This paper has been accepted for publication at IEEE Transactions on Communications, 2026. Please cite it as: A. Traspadini, M. Pagin, R. Ihamouine, R. Lucas, A. Noren, M. Zorzi, and M. Giordani, "End-to-End Simulation of 5G NR Integrated 1 Access and Backhaul Networks for Remote Maritime Connectivity," IEEE Transactions on Communications, to appear, 2026.
End-to-End Simulation of 5G NR Integrated Access and Backhaul Networks for Remote Maritime Connectivity
arXiv:2605.16531v1 [cs.NI] 15 May 2026
Alessandro Traspadini, Matteo Pagin, Raphaël Ihamouine, Rupert Lucas, Andrew Noren, Michele Zorzi, Fellow, IEEE, Marco Giordani, Senior Member, IEEE Abstract—Millimeter wave (mmWave) 5th generation (5G) networks offer high data rates but face coverage challenges due to severe path loss and blockage. These problems motivate the use of Integrated Access and Backhaul (IAB) as a flexible wireless backhaul solution that extends connectivity to cell boundaries and unfibered areas, including maritime environments. This paper overviews the latest 3GPP specifications for IAB networks in Releases 16 through 18. Then, it presents an ns-3 module for IAB, featuring a complete end-to-end protocol stack, including the backhaul adaptation protocol (BAP) layer, flexible slot and control configurations, and multiplexing schemes based on both time and frequency division. We test the IAB module via extensive system-level simulations in a custom maritime scenario where vessels, equipped with IAB-nodes, can simultaneously act as access points and relays, forming dynamic multi-hop networks that maintain connectivity via wireless backhaul to shore-based stations. We evaluate different topologies and channel conditions, providing insights into the design and deployment of mmWave IAB networks in offshore environments. Index Terms—Integrated Access and Backhaul (IAB), 5-th generation (5G), millimeter wave (mmWave) communication, maritime networks.
I. I NTRODUCTION Current 5th generation (5G) New Radio (NR) commercial networks are designed to operate in both Frequency Range 1 (FR1) below 6 GHz and Frequency Range 2 (FR2) in the millimeter wave (mmWave) spectrum [1]. The mmWave bands, in particular, enable gigabit-level data rates for enhanced Mobile Broadband (eMBB) services, but face significant propagation challenges due to severe path loss, blockage, and diffuse scattering. To ensure ubiquitous and continuous coverage, networks should be deployed as dense small cells, to reduce inter-site distance and establish stronger access channels [2]. This approach, however, involves high capital and operational expenditures (capex and opex) for network operators, primarily due to the high costs and logistical complexity associated with the installation of fiber-based backhaul. To address these challenges, the 3rd Generation Partnership Project (3GPP) standardized Integrated Access and Backhaul A. Traspadini, M. Zorzi, and M. Giordani are with the Department of Information Engineering, University of Padova. Padova, Italy. (E-mail: {alessandro.traspadini, marco.giordani, michele.zorzi}@dei.unipd.it). M. Pagin was with the Department of Information Engineering, University of Padova. Padova, Italy. He is now with Keysight Denmark. R. Ihamouine, R. Lucas, and A. Noren are with Viasat Inc, UK. (Email: {Raphael.Ihamouine, Rupert.Lucas, Andrew.Noren}@viasat.com). This work was partially supported by the European Union under the Italian National Recovery and Resilience Plan (NRRP) Mission 4, Component 2, Investment 1.3, CUP C93C22005250001, partnership on “Telecommunications of the Future” (PE00000001 - program “RESTART”).
