, K. W ang, Energy-efficient backscatter aided uplink NOMA roadside sensor communications under channel estimation errors, IEEE Transactions on Intelligent Transporta- tion Systems. [16] M. Asif, A. Ihsan, W . U. Khan, A. Ranjha, S. Zhang, S. X. W u, Energy-ef ficient backscatter-assisted coded cooperati ve-NOMA for B5G wireless communications, IEEE Transactions on Green Communications and Networking. [17] F . Zhao, W . Xu, W . Xiang, Integrated satellite-terrestrial networks with coordinated C-NOMA and relay transmission, IEEE Systems Journal 16 (4) (2022) 5270–5280. [18] I. Randrianantenaina, M. Kaneko, H. Dahrouj, H. ElSawy , M.-S. Alouini, Interference management in noma-based fog-radio access net- works via scheduling and power allocation, IEEE Transactions on Communications 68 (8) (2020) 5056–5071. [19] I. Budhiraja, N. Kumar , S. T yagi, Cross-layer interference management scheme for d2d mobile users using noma, IEEE Systems Journal 15 (2) (2020) 3109–3120. [20] A. Celik, M.-C. Tsai, R. M. Radaydeh, F . S. Al-Qahtani, M.-S. Alouini, Distributed cluster formation and power-bandwidth allocation for imper - fect noma in dl-hetnets, IEEE T ransactions on Communications 67 (2) (2018) 1677–1692. [21] A. J. Muhammed, Z. Ma, Z. Zhang, P . Fan, E. G. Larsson, Energy- efficient resource allocation for noma based small cell networks with wireless backhauls, IEEE T ransactions on Communications 68 (6) (2020) 3766–3781. [22] M. Ahmed, Z. Ali, W . U. Khan, O. W aqar , M. Asif, M. A. Javed, F . N. Al-W esabi, et al., Cooperative backscatter NOMA with imperfect SIC: towards energy efficient sum rate maximization in sustainable 6G networks, Journal of King Saud University-Computer and Information Sciences 34 (10) (2022) 7940–7947. [23] W . U. Khan, F . H. Memon, K. Dev , M. A. Javed, D.-T . Do, N. M. F . Qureshi, Ambient backcom in beyond 5G NOMA networks: A multi-cell resource allocation framework, Digital Communications and Networks. [24] A. Ihsan, W . Chen, M. Asif, W . U. Khan, Q. Wu, J. Li, Energy-efficient IRS-aided NOMA beamforming for 6G wireless communications, IEEE T ransactions on Green Communications and Networking 6 (4) (2022) 1945–1956. [25] J. Jiao, Y . Sun, S. Wu, Y . W ang, Q. Zhang, Network utility maximization resource allocation for noma in satellite-based internet of things, IEEE Internet of Things Journal 7 (4) (2020) 3230–3242. [26] J. Chu, X. Chen, C. Zhong, Z. Zhang, Robust design for noma-based multibeam leo satellite internet of things, IEEE Internet of Things journal 8 (3) (2020) 1959–1970. [27] X. Zhu, C. Jiang, L. Kuang, N. Ge, J. Lu, Non-orthogonal multiple access based integrated terrestrial-satellite networks, IEEE Journal on Selected Areas in Communications 35 (10) (2017) 2253–2267. [28] K. W ang, Y . Liu, Z. Ding, A. Nallanathan, M. Peng, User association and power allocation for multi-cell non-orthogonal multiple access networks, IEEE Transactions on W ireless Communications 18 (11) (2019) 5284– 5298. [29] H. Zhang, H. Zhang, W . Liu, K. Long, J. Dong, V . C. Leung, Energy efficient user clustering, hybrid precoding and power optimization in terahertz mimo-noma systems, IEEE Journal on selected areas in com- munications 38 (9) (2020) 2074–2085. [30] L. Qi, M. Peng, Y . Liu, S. Y an, Advanced user association in non- orthogonal multiple access-based fog radio access networks, IEEE T ransactions on Communications 67 (12) (2019) 8408–8421. [31] X. W ang, M. Chen, T . T aleb, A. Ksentini, V . C. Leung, Cache in the air: Exploiting content caching and delivery techniques for 5g systems, IEEE Communications Magazine 52 (2) (2014) 131–139. [32] L. Li, G. Zhao, R. S. Blum, A survey of caching techniques in cellular networks: Research issues and challenges in content placement and deliv ery strategies, IEEE Communications Surveys & Tutorials 20 (3) (2018) 1710–1732. [33] Z. Song, Y . Hao, Y . Liu, X. Sun, Energy-efficient multiaccess edge computing for terrestrial-satellite internet of things, IEEE Internet of Things Journal 8 (18) (2021) 14202–14218. [34] J. Li, K. Xue, D. S. W ei, J. Liu, Y . Zhang, Energy efficienc y and traffic offloading optimization in integrated satellite/terrestrial radio access networks, IEEE T ransactions on W ireless Communications 19 (4) (2020) 2367–2381. [35] Y . Fu, W . W en, Z. Zhao, T . Q. Quek, S. Jin, F .