arXiv CSopen access

Empowering Rural Areas with Multi-radio Microwave Backhaul Supported by Digital Twin for 5G IAB-based FWA

Empowering Rural Areas with Multi-radio Microwave Backhaul Supported by Digital Twin for 5G IAB-based FWA

arXiv:2607.21310v1 [cs.NI] 23 Jul 2026

Anselme Ndikumana1 , Kim Khoa Nguyen 1 , Adel Larabi2 , and Mohamed Cheriet1 1 Synchromedia Lab, École de Technologie Supérieure, Université du Québec, QC, Canada {anselme.ndikumana, kim-khoa.nguyen; Mohamed.Cheriet}@etsmtl.ca 2 GAIA Montreal, Ericsson Canada {adel.larabi}@ericsson.com

and bandwidth, it is often cost-prohibitive in low-density regions with uncertain returns on investment. As a costeffective alternative, 5G Fixed Wireless Access (FWA) enables connectivity by equipping homes with rooftop Customer Premises Equipment (CPE) that wirelessly links to fixed cellular base stations [2]. The number of FWA connections is projected to increase from 160 million by the end of 2024 to 350 million by 2030, representing 19% of all fixed broadband connections [3]. Mid-band and mmWave frequencies help to enhance the capacity of FWA networks [4]. However, 5G FWA is a singlehop solution, which limits its coverage range. To extend coverage beyond a single hop, additional technologies are needed to enable multi-hop FWA. In [5], the authors proposed a unified solution that considers Integrated Access and Backhaul (IAB) in FWA to cover larger areas. An IAB network [6] consists of an IAB donor and multiple IAB nodes. The IAB donor is a base station connected to the Core Network (CN) via a fiber-optic link, while the IAB nodes are additional base stations that connect to the donor via wireless backhaul links. As shown in Fig. 1, in the IAB-based FWA deployment, CPEs can connect to either the IAB donor or the IAB nodes using wireless links. Each IAB node comprises a Distributed Unit (DU) and a Index Terms—5G, Microwave Backhaul, IAB-based Mobile Termination (MT) unit. The DU serves both CPEs FWA, Digital Twin, Energy-Efficient and downstream MT. The MT allows the node to operate as a relay node when connecting to its parent DU. The I. Introduction IAB donor includes both a DU and a Centralized Unit A. Background and Motivations (CU). Access to high-speed internet, particularly 5G networks, Deploying fiber-optic backhaul for an IAB donor is remains uneven worldwide. Many regions, especially rural often not cost-effective, particularly in rural areas with low areas, are either unconnected or rely on slower network population density and uncertain returns on investment. To technologies such as 2G and 3G. According to the Interna- address this challenge, we consider using microwave backtional Telecommunication Union (ITU), 2.2 billion people haul as a practical and efficient alternative for connecting remain offline in 2025 [1]. Expanding high-speed internet an IAB donor to the core network for internet access. Longaccess in rural areas is essential for fostering economic haul microwave radios operating in frequency ranges such growth, improving education, and enhancing the quality as 11 GHz, 71 − 86 GHz, and 191.7 − 194.8 T Hz have of life. While fiber optic deployment offers high reliability been tested in rural areas, particularly agricultural zones with limited broadband access [7]. Despite its potential, This work was supported by NSERC (under project ALLRP 566589the integration of long-haul microwave, FWA, and IAB as 21) and InnovÉÉ (INNOV-R program) through the partnership with a unified network framework remains relatively unexplored Ericsson and ECCC. We thank the Ericsson’s Montréal GAIA team for their constructive and helpful comments, which have significantly in the literature. By integrating long-haul microwave, IAB, improved the quality and clarity of this manuscript (corresponding FWA, and high-frequency bands as a unified network, it author: Anselme Ndikumana). becomes possible to extend coverage across vast rural areas

Abstract—For digital inclusion, high-capacity Internet access should be provided to rural areas to support a range of services and applications. Due to the high operating costs of fiber-optic deployment, Fixed Wireless Access (FWA) is becoming a more attractive internet solution for rural areas. However, 5G FWA is a one-hop solution with limited coverage. A multi-hop solution is needed for wider rural coverage. This work considers a unified solution combining long-haul microwave, 5G Integrated Access and Backhaul (IAB), and FWA to provide a multi-hop network for extended coverage and high network capacity in rural areas. A key challenge for such a network is that energy consumption increases with the number of hops, a problem that has been overlooked in the existing literature. To address this, we propose energy-efficiency microwave backhaul for IAB-based FWA as the Physical Twin (PT). We develop an energy-efficient strategy to optimize radio start-up, serving, sleeping, and wake-up states for microwave backhaul connecting 5G IAB-based FWA serving rural areas. By operating the network at reduced capacity during low utilization, we aim to minimize energy consumption. Then, we present a Digital Twin (DT) of PT to improve its performance. We solve the formulated optimization problem using deep Q-learning in DT and the optimization solver in PT. The simulation results show that our approach satisfies the data rate requirements while reducing energy consumption.

Rural areas typically have lower population density compared to urban and suburban environments. As a result, radio resources allocated to CPEs and MTs often remain underutilized, leading to inefficient use of resources [13]. • Providing on-site technical support in rural areas is often logistically difficult and cost-prohibitive. Therefore, network automation using DT should be considered to enable remote network monitoring, diagnostics, and fault management. Despite its potential, the application of DT technology in rural connectivity scenarios is still largely underexplored in the existing literature. •

Optical fiber

CN

CU

IAB donor

DU

IAB node

IAB node MT DU

IAB node

MT DU

CPE

CPE

CPE

CPE Wireless backhaul

Figure 1: Illustration example of 5G IAB-based FWA.

while maintaining high data rates. We can use a Digital Twin (DT) to model a microwave backhaul connecting an IAB-based FWA network to the internet, where the DT continuously reflects the physical microwave’s processes, dynamics, and states. In other words, the DT serves as a virtual representation of the Physical Twin (PT), while the actual microwave constitutes the PT. DT [8] can play a transformative role in enhancing microwave backhaul service in rural areas. Instead of sending technicians to rural areas for every microwave backhaul malfunction, many network issues, such as remote antenna alignment or configuration changes, can be diagnosed and addressed remotely using the DT, significantly reducing operational costs and response times while improving microwave backhaul service reliability. B. Challenges for Deploying Long-haul Microwave and 5G IAB-based FWA Long-haul microwave and multi-hop IAB-based FWA networks present several critical challenges when deployed to serve rural areas: • Combining long-haul microwave with IAB-based FWA offers a promising solution for extending broadband coverage in rural areas. However, energy consumption tends to increase with the number of hops, and current research [7], [9]–[11] has not yet addressed this issue. • In multi-radio microwave backhaul, determining which radios to place in deep sleep mode during periods of low traffic, and when to reactivate them as demand increases, is a complex task. Efficient radio management is essential to achieving energy savings without degrading network performance [12]. • Energy-saving techniques, such as deep sleep, can adversely affect network performance during unexpected traffic surges. Delays in waking up radios may lead to service degradation or increased latency [12].