(IAB) in Release 16 [3], as a cost-efficient solution to support dense 5G deployments without requiring fiber connectivity at every site. In IAB, only a fraction of Next Generation Node Bases (gNBs), called IAB-donors, are directly connected to the 5G Core (5GC) via traditional fiber links. The others, called IAB-nodes, simultaneously operate as access points for User Equipments (UEs) and as wireless relays that forward the backhaul traffic toward the IAB-donor, possibly via multiple hops and at mmWave frequencies to maximize capacity [4], [5]. IAB provides a flexible and easy-to-deploy solution to improve coverage and accessibility, making it an attractive option for mmWave 5G networks [6], [7]. In fact, several operators have already shown interest in deploying IAB, with current estimates suggesting that approximately 10–20% of 5G sites could adopt this technology [1]. IAB is particularly beneficial in those scenarios where Line of Sight (LOS) propagation is blocked, such as in urban environments, or where fiber backhaul is impractical, such as in rural and remote areas [6], [7], as well as for Intelligent Transportation Systems (ITS), including railway and vehicular networks [8]–[11]. Beyond terrestrial applications, IAB can also play a crucial role to extend coverage in remote maritime coastal areas and seaports. In this context, IAB-nodes can be installed on vessels to form a dynamic multi-hop access and backhaul network interconnecting ships and shore stations with no or limited fiber capabilities [12], [13]. When sidelink communication is enabled, IAB can also support direct shipto-ship connectivity for exchanging critical information, such as navigation data and safety alerts, even beyond cellular coverage. Recent works have highlighted the role of sidelink IAB to enable autonomous maritime operations, e.g., to coordinate maneuvers among fleets of ships with low latency [14]. To properly design, optimize, and dimension IAB networks, accurate and realistic end-to-end evaluation is required. In this context, ns-3 enables full-stack simulation of complex network scenarios, and therefore represents a valid research tool in this domain. In 2018, we released an open-source 5G IAB simulation module for ns-3 [15]. However, that module was based on early 3GPP specifications and assumptions. Since then, the standardization of IAB has progressed significantly, and several key features have been introduced from Release 16. New features include: (i) a new backhaul adaptation protocol (BAP) layer, located above the Radio Link Control (RLC) layer, to support routing and forwarding of the packets across the IAB topology, i.e., from the IAB-donor to the access IAB-node; (ii) the support for Dual Connectivity (DC), which permits an IAB-node to concurrently connect to multiple IAB-
2
donors; (iii) new multiplexing and scheduling mechanisms; (iv) improved topology adaptation, interference mitigation, and mobility management schemes to support mobile Integrated Access and Backhaul (mIAB) nodes and UEs in dynamic environments. These innovations necessitate substantial updates of the existing ns-3 module to reflect current IAB research trends. Based on the above introduction, in this paper we provide the following contributions: • We present a comprehensive overview of the main 3GPP 5G NR IAB specifications from Releases 16 to 18. • We design and implement an open-source system-level simulation framework for 5G NR IAB networks based on ns-3, called ns3-mmwave-iab, which incorporates realistic physical layer modeling. Compared to the earlier implementation of this module presented in [15], the proposed framework is now aligned with the latest 3GPP IAB specifications. The framework captures realistic slot structures, multiplexing schemes (supporting both Time Division Multiplexing (TDM) and Frequency Division Multiplexing (FDM)), configurable control overhead, and multi-hop backhaul constraints. • Most of the literature on IAB is based on terrestrial networks, while the maritime scenario remains largely unexplored. However, emerging trends in commercial, military, security, safety-critical, navigation, and offshore monitoring operations rely on continuous and high-capacity connectivity, and highlight the strategic role of IAB in maritime scenarios. To address this gap, our ns3-mmwave-iab framework also integrates a maritime-specific mmWave propagation and channel model, to evaluate the effects of sea-surface reflections and offshore propagation characteristics. • Leveraging this framework, we conduct an extensive simulation campaign to evaluate the performance of maritime IAB networks under different topologies, weather, and resource allocation configurations. We demonstrate that single-hop deployments offer low latency in good propagation conditions, but suffer from severe rain attenuation. In turn, multi-hop networks suffer more interference, though this impact can be mitigated by heavy rain. Resource allocation is also crucial: in a multi-hop IAB network, we discuss how to distribute Orthogonal Frequency Division Multiplexing (OFDM) symbols to avoid bottleneck effects. The remainder of this paper is organized as follows. Sec. II reviews the related work on IAB. Sec. III outlines the 3GPP IAB standard from Release 16 to Release 18. Sec. IV describes the proposed IAB ns-3 simulator, focusing on the implemented protocol stack, maritime channel model, slot formats, control symbols, and multiplexing. Sec. V presents our main simulation results. Finally, Sec. VI concludes the paper with directions for future research. II. R ELATED W ORKS The existing literature on IAB has investigated key research aspects such as resource allocation, topology optimization, and mobility support, under both static and dynamic conditions, which are reviewed in the following paragraphs.