-C. Zheng, Dynamic power control for noma transmissions in wireless caching networks, IEEE W ireless Communications Letters 8 (5) (2019) 1485–1488. [36] Z. Y ang, Y . Liu, Y . Chen, N. Al-Dhahir, Cache-aided noma mobile edge computing: A reinforcement learning approach, IEEE Transactions on W ireless Communications 19 (10) (2020) 6899–6915. [37] Z. Ding, P . Fan, G. K. Karagiannidis, R. Schober, H. V . Poor , Noma assisted wireless caching: Strategies and performance analysis, IEEE T ransactions on Communications 66 (10) (2018) 4854–4876. [38] A. A. Shamsabadi, A. Y ada v , H. Y anikomeroglu, Joint beamforming and user association design for integrated haps-terrestrial networks, arXiv preprint [39] G. B. Koc ¸ , B. C ¸ ilo ˘ glu, M. ¨ Ozt ¨ urk, H. Y anikomeroglu, Haps-enabled sustainability provision in cellular networks through cell-switching, arXiv preprint [40] S. Alfattani, A. Y ada v , H. Y anikomeroglu, A. Y ongacoglu, Resource- efficient haps-ris enabled beyond-cell communications, IEEE W ireless Communications Letters 12 (4) (2023) 679–683. [41] H. Zheng, M. Atia, H. Y anikomeroglu, Analysis of a haps-aided gnss in urban areas using a raim algorithm, IEEE Open Journal of the Communications Society 4 (2023) 226–238. [42] E. Erdogan, O. B. Y ahia, G. K. Kurt, H. Y anikomeroglu, Optical haps eav esdropping in vertical heterogeneous networks, IEEE Open Journal of V ehicular T echnology 4 (2023) 208–216. [43] R. Shafie, M. J. Omidi, O. Abbasi, H. Y anikomeroglu, Power allocation for a haps-enabled mimo noma system with spatially correlated channels, in: 2022 IEEE Globecom W orkshops (GC Wkshps), IEEE, 2022, pp. 1377–1382. [44] ˙ I. Cumalı, B. ¨ Ozbek, G. K. Kurt, H. Y anikomeroglu, User selection and codebook design for noma-based high altitude platform station (haps) communications, IEEE Transactions on V ehicular T echnology 72 (3) (2022) 3636–3646. [45] A. A. Shamsabadi, A. Y adav , O. Abbasi, H. Y anikomeroglu, Handling interference in integrated haps-terrestrial networks through radio re- source management, IEEE Wireless Communications Letters 11 (12) (2022) 2585–2589. [46] T . Ovatman, G. K. Kurt, H. Y anikomeroglu, An accurate model for computation offloading in 6g networks and a haps-based case study , IEEE Open Journal of the Communications Society 3 (2022) 1963– 1977. [47] O. Abbasi, H. Y anik omeroglu, A cell-free scheme for uav base stations with haps-assisted backhauling in terahertz band, in: ICC 2022-IEEE International Conference on Communications, IEEE, 2022, pp. 249–254. [48] X. Lin, S. Rommer , S. Euler , E. A. Y avuz, R. S. Karlsson, 5g from space: An overvie w of 3gpp non-terrestrial networks, IEEE Communications Standards Magazine 5 (4) (2021) 147–153. [49] X. Lin, An ov erview of 5g advanced e volution in 3gpp release 18, IEEE Communications Standards Magazine 6 (3) (2022) 77–83. [50] Y . Liu, W . Y i, Z. Ding, X. Liu, O. A. Dobre, N. Al-Dhahir , Developing noma to next generation multiple access: Future vision and research opportunities, IEEE Wireless Communications 29 (6) (2022) 120–127. [51] Y . Liu, S. Zhang, X. Mu, Z. Ding, R. Schober, N. Al-Dhahir , E. Hossain, X. Shen, Evolution of noma toward next generation multiple access (ngma) for 6g, IEEE Journal on Selected Areas in Communications 40 (4) (2022) 1037–1071. [52] W . U. Khan, X. Li, A. Ihsan, Z. Ali, B. M. Elhalawany , G. A. S. Sidhu, Energy efficiency maximization for beyond 5g noma-enabled hetero- geneous networks, Peer-to-Peer Networking and Applications 14 (5) (2021) 3250–3264. [53] Y . Y ang, S. Zhang, J. P . Lie, R. Zhang, Flexible multi-group single- carrier modulation: Subcarrier mapping and power allocation optimiza- tion, IEEE Transact ions on Communications 67 (3) (2018) 2002–2014. COMPUTER COMMUNIA CTION 15 [54] W . U. Khan, A. Ihsan, T . N. Nguyen, Z. Ali, M. A. Javed, Noma-enabled backscatter communications for green transportation in automotive- industry 5.0, IEEE T ransactions on Industrial Informatics 18 (11) (2022) 7862–7874. [55] W .-S. Lai, T .-H. Chang, T .-S. Lee, Joint power and admission control for spectral and energy efficienc y maximization in heterogeneous ofdma networks, IEEE T ransactions on W ireless Communications 15 (5) (2016) 3531–3547. [56] S. Boyd, S. P . Boyd, L. V andenberghe, Conv ex optimization, Cambridge univ ersity press, 2004.