C. Contributions In this paper, we propose an energy-efficient microwave backhaul framework, supported by DT, to connect an IABbased FWA network serving rural areas to address the key challenges discussed above. The main contributions of this paper are summarized as follows: • We propose a multi-radio microwave backhaul system where each microwave node and IAB donor is equipped with multiple microwave radios. These radios, considered part of the Physical Twin (PT), can dynamically enter low-power states to reduce energy consumption during periods of low network utilization. Unlike existing literature, which typically models radios in only three states, off, on, and deep sleep, and ignores the latency associated with transitioning between these states, we introduce a more realistic five-state model: completely off, startup, serving, deep sleep, and wake-up. We then define state and action spaces to manage state transitions while minimizing energy consumption. • We develop a DT model for the microwave backhaul that receives network metrics from the PT. The DT uses deep Q-learning (DQL) [14] to optimize state transitions of the microwave radios, aiming to minimize energy consumption when the network is underutilized. The optimized transition matrix (i.e., action-state mapping) is then sent to the PT to guide its operations. • Upon receiving feedback from the DT, at the PT, the PT uses an optimized transition matrix to minimize energy consumption based on Age of Processing (AoP), while ensuring the data rate requirements of the IABbased FWA serving rural areas are met. We solve this problem using optimization solver. In this work, we adopt the AoP metric [15] to evaluate the timeliness of the PT’s sending of network metrics to the DT. The DT then optimizes radio states, sends the optimized states as feedback to the PT, and the PT subsequently uses this feedback to improve its operations. AoP is an extension of the Age of Information (AoI). This metric represents the time elapsed between the generation of a status at the source node and its most recent update at the destination node [16]. However, in practical systems, such as microwave backhaul, useful status information can be obtained only after processing the collected data from

the PT, e.g., computation or inference. Therefore, AoP on mmWave channel modeling for a 5G FWA network incorporates this additional computation delay into the operating at 60 GHz. Furthermore, [25] analyzed FWA AoI framework, offering a more comprehensive view of performance across multiple frequency bands, specifically end-to-end freshness. It has been applied in various real- 28 GHz, 60 GHz, and 140 GHz, highlighting trade-offs time applications, including data sampling, offloading, and between higher-frequency path loss and the benefits of processing in Internet of Things (IoT) systems [15], as well larger bandwidths, which enable higher channel capacity. In as data offloading for vehicular networks [17], [18]. [26], the authors evaluated the deployment of 60 GHz FWA. The remainder of the paper is organized as follows: This study collectively emphasizes the potential of highSection II reviews related work, and Section III presents the frequency bands in delivering high-capacity FWA services. system model. In Section IV, we formulate the optimization For rural connectivity, the authors in [27], [28] proposed problem, and Section V outlines the proposed solution. using digital twins and closed-loop systems to manage radio Section VI includes the performance evaluation. Finally, resource allocation in FWA networks. Although promising, we conclude the paper in Section VII. the proposed approaches are also limited to single-hop FWA networks. II. Literature Review IAB-based FWA and Resource Allocation: In [5], the We categorize the existing related works into three main authors proposed an integrated approach combining IAB groups: (i) microwave backhaul and radio link bonding, (ii) and FWA in the high-frequency 28 GHz band to deliver FWA and resource allocation supported by DT, and (iii) high-capacity connectivity to residential homes. In [29], IAB-based FWA and resource allocation. the use of unmanned aerial vehicles as IAB nodes was Microwave Backhaul and Radio-link Bonding: Microwave explored to enhance the flexibility and adaptability of IAB is widely used across a range of frequency bands and is network topologies. Given that high-frequency bands are expected to remain a vital transport technology for 5G prone to significant path loss, especially when obstacles networks, as noted in [19]. In [7], long-haul microwave are present, [11] proposed a multi-band solution combining radios operating in the 11 GHz, 71–86 GHz, 191.7–194.8 microwave and millimeter-wave in 5G new radio systems. GHz spectrum were evaluated for use in a rural agricultural Their approach mitigates outages by instantly rerouting region lacking broadband connectivity. A key technique traffic to sub-6 GHz links when mmWave connections for increasing capacity in such a network is radio-link become unavailable. Similarly, [30] introduced an archibonding, which aggregates data across multiple frequency tecture that integrates mmWave and sub-6 GHz bands carriers. The authors in [20] explored triple-band scheduling to address mmWave blockages and intermittent connecinvolving the 28 GHz band, the E-band (71–76 GHz paired tivity. They focused on packet scheduling strategies that with 81–86 GHz), and the Terahertz (THz) band. The leverage both interfaces to maintain service reliability. integration of these high-frequency bands is better suited In rural environments, where the population density is to high-density urban environments, given their limited low, radio resources allocated to CPEs and MTs are coverage and high path loss. In [21], the potential of often underutilized. Addressing this, [10] highlighted the high-power amplifier modules was examined for enhancing importance of radio resource allocation coordination in long-reach E-band systems in a radio-link bonding setup, IAB networks, which can be either distributed (managed improving backhaul performance for long distances. The by individual IAB nodes) or centralized (managed by authors in [12] pointed out that microwave backhaul links the IAB donor). They proposed a hybrid coordination are often underutilized, with usage rarely exceeding 50%. method that combines both centralized and distributed This underutilization presents an opportunity to reduce strategies. However, their approach does not address the energy consumption by powering down some microwave increasing energy consumption and latency that occur radios during periods of low demand. However, managing as the number of hops in the IAB network grows. In energy efficiency in the microwave network with multiple summary, while existing studies focus on the design of radios and multi-band configurations presents significant IAB-based FWA networks and radio resource allocation challenges. Specifically, it is difficult to determine optimal strategies, the challenge of minimizing energy consumption times and conditions for putting specific radios or carriers in multi-hop IAB-based FWA networks, particularly for into deep sleep or awakening them, ensuring that traffic rural deployments, remains largely unaddressed in the demands are met without network service disruption when literature. capacity needs suddenly increase. Based on the related works reviewed above, the inteFWA and Resource Allocation Supported by DT: In [22], gration of long-haul microwave supported by DT with the authors proposed a method to determine the optimal multi-hop IAB-based FWA for rural connectivity has not number of base stations required for FWA capacity and been thoroughly explored in the existing literature. To coverage planning, using an urban residential area as a the best of our knowledge, this is the first study to focus case study. Similarly, in [23], the authors introduced an on minimizing energy consumption in such networks by approach to estimate the maximum number of houses leveraging multi-radio microwave backhaul supported by that can be simultaneously connected to an FWA network DT to connect IAB-based FWA serving rural areas. while meeting minimum target bit rates, based on available network resources and cell radius. The work in [24] focused

Microwave Backhaul for IAB-based FWA

DT CPE

CPE

CN

MEC server

Radio 1

Radio 1

Radio M

Radio M CU DU

Fiber Sense

MT

Act

Microwave node

Sense

IAB node

CPE

CPE

Act IAB donor

Figure 2: Illustration of the system model. Table I: Summary of key notations. Notation

Definition

M J V D K Sj Aj j Pm,k

Set of microwave radios |M| = M Set of microwave nodes |J | = J Set of CPE and IAB-MT |V| = V DL data rate requirement Total number of states States of node j Actions of node j Power consumption for radio m in state k at node j Transmission power for radio m in state k Channel gain Deep sleep threshold Radio wake-up threshold: Time between sending update i to DT and receiving feedback State transition matrix The average AoP Time of freshest status update at the PT m Vector of state selection variables Vector of KPI/SLA satisfaction variables at PT Vector of KPI/SLA satisfaction variables at DT Vector of linearization variables Area of triangle Area of parallelogram Desynchronization time between DT and PT Energy consumption of DT Reward function in DT Achievable data rate for radio m of node j