a) Resource allocation: IAB-nodes can operate in both in-band and out-of-band scenarios. The out-of-band mode involves using different bands for access and backhaul, while in the in-band mode the backhaul and access links share a common band, which provides more flexibility at the cost of increased complexity [8]. Resource allocation has primarily focused on in-band backhauling networks [1], [16]–[22], while fewer works, e.g., [7], [23], have addressed out-of-band backhauling due to the higher complexity. In [1], the authors proposed a joint resource allocation and link scheduling framework for multi-hop mmWave IAB systems, considering half-duplex constraints and flexible Transmission Time Interval (TTI) configurations, showing that multiple beams at the IAB-donor can significantly improve the throughput. Similarly, [16] investigated capacity maximization in a multi-hop in-band IAB network using approximation techniques to solve the underlying optimization problem. Centralized and semi-centralized resource allocation strategies were explored in [17], [18], showing that centralized decisions can improve the performance but are sensitive to outdated channel quality information, e.g., channel state information (CSI), obtained from child nodes. To address this limitation, semi-centralized resource allocation allows local schedulers to refine centralized decisions based on local information, provided that timely information exchange is feasible, thereby improving the network performance. Similarly, in [20] the authors proposed an integrated resource allocation scheme, where the bandwidth assigned by a donor IAB to the backhaul for a given IAB-node depends on the access load of the node itself. This scheme can be interpreted as a centralized allocation framework with some adaptation to local conditions, and provides superior coverage probability compared to static resource allocation. Alternative approaches include game-theoretic formulations [19], which optimize resource allocation via centralized and distributed algorithms converging to the Nash bargaining solution, and methods based on deep reinforcement learning (DRL) [21], which adaptively allocate spectrum in dynamic IAB networks while handling time-varying conditions and heterogeneous setups. An experimental approach was presented in [22], where the authors implemented a testbed consisting of an IAB-donor, an IAB-node, and a user operating in the mmWave band, and evaluated dynamic resource allocation between access and backhaul links in terms of coverage and throughput. In the out-of-band context, the authors in [7] proposed an Open Radio Access Network (O-RAN) optimization framework for out-of-band IAB, formulating a Mixed-Integer Linear Programming (MILP) to jointly optimize user-level radio resources, bandwidth partitioning, and mmWave backhaul routing. Similarly, out-of-band mmWave IAB networks were studied in [23] with both centralized and distributed resource allocation algorithms, complemented by a geometric analysis to approximate co-channel interference and provide system design insights. b) Topology optimization: Topology design and routing strategies are fundamental to achieve efficient multi-hop communication. Several works have addressed topology optimization using centralized and distributed approaches [24]–[27].
3
Fig. 1: 3GPP IAB protocol stacks for UP and CP. In both cases, packets are routed from the IAB-Donor to IAB-Node 2 through the wireless backhaul of IAB-Node 1.
In [24], the authors proposed a distributed stochastic optimization framework for joint resource allocation and path selection. Similarly, [25] investigated joint routing and resource allocation to maximize the minimum node throughput under time and resource constraints in a Time Division Multiple Access (TDMA) multi-hop IAB network. Distributed routing strategies (in which each IAB-node makes the next-hop decision) were studied in [26], where different path selection strategies were compared, including those based on the highest Signal to Interference plus Noise Ratio (SINR), the shortest path, and the load conditions on the routes. The authors showed that the shortest path strategy, which always selects an IAB-donor as the next hop, if reachable, is ineffective in sparsely deployed networks and for low SINR regimes. Mesh and tree-based IAB topologies were compared in [27], showing that mesh deployments can improve throughput and latency due to increased path diversity. Nevertheless, these studies focused primarily on algorithmic optimization, and do not capture the full protocol stack or realistic physical layer constraints. c) Mobility: Mobility introduces additional challenges due to dynamic topology changes and varying channel conditions. To mitigate interference in mIAB networks, silent slots
in the Time Division Duplexing (TDD) frame structure were proposed in [9]. The integration of unmanned aerial vehicles (UAVs) as mIAB-nodes has also been widely studied [28]– [30], showing that aerial relays can enhance coverage and enable flexible network deployment. For instance, [29] proposed joint optimization of power allocation, resource allocation, and UAV positioning to improve throughput, while [30] explored dynamic rerouting strategies using mathematical models to maintain connectivity in the presence of link blockages. III. IAB IN THE 3GPP In this section we review the 3GPP IAB specifications from Release 16. Sec. III-A introduces the IAB architecture, Sec. III-B describes the BAP, Sec. III-C presents the supported IAB topologies, Sec. III-D discusses multiplexing and scheduling mechanisms, and Sec. III-E focuses on mobility management enhancements introduced in Release 18. A. IAB Architecture IAB was standardized in the 3GPP 5G NR specifications in Release 16 [3], and is based on the functional split paradigm
4
introduced in Release 15. In this architecture, illustrated in Fig. 1, we distinguish three types of nodes: • UEs, acting as end users. The Uu interface connects the UE and its parent IAB-node. • IAB-nodes, i.e., gNBs with wireless backhaul. Each IABnode is divided into two logical units: a mobile terminal (MT), for the wireless backhaul connection toward an upstream IAB-node or IAB-donor, and a distributed unit (DU), for the access connection to the UEs or the downstream MTs of other IAB-nodes. • IAB-donors, i.e., full-blown gNBs with a classical wired (fiber) connection toward the core network. Each IABdonor is also divided into two logical units: a centralized unit (CU) and a DU. Specifically, the DU is closer to the radio and antenna elements, and implements a limited subset of time-critical radio functions such as scheduling, segmentation, and packet retransmission [31]. According to “Option 2” in TR 38.801 [32], the DU only terminates the RLC, Medium Access Control (MAC), and Physical (PHY) layers. For the User Plane (UP), the CUUP of the IAB-donor and the UE terminate the Service Data Adaptation Protocol (SDAP) and Packet Data Convergence Protocol (PDCP) layers. Data packets are exchanged via the F1-U interface, which consists of GPRS Tunnelling Protocol (GTP)-U tunnels from the CU-UP and the serving IAB-node DU. For the Control Plane (CP), the CU-CP of the IAB-donor and the UE terminate the Radio Resource Control (RRC) and PDCP layers. Control packets are exchanged via the F1-C interface, which uses the Stream Control Transport Protocol (SCTP) between the CU-CP and the serving IAB-node DU. B. Backhaul Adaptation Protocol (BAP) The BAP is an additional layer located above the RLC, as depicted in Fig. 1, which was introduced by the 3GPP to route the packets across the IAB topology, i.e., from the IAB-donor to the access IAB-node [33]. Each IAB-node is assigned a unique BAP address by the IAB-donor. For downlink (DL) packets, the BAP layer of the IAB-donor adds a BAP header that includes the destination BAP address of the access IABnode and the path ID, to discriminate among different routes to the destination. Similarly, for uplink (UL) packets, the access IAB-node adds a BAP header containing the destination BAP address of the IAB-donor and the path ID. Each IAB-node is configured with routing tables for UL and DL, indicating the child node (in the case of DL) or parent node (in the case of UL) to which the packet should be forwarded. Upon receiving a packet, the BAP layer checks the destination address in the BAP header. If it matches its own address, the packet is forwarded to the higher layers of the DU. Otherwise, the IAB-node delivers the packet to its DU and forwards it to the next node based on the routing table. The BAP layer is also responsible for mapping ingress and egress backhaul RLC channels to ensure that packets meet the desired Quality of Service (QoS) requirements. For bearers with strict QoS demands, a 1 : 1 mapping can be used, to dedicate an individual backhaul RLC channel to each hop. Otherwise, a 1 : N mapping is employed to multiplex packets from N bearers over a single backhaul RLC channel [31].