TX Pm,k GRX m,6 D̃rds D̃w i Em

Φ(S j , Aj ) B̃m Um x y z o i2 Hm i1 Hm ∆τ Lmec R(S j , Aj ) j Dm

Figure 3: Illustration example of an antenna with two microwave radios [31]. ON (1)

OFF (0)

Completely off (2)

Startup (4)

Deep sleep (3)

Wakeup (5)

Serving (6)

Figure 4: Illustration of microwave radio states. CPEs and IAB MTs that use microwave backhaul. We denote dv as the downlink (DL) data rate requirement for terminal v, i.e., CPE or IAB MT. Hereafter, the term terminal denotes either a CPE or an IAB-MT. Since we have multiple terminals, the DL data rate requirement can be denoted as: V X D= dv . (1) v=1

III. System Model

The data rate D passes through the microwave node and We illustrate our system model in Fig. 2, while the key the IAB donor, each of which has multiple radios. notations used in this paper are summarized in Table I. We model the microwave backhaul connecting the IABOur system model considers the microwave backhaul based FWA network to the CN as a physical twin (PT). A between a microwave node and the IAB donor. The corresponding DT is deployed on a Multi-access Edge Commicrowave node is connected to the Core Network (CN) puting (MEC) server [32] in CN to represent the microwave via optical fiber. Each microwave node and IAB donor is backhaul. The DT continuously receives network metrics equipped with more radios. Fig. 3 shows an example of from the PT and optimizes the radios’ state transitions a multi-band booster antenna with two radios, which can to minimize energy consumption during periods of low be attached to the microwave node and the IAB donor. network utilization. The resulting optimized transition Let M denotes a set of microwave radios and J a set of matrix (i.e., action-state mapping) is then communicated microwave nodes and IAB donor, where M j represents the back to the PT as feedback. Upon receiving feedback from number of radios associated with microwave node or IAB the DT, the PT uses it to minimize both the AoP and donor j ∈ J . We consider multi-band microwave radios energy consumption while meeting the IAB-based FWA’s that can operate in narrow (e.g., 6 – 15 GHz), wide (e.g., data rate requirements. In IAB-based FWA, the IAB donor is equipped with 18 – 42 GHz), or very wide (e.g., 71 – 76 GHz and 81 – 86 GHz) frequency bands, depending on the data rate microwave radios for backhauling, as well as other radio(s) to serve IAB nodes and CPEs for home internet access. We requirements of IAB-based FWA serving rural area. In IAB-based FWA, we consider V to be the set of use the term IAB station to refer to either an IAB donor

or an IAB node. Each IAB station supports dual-band achievable SNR, the maximum achievable DL data rate for operation with mid-band and millimeter-wave (mmWave) microwave radio m of IAB donor j is given by: interfaces. Here, we remind that in IAB-based FWA, the j b (5) Dm = xjm,6 ωm log2 (1 + δm ) , DU has a MAC scheduler that allocates resource blocks b (RBs). Since each IAB station includes an O-DU, we use where ωm is the channel bandwidth. Considering all the terms DU and IAB station interchangeably unless microwave radios M j used at IAB donor j, the data rate otherwise specified. The proposed DT and multi-state should satisfy: Mj approach are applied exclusively to the microwave backhaul X j radios, while the access radios serving IAB nodes and CPEs Dm ≥ D, (6) m=1 are considered to operate continuously. Here, we focus on modeling microwave backhaul for connecting IAB-based At the time tjm , when the existing microwave radios in FWA, while the IAB-based FWA modeling, including self- the serving state can not satisfy (Eq. 6), the controller backhauling, is discussed in our previous works in [?], [33] considers switching the existing microwave radio(s) from the completely off state to the startup state. Therefore, we A. Microwave Backhaul for Connecting IAB-based FWA define an action a1,j m,4 of switching from completely off to As shown in Fig. 4, we consider two physical states for 0,j the microwave radio: OFF (denoted 0) or ON (denoted 1). the startup state. Otherwise, action ãm,2 is taken, where radio remains in a completely off state. By Also, the microwave node has a processor unit equipped the microwave 1,j taking action a , the controller sends a command to the m,4 with the controller. At the controller, we consider five microwave radio to switch physically from the OFF state states: completely off (denoted 2), deep sleep (denoted to the ON state. This action is denoted a1,j m . When the 3), startup (denoted 4), wakeup (denoted 5), and serving microwave radio starts, it stays in the startup state for state (denoted 6). When combining the physical states and controller states, we use K to denote the total number of states. The state {0{2, 3} and {1{4, 5, 6}} × {1, 2, . . . , M j } can be represented as:

τ X

xjm,4 tjm ≤ tjm,4 ,

(7)

tjm =1

where tjm,4 is the time required for the microwave radio to complete the startup process. In other words, microwave In the OFF(0) state, we distinguish between deep sleep radio takes action ã1,j to stay in the startup state until m,4 and being completely off. Completely off refers to a unit tjm,4 expires. If the startup process is not completed within that is either new or unused in a serving state for a long j period, such as consecutive days. In a deep sleep state, we startup time tm,4 , the microwave radio0,jreturns to the ˜ m,2 and tries to consider time on a small scale rather than on a daily scale. completely off state by taking action a∗ start again. Otherwise, the controller considers switching In the ON(1) state, the controller considers startup (4), the microwave radio from the startup state to the serving wake-up (5), and serving (6) states. 1,j j j state by taking action a . m,6 We denote Pm,k (tm ) as the power consumption of When network traffic falls below a certain threshold, microwave radio m of node j at the state k at time tjm . Also, we define xj = {xjm,k } as a vector of decision variables certain serving microwave radios can enter a deep sleep that indicate the state microwave radio belongs to. The state to reduce energy consumption. Therefore, we define the following data traffic D̃rds as the deep sleep threshold decision variable xjm,k is defined as: at the microwave node or at the IAB donor:   1 if microwave radio m of microwave node j Mj  X j j Ψrds = max{( Dm (tjm ) − D(tjm )), D̃rds }. (8) xm,k = is in the state k,   m=1 0, otherwise. (3) When Ψrds = D̃rds , we consider switching from the serving To satisfy the data rate requirement D that passes state to the deep-sleep state for microwave radio in the through the microwave backhaul, let us consider m and n as ON state. We denote this action as a0,j m,3 . By taking action two microwave radios. Radio n is located at the microwave a0,j , the controller sends a command to the microwave m,3 node, and radio m is located at the IAB donor. Then, we radio to switch physically to the OFF state. This action is express the achievable SNR δm at microwave radio m as denoted a0,j . Otherwise, microwave radio remains in the m follows: serving state by taking action ã1,j 2 TX m,6 . |GRX | P m,6 n,6 When backhaul traffic increases again, a microwave radio δm = , (4) 2 σmn in a deep sleep state can wake up. Therefore, we define the 2 TX where σmn is the noise power, Pn,6 is the transmission following data traffic D̃w as the radio wakeup threshold at the microwave node or at the IAB donor: power in serving state, and GRX m,6 is the channel gain Mj for microwave in ON and serving states. Based on the X j Ψw = max{( Dm (tjm ) − D(tjm )), D̃w }. (9) j S j = {Sm,k }|m = 1, 2, . . . , M j , k = 0, 1, 2, . . . , K.

(2)

m=1

Sense

Act Feedback

Think, Act, and Plan

Network/metrics data

Data Acquisition

Microwave Backhaul Digital Twin for IAB-based FWA

Microwave Backhaul Physical Twin for IABbased FWA

Network representation, SLA, and KPI definition

Nodes, links, network demand monitoring KPI and SLA against network performance Parameter tuning and self-healing

Maintenance plan

Figure 5: Interaction between PT and DT of microwave backhaul.

When Ψw = D̃w , we consider switching from the deep sleep state to the wakeup state. We denote this action as a1,j m,5 . Otherwise, microwave radio remains in a deep sleep j state by taking action ã0,j m,3 . We consider tm,5 as the time required so that the microwave radio completes the wakeup process, such that: τ X xjm,5 tjm ≤ tjm,5 . (10) t=1

When a microwave radio is in a startup state and cannot complete the wakeup process within the startup time tjm,5 , it fails, returns to the deep sleep state, and tries again. ˜ 0,j This action is denoted a∗ m,3 . Otherwise, the microwave radio switches from the wakeup state to the serving state. This action is denoted a1,j m,7 . When the microwave radio is in deep sleep for a long period T (in days), it can be switched to the completely off state. In other words, the completely off state happens when: T X j xjm,3 Dm (tjm ) = 0. (11) tjm =1

We denote a0,j m,2 as an action of switching from deep sleep to the completely off state. Otherwise, this action ã0,j m,3 is taken, where the microwave radio remains in the deep sleep state. Based on the above-defined actions, we define Aj = b,j {ab,j m,k+1 , am } as an action space for b = 0, 1 for physical states of microwave radio and k for states of microwave radios at the controller. Furthermore, we define Φ(S j , Aj ) as a transition matrix from one state to another, which depends on states S j and actions Aj . To reduce computational overload on controllers at each microwave node and IAB donor, we propose DT, which

represents the microwave backhaul connecting the IABbased FWA network to the CN. The DT is constructed using microwave backhaul data and deployed on the MEC server in the CN. After DT construction, state transitions S j and actions Aj are computed within the DT and then transmitted to the microwave node and IAB donor for acting. As illustrated in Fig. 5, the DT in the MEC server consists of two main modules: the Data Acquisition module and the Think, Act, and Plan module, each with several sub-modules: • Network Representation, Service Level Agreement (SLA), and Key Performance Indicator (KPI) Definition Sub-module: This sub-module collects information on microwave nodes and IAB donor, such as available microwave radios, and their configurations (e.g., frequency bands). This process is used to create or update the DT model of the microwave backhaul. • Nodes, Links, and Network Demand Monitoring Submodule: This sub-module gathers performance-related data, including data rate, delay, transmission power, energy consumption, microwave radio utilization, and radio states. This process, referred to as network sensing, ensures that the collected information is then passed to the Think, Act, and Plan module. • KPI and SLA Evaluation Sub-module: The monitored data are compared against predefined KPIs and SLAs (e.g., throughput requirements, latency constraints, and power consumption targets). If the performance meets the KPIs and SLAs, the DT instructs the Physical Twin (PT) to maintain the current microwave radio states. • Parameter and Self-Healing Sub-module: If network performance does not meet the KPIs and SLAs, the DT adjusts the microwave radio states accordingly to

Time of the freshest status update i at the PT node using microwave radio m is given by:

Microwave Backhaul DT

Microwave Backhaul PT i+1 Send update at 𝑈m

i i Um = max{Um , Em ≤ t; ∀i},

DT processing

𝑚 period Sense for 𝐶𝑖+1

where time t (with t = ̸ tjm ) is the current time. From (14), we can calculate instantaneous AoP Bm (t) of new status update i + 1 at time t as follows:

Feedback (updated microwave states)

Act at 𝑬im

Age of Processing

Bm (t) = t − Um .

Figure 6: AoP and interaction between PT and DT.

𝑖1 𝐻𝑚 𝑖2 𝐻𝑚

1 𝑈𝑚

0 𝐸𝑚

𝐿0𝑚

0 𝐶𝑚

1 𝐸𝑚

𝐿1𝑚

2 𝑈𝑚 1 𝐶𝑚

𝑖−1 𝐸𝑚

𝑖−1 𝑈𝑚

𝐿𝑖−1 𝑚

𝑖−1 𝐶𝑚

𝑖 𝐸𝑚

𝑖 𝑈𝑚

𝐿𝑖𝑚

𝑖+1 𝑈𝑚

time

𝑖 𝐶𝑚

Figure 7: AoP modeling for PT.

IV. Problem Formulation To reduce both computation and network delay in the microwave backhaul between DT and PT, we consider AoP. In other words, AoP combine computation delay with the total network delay (i.e., processing delay, queuing delay, transmission delay, and propagation delay). As shown in Figs. 6 and 7, at PT, we define the total time for sending update i (collected network information) to DT and receiving feedback (optimized states and actions) from i DT as Em . Also, we define Lim as the period between sending update i to DT and receiving DT’s feedback on i−1 the already submitted update at time Um . By considering computation at DT and network delay, the feedback on submitted update i is received at the microwave node or the IAB donor that uses microwave radio m at time: (12)

Furthermore, we assume that the PT continues to collect i backhaul network metrics. Therefore, from Em , the PT i generates a new update to send to DT after time Cm ≥ 0. i+1 The PT samples new status update i + 1 at time Um , i+1 where Um is given by: i+1 i i Um = Em + Cm .

We simplify (16) by decomposing the integral into a series of areas of parallelograms and triangles as shown in Fig. 7. i1 The area of parallelogram Hm can be expressed as follows: i1 i−1 i Hm = (Li−1 m + Cm )Lm ,

satisfy them. Once adjustments are applied, the DT validates whether the new states satisfy the KPIs and SLAs. If successful, the DT sends the adjusted states to the PT for implementation. • Maintenance Plan Sub-module: If the adjusted states still fail to meet the KPIs and SLAs, the DT activates the maintenance plan, which may involve scheduling a site visit for a technician to inspect the microwave backhaul and make proper setup and configuration to meet the KPIs and SLAs.

i i Em = Um + Lim .

(15)

In other words, Bm (t) is expressed as a time period. Then, we can use Bm (t) to compute the AoP Bm , where Bm is expressed as: Z 1 t Bm (t)dt. (16) Bm = lim t→∞ t 0

Age: 𝐵𝑚 (𝑡)

0 𝑈𝑚

(14)

(13)

(17)

i2 while the area of triangle Hm can be defined as follows:

1 i i 2 (L + Cm ) . (18) 2 m Therefore, by considering (17) and (18), the average AoP B̃m is given by: P i2 H i1 + Hm B̃m = Pm→∞ m . (19) i i m→∞ Lm + Cm i2 Hm =

Based on the average AoP B̃m , we formulate the following optimization problem that aims to minimize energy consumption while satisfying the IAB-based FWA’s performance constraints. Energy consumption is modeled j as the product of the power consumption Pm,k , which depends on the microwave states, and the AoP. Since power consumption is measured in watts and AoP in seconds, the resulting energy is expressed in joules: j

min x

K X M X

j xjm,k Φ(S j , Aj ) F Pm,k B̃m

(20)

k=0 m=1

subject to: j

K X M X

j

TX xjm,k Pm,k ≤

k=6 m=1

K X M X

TX P̃m,k ,

(20a)

k=6 m=1

j

K X M X

j Dm ≥ D,

(20b)

xjm,k ≥ 1,

(20c)

k=6 m=1 j

K X M X

k=6 m=1 xjm,k ∈ {0, 1}.