C. IAB Topologies IAB deployments exhibit a possibly multi-hop topology, where a well-defined parent-child relationship is present. Parent nodes can be represented by either an IAB-donor or an IAB-node; child nodes by either UEs or downstream IABnodes. In 3GPP 5G NR Release 16, the standard supported two types of IAB topologies [3]: spanning tree (ST) and directed acyclic graph (DAG). In the former, IAB-nodes are connected to a single parent, while in the latter multiple child-to-parent connections can be established. Clearly, an ST topology is less complex, but at the same time comes with resiliency constraints and performance degradation. For instance, in an ST, backhaul Radio Link Failures (RLFs) are likely to result in service interruptions for the end users, due to the lack of alternative backhaul routes. Instead, the DAG topology provides backhaul route redundancy, which can be used both to increase service availability and for load balancing. Notably, the 3GPP does not set an upper limit on the depth of either topology. Therefore, IAB protocols must provide support for an arbitrary number of backhaul hops [3]. Since Release 17, the 3GPP also supports an IAB-node MT to concurrently connect to two IAB-donors via NR DC [34, Sec. 11.3]. This approach effectively interconnects two IABtopologies, with the dual-connected IAB-node taking the role of the boundary node. Accordingly, it is in charge of possibly overwriting the BAP header routing information whenever packets are routed across the two topologies. D. Multiplexing and Scheduling In an IAB network, the backhaul link between a parent node DU and the child node MT mimics the link between a fiberequipped gNB and a typical 5G NR UE. Accordingly, there exist three possible types of time-domain resources just like for 5G NR UEs, i.e., DL, UL, or flexible (F) [35, Sec. 11.1]. However, in practice, the actual transmission direction (UL or DL) depends on the current availability of resources, and the IAB configuration, i.e., half- or full-duplex. Specifically, in case of half-duplex (when “IAB Simultaneous Operation” is disabled [36]), DUs/MTs cannot concurrently transmit and receive data. To reflect this constraint, time resources can also be marked as [37]: • Hard. The resource is always available to the DU, regardless of the MT configuration. Accordingly, other halfduplex MTs will mark these resources as “Not Available”. • Soft. The resource is available to the DU if and only if it does not collide with other MT transmissions/receptions. When “IAB Simultaneous Operation” is enabled, this is equivalent to a “Hard” resource. The actual availability of “Soft” resources can be inferred either implicitly or explicitly (in the latter case, based on the current resource allocations of the MTs). • Not Available. The resource is unavailable due to halfduplexing constraints. Since Release 17, the 3GPP also supports FDM of MTs and DUs by marking resources as “Hard,” “Soft,” and “Not Available” on a per-RB basis. This approach effectively enables concurrent transmission/reception by the MT and DU
5
S1-U F1-U BH-RLC
BH-RLC NR-Uu
PGW/SGW
(a) PDU session traffic (UE-anchored). S1-U
BH-RLC
BH-RLC
IAB-Node 3
PGW/SGW
(b) Non-PDU traffic (LAN-anchored).