(20d)

The objective function above applies to each microwave node and IAB donor and depends on the number of states and microwave radios, since each microwave radio does not operate independently. Constraint (20a) ensures that the transmission power does not exceed the maximum allowable TX power limit P̃m,k for microwave radios in serving state.

Constraint (20b) guarantees that the microwave radios in serving state should support the required downlink data rate for IAB-based FWA. Constraint (20c) enforces that each microwave node and IAB donor has at least one radio in the serving state. Finally, Constraint (20d) ensures that xjm,k is binary decision variable. V. Solution Approach In (20), the transition matrix Φ(S j , Aj ) can, in principle, be obtained from historical state–action data of microwave radios. However, such data is not available in practice, making the transition matrix unknown. Moreover, estimating Φ(S j , Aj ) locally at each microwave node or IAB donor would incur significant computational overhead. To address this challenge, we leverage DT deployed at the MEC server in the CN to estimate the transition matrix. Specifically, the DT continuously collects real-time data from the PT and constructs a virtual representation of the network. Based on this virtual model, the DT learns and updates microwave state transition behaviors, i.e., Φ(S j , Aj ), and enables safe evaluation of state-control decisions before their deployment in the PT. Placing the DT at the MEC server further reduces the computational burden on individual microwave nodes and IAB donor, enabling efficient, adaptive, and SLA-compliant decisionmaking in a centralized manner. The PT continuously monitors the network, collects runtime measurements, and transmits them to the DT. The DT processes this information and evaluates Φ(S j , Aj ), which is then used to guide control decisions in the PT. The data acquisition framework includes three key components: (i) the backhaul network representation, which defines the virtual model of the PT, (ii) the service level agreement (SLA), which specifies the required data and service constraints for CPEs/MTs, and (iii) the key performance indicators (KPIs), which define measurable performance targets. Within the DT, the collected runtime metrics from the PT include radio states, achieved data rates, availability, power consumption, throughput, and resource utilization. These measurements are compared against predefined SLA and KPI requirements. Accordingly, we define a decision variable ym for each microwave radio m as: ( 1, if PT monitoring results satisfy KPI and SLA, ym = 0, otherwise. (21) If ym = 1, the DT confirms that the current configuration of microwave radio is satisfactory and instructs the PT to maintain the existing microwave radio states while continuing data collection. Otherwise (ym = 0), the DT explores alternative configurations, such as putting in serving state a microwave radio from deep sleep or turning on a completely off radio, in order to satisfy SLA and KPI requirements. These updated configurations are used to recompute Φ(S j , Aj ) within the DT environment before applying it to PT. The configurations are validated within

the DT, leading to a second decision variable zm defined as:  j j  1, if configuration at the DT using Φ(S , A ) zm = satisfies KPI and SLA requirements,   0, otherwise. (22) During the computation and testing of Φ(S j , Aj ) at the DT, a synchronization gap may arise between PT and the DT due to processing delays at the MEC server. We define the resulting desynchronization time ∆τ as: ∆τ =

D̃Λmec , fmec

(23)

where fmec denotes the CPU frequency of the MEC server, Λmec represents the number of CPU cycles required for DT processing, and D̃ is the size of the collected network/monitoring data. Furthermore, the energy consumption associated with DT processing at the MEC server is given by: 2 Lmec = D̃ ιmec Λmec fmec , (24) where ιmec is a constant determined by the MEC CPU architecture. Based on these definitions, we define reward function R(S j , Aj ) in DT as: PM j j j m=1 (ym + zm ) R(S , A ) = PK PM j . j j j k=0 m=1 ∥Φ(S , A )∥F Pm,k ∆τ + Lmec (25) Maximizing R(S j , Aj ) also helps to minimize the total energy cost in the denominator. Specifically, the denominator captures both (i) the communication-related energy consumption, which depends on the power consumption determined by microwave radio states and desynchronization time ∆τ , and (ii) the computational energy consumption Lmec at the MEC server. Moreover, even when no CPE or IAB-MT is active, microwave nodes and IAB donors still consume a non-zero amount of power due to control signaling operations. Therefore, the denominator of (25) never becomes zero, ensuring that the reward function remains well-defined. Finally, if R(S j , Aj ) = 0, then ym + zm = 0, meaning that neither the PT nor the DT can satisfy the SLA and KPI requirements. Therefore, a maintenance plan should be considered, which may require physical intervention. At DT, maximizing R(S j , Aj ) requires to estimate Φ(S j , Aj ), where we use a Reinforcement Learning (RL) approach [14] to approximate Φ(S j , Aj ). In particular, we use Deep Q-Learning (DQL) [34], which evaluates the quality of state-action pairs through the Q-function: Q : S j × Aj → R(S j , Aj ).

(26)

We chose DQL over other approaches because the transition matrix for the microwave node and IAB donor is unknown, making classical optimization intractable. As a modelfree method, DQL learns optimal state–action policies directly from interactions between the DT and PT without requiring an explicit transition model.

Feedback (transition matrix, i.e., actionsDT: maximizes reward = minimize energy consumption = Power (microwave status) × de- states mapping ) synchronize time (DT processing time) subject to service requirements of IAB-based FWA

Solution: Reinforcement Learning (RL) approach to calculates the quality of the state-actions combination

PT: minimizes energy consumption = Power (depends on microwave status from DT) × age of information subject to service requirements of IAB-based FWA

Solution: Optimization solver

Generating new data and send it to DT ( every 1 second after receiving feedback) Figure 8: Summary of the proposed solution approach for (26) and (20). As illustrated in Fig. 8, the transition matrix Φ(S j , Aj ) validation indicators, respectively, each having dimension is obtained from DQL in maximizing R(S j , Aj ) and n. In the worst case, the DT must compare all possible transmitted as feedback to the PT. This feedback enables state-validation combinations, resulting in a computational each microwave node and IAB donor to solve (20) with complexity of O(n2 ). In the PT of the microwave backhaul, the computational known Φ(S j , Aj ). In other words, the problem (20) can be solved using standard solvers such as Gurobi or CPLEX. complexity mainly depends on the number of microwave After solving (20), each microwave node and IAB donor radios and operational states. Let n denote the total collect new network metrics and transmit them to the DT number of microwave radio-state combinations. Solving i as a status update. In the proposed solution, we use Cm = 1 (20) requires iterating through the radio-state decision j second; however, this choice is not restrictive, and other variable xm,k , resulting in a linear complexity of O(n). The total computational complexity of our proposal can update intervals can be applied. Nonetheless, selecting a  P  i 2 larger value of Cm may prevent the DT from accurately be expressed as O µ L n n + O(n ) + O(n). Since l−1 l l=1 reflecting the PT’s real-time status. the DNN training is performed offline, the quadratic verifiRemark 1. Computational complexity of the proposed cation process dominates the online execution. Therefore, the overall worst-case computational complexity of the solution approach is O(n2 ). proposed solution can be approximated as O(n2 ). In the DT, DQL employs a deep neural network (DNN) VI. Simulation Results and Analysis [35], [36] to approximate the Q-value function Q in (26). In this section, we describe the simulation setup and The primary computational overhead originates from the present the results of the performance evaluation. The DNN training phase, where each parameter update requires simulation environment is implemented using Python 3.8.10 both forward and backward propagation through the netfor numerical analysis and CVXPY [37] with embedded work. The corresponding computational complexity can be  P  L conic solver (ECOS) for solving optimization problem in expressed as O µ l=1 nl−1 nl , µ denotes the minibatch PT. Furthermore, for DQL in DT, we use PyTorch [38]. size, L is the number of neural network layers, and nl represents the number of neurons in layer l. Therefore, A. Simulation Setup the complexity increases with the DNN depth and the To construct the network topology, we randomly selected size of the state-action space. However, in this work, a rural area in Quebec, Canada, as illustrated in Fig. 9. the DNN is trained offline at the DT hosted on the The microwave node (i.e., the red node) is located near the MEC server in the core network (CN). Consequently, the town of Val-des-Sources. It is connected to an IAB donor online DT operation performs inference only using the (node in yellow) located 17.95 km away, providing service pre-trained model, thereby significantly reducing real-time coverage for 35 houses (i.e., CPEs are the nodes in green) computational overhead. Moreover, since rural microwave in the surrounding rural area. backhaul networks experience relatively slow variations For the microwave node and the IAB donor, each is in traffic demand and microwave radio states compared equipped with two radios: one operating in the 7 GHz with dense urban deployments, the DT computational band with 64 MHz bandwidth, and another in the 42 GHz requirements remain manageable. band with 500 MHz bandwidth. The power states of the To evaluate whether the estimated transition matrix microwave radios are modeled as follows: 0 W (completely Φ(S j , Aj ) satisfies the SLA and KPI requirements, the DT off), 3 W (deep sleep), 55 W (startup), 50 W (wakeperforms verification operations over the microwave radio up), and 80 W (serving). In practice, however, these states and monitoring results. Let y and z denote vectors power levels can be directly measured from the microwave associated with microwave radio states and SLA/KPI equipment. Furthermore, we set tj = 10 seconds and m,4

1e9

4

Data rate (bps)

3 Data rate: 7 GHz and 42 GHz Minimum data rate for IAB-based FWA Data rate: only 7 GHz Data rate: only 42 GHz

2 1

-08

-07

26 -01 20

26 -01

-06 20

26 -01

-05 20

26 -01

-04 20

26 -01

-03 20

-02

26 -01 20

26 -01 20

20

26 -01

-01

0

Date Figure 11: Data rate requirements with radio states. Radios of 42 and 7 GHz in serving state Our data rate (5 states with DT) Data rate requirement

1e9

4.0

Data rate (Bps)

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0

Microwave backhaul traffic (Mbps)

Figure 9: Nodes in rural area under consideration.

4000 3500 3000 2500 2000 1500 1000 500

100

200

300 400 Iterations

500

600

700

Figure 12: Our approach in satisfying data rate requirements.

-08

26

20

-01

-07

26

20

-01

-06

26

20

-01

-05

26

20

-01

-04

26

20

-01

-03

26

20

-01

-02

26

20

-01

26

20

-01

-01

In IAB-based FWA, 7 IAB nodes (i.e., nodes in blue in Fig. 9) are connected to the IAB donor to serve 35 houses with CPEs. As the signal attenuates with distance, we assume each CPE is connected to the nearby IAB station. Nearby CPEs or IAB-MTs to the parent IAB station operated in the mmWave band, while distant ones used the mid-band. The IAB-based FWA network operated at 38 GHz for mmWave and 6 GHz for mid-band, with bandwidths within the range of 10 to 1600 M Hz and subcarrier spacing in the range 15 to 480 kHz [40]. The results of the use of mmWave and mid-band for IAB-based Date FWA are discussed in our previous work in [33], [41]. In this Figure 10: Downlink microwave backhaul data for performance evaluation, we focus on the use of microwave IAB-based FWA. backhaul to support IAB-based FWA. The network traffic in the considered rural scenario is synthetically generated over a one-week period with a tjm,5 = 8 seconds. These values are reasonable for microwave temporal resolution of one second. The generated traffic backhaul radios, which typically maintain stable operation model captures both short-term variations and recurring by aggregating traffic from multiple access radios, thereby daily patterns, providing a realistic and flexible repreenabling continuous traffic availability. As a result, frequent sentation of aggregated data demand for performance state transitions at a second or millisecond scale during the evaluation in rural environments. As expected in rural day are uncommon, an assumption supported by [12], [39]. access networks, traffic demand is higher during daytime However, these parameter values are not suitable for IAB- periods and significantly lower during nighttime hours. The based FWA radios that operate under different resource generated traffic profile is then used as an input to the allocations. proposed approach, which is designed to support arbitrary

40

1e9

4.0 3.0

Reward

Data rate (Bps)

3.5 Data rate all radio serving Our data rate (5 states with DT) Our data rate (5 states no DT) Data rate requirement Data rate (3 states)

2.5 2.0 1.5

20 10 0 10

1.0

0

300 400 500 Episode Figure 15: Reward maximization at DT.

0.5 0

100

200

300 400 Iterations

500

600

700

Figure 13: Comparison of our approach with other methods.

140 Energy consumption all radio serving Our energy consumption (5 states with DT, 47.34 % improvement ) Our energy consumption (5 states no DT, 46.91 % improvement ) Energy consumption (3 states), 45.34 % improvement )

130 120 110

Seconds

150

100

200

600

Optimization computation delay DQL computation delay AoP Network delay

20

160 Energy consumption (Joule)

DQL reward

30

15 10 5

100 90 80

0 0

100

200

300 400 Iterations

500

600

700

100

200

300 400 Iterations

500

600

Figure 16: Age of processing and delay.