Fig. 2: User-plane protocol stacks implemented in our IAB simulator. In Fig. 2a, PDU session traffic is anchored at the IAB-donor CU-UP, and uses the “inner PDCP” entity to establish internal GTP-U tunnels over the IAB backhaul. In Fig. 2b, non-PDU traffic bypasses the SDAP, PDCP, and “inner PDCP” layers, and is forwarded across the IAB backhaul via BAP-only routing.
interfaces, utilizing frequency division and spatial multiplexing techniques [34, Sec. 11.3]. Moreover, Release 17 further promotes space-division multiplexing (SDM) in IAB networks by introducing signaling between neighboring IAB-nodes regarding restricted and/or preferred beams, power adjustments, and transmission/reception timing alignments [34, Sec. 11.3]. E. Mobility Management 3GPP Release 18 introduces innovations for topology adaptation, interference mitigation, and mobility management to support IAB-nodes and UEs in dynamic environments [38]. Focusing on the latter aspect, a key improvement is the management of mIAB-nodes, such as those deployed in moving vehicles like buses or trains [39]. These nodes can provide 5G connectivity, but their mobility introduces challenges during IAB-donor switching, potentially causing disruptions for the UEs. To address this issue, Release 18 introduces a dedicated mobile CU (m-CU) with an Xn interface that connects to multiple IAB-donors [40]. The m-CU enables seamless transitions between IAB-donors within their coverage areas, ensuring uninterrupted handovers and maintaining stable UE connectivity. This innovation enables the 5G network to effectively support mIAB-nodes in highly dynamic environments. IV. E ND - TO - END SIMULATION OF IAB NETWORKS In this work, we designed and developed an open-source ns3 module for 5G IAB networks, named ns3-mmwave-iab1 . This module is built upon the ns3-mmwave framework for 3GPP NR networks [41] and a previous version of the code 1 https://github.com/signetlabdei/ns3-IAB.git.
presented in [15], that we extended to incorporate the latest IAB functionalities from Release 16,2 as described in Sec. III. Notably, the module was extended to simulate maritime scenarios, especially through a dedicated maritime-specific channel model and offshore scenario configurations. Still, the module remains sufficiently general to support future studies combining maritime and terrestrial segments and comparisons with alternative architectures (e.g., terrestrial-only IAB, donorbased deployments, and relays). In the remainder of this section, we present the main components of the proposed IAB simulation framework, specifically our BAP adaptation layer implementation (Sec. IV-A), the end-to-end protocol stack and data flow (Sec. IV-B), a maritime-specific channel model (Sec. IV-C), the slot format and control signaling structure (Sec. IV-D), and the MT and DU multiplexing schemes (Sec. IV-E). A. BAP Implementation First, we implemented in ns-3 the BAP adaptation layer (see Sec. III-B). This layer is located on top of the RLC layer, and is in charge of routing packets between the IAB-donor and the access IAB-node via wireless backhauling. In the module, we model the BAP Data Packet Data Units (PDUs), which are used to transport upper-layer data [33, Sec. 6.1.1], while the implementation of Control PDUs is left for future work. The header of a BAP Data PDU is of 3 bytes, and comprises the following fields [33, Sec. 6.3]: 2 The module is compliant with the latest 3GPP IAB specifications from the Radio Access Network (RAN) perspective. The Core Network (CN) is modeled according to the Long Term Evolution (LTE) specifications as defined in the ns3-mmwave module [41].