Figure 14: Comparison of energy consumption. with and without DT support, both satisfy minimum data rate requirements. and dynamic traffic patterns. The resulting backhaul traffic Fig. 14 compares the energy consumption of the proposed demand for the IAB-based FWA scenario is presented in approach with and without DT support. The results are Fig. 10. further benchmarked against a three-state power model and a baseline scenario in which all radios remain continuously B. Simulation Results serving. The simulation results show that the proposed Considering the five operational states of each microwave approach achieves lower energy consumption compared radio, where each microwave node and IAB donor are to the baselines. In this figure, energy is expressed in j equipped with two microwave radios (M = 2) like joules; note that 1 W corresponds to 1 J/s. The results the ones shown in Fig. 3. The DT decides on states to show an improvement of approximately 47.34% over all minimize energy consumption while satisfying the required radios that remain continuously serving and 2% over the microwave backhaul data rate. As illustrated in Figs 11 3-state baseline. This gain can be even more significant and 12, when the data rate requirement is relatively low, when scaled to a network with multiple microwave links, a 42 GHz microwave radio with 500 M Hz bandwidth resulting in substantial cumulative energy savings for is sufficient to meet the backhaul data rate requirement. network operators. In this case, the 7 GHz microwave radio with 64 M Hz In the DT, we employ DQL to maximize the reward forbandwidth can enter a deep-sleep state to reduce energy mulated in (26). Fig. 15 illustrates the reward maximization consumption. Conversely, when the demand exceeds the process over 700 episodes. In the context of DQL, an episode capacity of the 42 GHz radio, the 7 GHz radio must wake is defined as a complete sequence of interactions between up to support the 42 GHz microwave radio and ensure that the agent and the environment, starting from an initial the data rate requirements are met. When both the 42 state and continuing until a terminal state is reached. The GHz and 7 GHz microwave radios are in serving states, a DQL model is implemented as a fully connected feedforward rate of 4.2 Gbps can be achieved for microwave backhaul. neural network. Its architecture comprises two hidden layers In Fig. 13, we compare our five-state approach with a with 64 neurons each, with ReLU activation functions to baseline in [12]. Furthermore, we evaluate the performance introduce nonlinearity. The final linear layer projects the of our approach with and without DT support [?]. The relearned hidden representation onto the action-value space, sults demonstrate that our proposal outperforms the threethereby estimating the Q-values for all possible actions Aj state baseline in satisfying the data rate requirement of in a given states S j . This architecture achieves a balance IAB-based FWA. However, when considering our approach

1e6

Energy efficiency (Bit/Joule)

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0

100

200

300 Iteration

5 states with AoP and DT All radios are serving 5 states without AoP and DT 400 500 600

Average data rate (Mbps)

Figure 17: Comparison of energy efficiency.

200 180 160 140 120 100 80 60

VII. Conclusion

0

CPE: Achievable data rate (baseline) CPE: Achievable data rate (our dynamic configuration) CPE: Minimum data rate requirement 20000 40000 60000 80000 100000 120000 Iterations

Figure 18: Average achievable data rate per CPE. 60 Transmission power (Our dynamic configuration) Transmission power (baseline) 50

Average transmission power (W)

and energy consumption, the results in Fig. 17 demonstrate that our proposed approach with DT and AoP achieves higher energy efficiency. We consider an IAB-based FWA network using 38 GHz for the mmWave band and 6 GHz for the mid-band. A numerology-based resource block allocation scheme from our previous work in [?] is employed to select channel bandwidths ranging from 10 M Hz to 1600 M Hz and subcarrier spacings between 15 kHz and 480 kHz [40]. The proposed dynamic configuration approach is compared with a baseline scenario that uses fixed numerologies and radio resources, namely 264 resource blocks for the midband and 273 resource blocks for the mmWave band, both with a 30 kHz subcarrier spacing [5]. Figure 18 illustrates the achievable data rate, while Fig. 19 presents the corresponding transmission power. The results show that the proposed dynamic configuration achieves the same data rate as the baseline while requiring lower transmission power.

40 30 20 10 0 0

20000 40000 60000 80000 100000 120000 Iterations

Figure 19: Average transmission power for IAB station.

This paper investigated an energy-efficient multi-hop wireless backhaul for rural connectivity by considering long-haul microwave links, 5G IAB, and FWA as a unified solution. To address the overlooked challenge of increasing energy consumption with the number of hops, we modeled the microwave backhaul as a multi-state system, incorporating radio-off, start-up, serving, deep sleep, and wake-up states, and formulated an energy minimization problem that satisfies the data rate requirement of IABbased FWA serving rural areas. To efficiently solve this problem, we leveraged deep Q-learning within the DT to learn optimal control policies. In contrast, we solve the optimization problem in PT for changing microwave radio states to minimize energy consumption while satisfying data rate requirements. Simulation results demonstrate that the proposed framework achieves substantial energy savings compared to conventional baseline strategies while still satisfying the required data rate constraints for IABbased FWA. These findings highlight the effectiveness of DT-enabled intelligence for managing energy-performance trade-offs in multi-hop rural wireless networks and provide a promising direction for sustainable, scalable rural broadband deployment. References

between simplicity and computation requirement, making it well-suited for reinforcement learning tasks with moderate state and action dimensions. Considering DQL within the DT and AoP, Fig. 16 illustrates both the network delay between the PT and DT and the processing time at the DT (DQL computation delay). The figure also shows the computation delay in solving the optimization problem. Notably, distributing DQL to each microwave node and IAB donor without leveraging the DT may consume more energy than executing DQL operations centrally at the MEC server. Considering both data rate

[1] ITU, “Global number of internet users increases, but disparities deepen key digital divides,” https: //www.itu.int/en/mediacentre/Pages/PR-2025-11-17-Factsand-Figures.aspx#:~:text=Geneva%2C%2017%20November% 202025,as%20artificial%20intelligence%20(AI)., 2025, published: November 17, 2025, Accessed April 11, 2026. [2] Maravedis LLC, “5G fixed wireless gigabit services today: An industry overview,” https://shop.maravedis-bwa.com/products/ 5g-fixed-wireless-gigabit-services-today-an-industry-overview, [Online; accessed April 1, 2026]. [3] Ericsson, “Ericsson mobility report,” https://www.ericsson. com/4adb7e/assets/local/reports-papers/mobility-report/ documents/2024/ericsson-mobility-report-november-2024.pdf, November, 2024. [4] K. R. Chaudhuri, E. C. Neto, L. Falconetti, R. Fassbinder, S. Guirguis, A. Halder, M. Irizarry, R. D. Patel, N. Saxena, and

S. Sorlescu, “Extended range mmwave for fixed wireless applications,” in Proceedings of 97th ARFTG Microwave Measurement Conference (ARFTG). IEEE, 2021, pp. 1–4. [5] M. Hashemi, M. Coldrey, M. Johansson, and S. Petersson, “Integrated access and backhaul in fixed wireless access systems,” in 2017 IEEE 86th Vehicular Technology Conference (VTC-Fall). IEEE, 2017, pp. 1–5. [6] 3GPP-TR, “Technical specification group radio access network; nr; study on integrated access and backhaul (release 16). (3GPP TR tr 38.874 v16.0.0 ),” 2018-12. [7] G. Zu, M. Nadim, S. Reddy, T. U. Islam, S. Babu, T. Zhang, D. Oiao, H. Zhang, and A. Arora, “Arahaul: Multi-modal wireless x-haul living lab for long-distance, high-capacity communications,” in 2023 IEEE Future Networks World Forum (FNWF). IEEE, 2023, pp. 1–6. [8] F. Tao, B. Xiao, Q. Qi, J. Cheng, and P. Ji, “Digital twin modeling,” Journal of Manufacturing Systems, vol. 64, pp. 372– 389, 2022. [9] B. Zhang, F. Devoti, I. Filippini, and D. De Donno, “Resource allocation in mmwave 5G IAB networks: A reinforcement learning approach based on column generation,” Computer Networks, vol. 196, p. 108248, 2021. [10] M. Yu, Y. Pi, A. Tang, and X. Wang, “Coordinated parallel resource allocation for integrated access and backhaul networks,” Computer Networks, vol. 222, p. 109533, 2023. [11] V. Begishev, E. Sopin, D. Moltchanov, R. Pirmagomedov, A. Samuylov, S. Andreev, Y. Koucheryavy, and K. Samouylov, “Performance analysis of multi-band microwave and millimeterwave operation in 5g nr systems,” IEEE Transactions on Wireless Communications, vol. 20, no. 6, pp. 3475–3490, 2021. [12] A. Frithiofson and M. Vacante, “Energy efficient control of microwave networks,” 2022. [13] A. Ndikumana, K. K. Nguyen, and M. Cheriet, “Renewable energy powered and open ran-based architecture for 5g fixed wireless access provisioning in rural areas,” IEEE Transactions on Green Communications and Networking, 2024. [14] D. Ernst and A. Louette, “Introduction to reinforcement learning,” Feuerriegel, S., Hartmann, J., Janiesch, C., and Zschech, P, pp. 111–126, 2024. [15] R. Li, Q. Ma, J. Gong, Z. Zhou, and X. Chen, “Age of processing: Age-driven status sampling and processing offloading for edgecomputing-enabled real-time iot applications,” IEEE Internet of Things Journal, vol. 8, no. 19, pp. 14 471–14 484, 2021. [16] R. D. Yates, Y. Sun, D. R. Brown, S. K. Kaul, E. Modiano, and S. Ulukus, “Age of information: An introduction and survey,” IEEE Journal on Selected Areas in Communications, vol. 39, no. 5, pp. 1183–1210, 2021. [17] A. Ndikumana, K. K. Nguyen, and M. Cheriet, “Age of processing-based data offloading for autonomous vehicles in multirats open ran,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 11, pp. 21 450–21 464, 2022. [18] N. Anselme, K. K. Nguyen, and M. Cheriet, “Age of processingaware offloading decision for autonomous vehicles in 5g open ran environment,” IEEE Transactions on Mobile Computing, vol. 23, no. 7, pp. 7865–7877, 2023. [19] M. G. Sellin, M. Edberg, M. Berggren, M. Coldrey, J. Edstam, D. Eriksson, J. Flodin, J. Hansryd, A. Olsson, M. Ohberg et al., “Enhancing 5g with microwave,” Ericsson, Oct, 2020. [20] Y. Wang, H. Wu, Y. Niu, J. Ding, S. Mao, B. Ai, Z. Zhong, and N. Wang, “Triple-band scheduling with millimeter wave and terahertz bands for wireless backhaul,” Journal of Communications and Networks, vol. 24, no. 4, pp. 438–450, 2022. [21] A. Colzani, M. Fumagalli, A. Fonte, A. Traversa, and E. Ture, “Long-reach e-band hpa for 5g radio link,” in Proceedings of 52nd European Microwave Conference (EuMC). IEEE, 2022, pp. 760–763. [22] P. Rahmawati, M. I. Nashiruddin, A. T. Hanuranto, and A. Akhmad, “Assessing 3.5 ghz frequency for 5g new radio (nr) implementation in indonesia’s urban area,” in Proceedings of 12th Annual Computing and Communication Workshop and Conference (CCWC). IEEE, 2022, pp. 0876–0882. [23] A. Lappalainen, Y. Zhang, and C. Rosenberg, “Planning 5g networks for rural fixed wireless access,” IEEE Transactions on Network and Service Management, vol. 20, no. 1, pp. 441–455, 2022.

[24] B. De Beelde, Z. Verboven, E. Tanghe, D. Plets, and W. Joseph, “Outdoor mmwave channel modeling for fixed wireless access at 60 ghz,” Radio Science, vol. 57, no. 12, pp. 1–14, 2022. [25] B. De Beelde, M. Vantorre, G. Castellanos, M. Pickavet, and W. Joseph, “Mmwave physical layer network modeling and planning for fixed wireless access applications,” Sensors, vol. 23, no. 4, p. 2280, 2023. [26] G. Castellanos, B. De Beelde, D. Plets, L. Martens, W. Joseph, and M. Deruyck, “Evaluating 60 ghz fwa deployments for urban and rural environments in belgium,” Sensors, vol. 23, no. 3, p. 1056, 2023. [27] A. Ndikumana, K. K. Nguyen, and M. Cheriet, “Digital twin backed closed-loops for energy-aware and open ran-based fixed wireless access serving rural areas,” IEEE Transactions on Mobile Computing, 2024. [28] N. Anselme, K. K. Nguyen, and M. Cheriet, “Digital twin assisted closed-loops for energy-efficient open ran-based fixed wireless access provisioning in rural areas,” in GLOBECOM 2023-2023 IEEE Global Communications Conference. IEEE, 2023, pp. 6285–6290. [29] N. Tafintsev, D. Moltchanov, A. Chiumento, M. Valkama, and S. Andreev, “Airborne integrated access and backhaul systems: learning-aided modeling and optimization,” IEEE Transactions on Vehicular Technology, 2023. [30] G. Yao, M. Hashemi, R. Singh, and N. B. Shroff, “Delayoptimal scheduling for integrated mmwave–sub-6 ghz systems with markovian blockage model,” IEEE Transactions on Mobile Computing, vol. 22, no. 9, pp. 5124–5139, 2022. [31] Ericsson, “Long haul microwave solutions for high capacities over long distances,” https://www.ericsson.com/en/mobiletransport/long-haul, [Online; accessed April 15, 2026]. [32] A. Ndikumana, N. H. Tran, T. M. Ho, Z. Han, W. Saad, D. Niyato, and C. S. Hong, “Joint communication, computation, caching, and control in big data multi-access edge computing,” IEEE Transactions on mobile Computing, vol. 19, no. 6, pp. 1359–1374, 2019. [33] A. Ndikumana, K. K. Nguyen, and M. Cheriet, “Energy-efficient 5g integrated access and backhaul open ran-based fixed wireless access provisioning in rural areas,” IEEE Transactions on Green Communications and Networking, vol. 10, pp. 1642–1653, 2025. [34] J. Clifton and E. Laber, “Q-learning: Theory and applications,” Annual Review of Statistics and Its Application, vol. 7, no. 1, pp. 279–301, 2020. [35] V. Mnih, K. Kavukcuoglu, D. Silver et al., “Human-level control through deep reinforcement learning,” Nature, vol. 518, no. 7540, pp. 529–533, 2015. [36] H. Van Hasselt, A. Guez, and D. Silver, “Deep reinforcement learning with double q-learning,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 30, no. 1, 2016. [37] S. Diamond and S. Boyd, “Cvxpy: A python-embedded modeling language for convex optimization,” Journal of Machine Learning Research, vol. 17, no. 83, pp. 1–5, 2016. [38] N. Ketkar and J. Moolayil, “Introduction to pytorch,” in Deep learning with python: learn best practices of deep learning models with PyTorch. Springer, 2021, pp. 27–91. [39] Ericsson, “How to reduce power consumption in microwave networks,” https://www.ericsson.com/en/reports-andpapers/microwave-outlook/articles/reducing-powerconsumption-in-microwave-networks, [Online; accessed April 29, 2026]. [40] T. ETSI, “User equipment (UE) radio transmission and reception; part 2: Range 2 standalone (3GPP TS 38.101-2 version 17.6.0 release 17),” 2022-08. [41] A. Ndikumana, K. K. Nguyen, A. Larabi, and M. Cheriet, “Energy-efficient multi-radio microwave and iab-based fixed wireless access for rural areas,” IEEE Transactions on Green Communications and Networking, vol. 10, pp. 3541–3554, 2026.