6
D/C (1 bit), indicating whether the packet represents a BAP Data PDU or Control PDU; • DESTINATION (10 bits), i.e., the BAP address of the destination IAB-node (DL) or IAB-donor (UL) DU; • PATH (10 bits), representing the BAP path ID; • R (3 bits), reserved for future use. Given that both the DESTINATION and the PATH fields are of 10 bits, the system can support up to 1024 potential pathways, and up to 1024 IAB-nodes, including the IABdonor. Additionally, we implemented the logic for handling BAP PDUs and Service Data Units (SDUs), as described in Sec. IV-B. •
B. Protocol Stack and End-to-End Data Flow In line with the 3GPP [42], in our simulator we reuse most 5G NR primitives to model the IAB protocol stack. Specifically, we model IAB-nodes as wireless backhauled base stations which feature two interfaces, i.e., the DU and the MT, as depicted in Fig. 2. The former provides connectivity to the UEs and the child MTs, like traditional base stations, while the latter connects to upstream DUs. 1) User Plane (PDU traffic): In the UP, each interface implements the 5G NR RLC, MAC, and PHY layers. Additionally, both MTs and DUs also implement the BAP adaptation layer, to handle routing and forwarding. Finally, we introduced in both IAB-node and IAB-donor DUs a so-called “inner PDCP” entity, as illustrated in Fig. 2a, which represents the GTP-U tunnel endpoint for the backhaul network. Accordingly, in our ns-3 implementation, DL UP packets are handled as follows. 1) Packets enter the CN from the Data Network (DN) via the Packet Gateway (PGW) and Serving Gateway (SGW), which together emulate the functionality of the 5GC User Plane Function (UPF). 2) Packets are then encapsulated in GTP-U tunnels. Each tunnel is identified by a Tunnel Endpoint Identifier (TEID) associated with a data bearer. In our implementation, Traffic Flow Templates (TFTs) [43] are used to classify packets into the corresponding tunnels according to LTE S1-U interface procedures [44]. Therefore, the role of TFTs is similar to that of the SDAP layer in 5G NR, i.e., mapping a QoS flow from the 5GC to a data bearer via the GTP-U extension header [45, Sec. 6.5]. 3) Once TEIDs are mapped to radio bearers, packets are forwarded to the corresponding PDCP instance. Then, the “inner PDCP” entity encapsulates UP packets in internal GTP-U tunnels, associating radio bearers to TEIDs and storing this mapping at the IAB-donor [3, Sec. 6.3.1]. GTP-U, User Datagram Protocol (UDP), and IP headers are then appended and packets are encapsulated over the Transport Network Layer (TNL); the IP Differentiated Services Code Point (DSCP) field is set according to the assigned TEID. 4) Subsequently, the PDCP instance forwards the encapsulated UP packet to the BAP layer. The latter leverages the TNL header to: (i) determine the BAP destination address and path ID; (ii) encode them in a BAP header; and (iii)
forward the BAP Data PDU to the corresponding RLC channel [3] [33, Sec. 6.3.2]. In our implementation, the mapping between PDCP entities and the corresponding BAP destination address and path ID is handled through a forwarding table maintained by the IAB-donor. 5) Once the BAP PDU is generated, it is propagated across the IAB backhaul. At each intermediate IAB-node, the local BAP instance checks whether the destination BAP address matches its own. If the address does not match, the node identifies the appropriate egress RLC channel associated with the indicated path ID and forwards the PDU to the next hop. When the BAP PDU reaches the destination IAB-node, the outer BAP, GTP-U, UDP, and IP headers are removed, thus terminating the internal GTP-U tunnel. 6) The packet is then passed to the RLC entity of the serving DU and transmitted over the access link to the destination UE. UL UP packets are handled as in the DL case, with the only difference that the invocation of the “inner PDCP” instance, and the corresponding encapsulation into the internal GTP-U tunnel, are performed at the access IAB-node rather than at the IAB-donor. 2) User Plane (non-PDU traffic): In addition to DL and UL traffic generated at the DN (in the 5GC or the UE, respectively), our simulator can support traffic that is not attached to a UPF, but rather terminates at the IAB-donor DU, as represented in Fig. 2b (in IAB-node 3). Such traffic may originate from, or be destined to, physical Local Area Network (LAN) interfaces located onboard IAB-nodes. To enable this functionality, we extend our baseline BAP implementation by using one of the three reserved (R) bits in the BAP header (see Sec. IV-A) to distinguish between PDU and non-PDU traffic. In the latter case, routing within the backhaul is handled as follows. When a data packet is received at the donor CU-UP, it is forwarded directly to the outbound BAP interface, bypassing the SDAP, PDCP, and “inner PDCP” layers at the IAB-donor. The packet is then transported through the IAB backhaul according to a pre-configured BAP destination address and path ID, and finally delivered to the LAN interface at the target IAB-node. 3) Control Plane: To simplify the CP design, we introduce an ideal logical link between each IAB-node DU and the Access and Mobility Management Function (AMF). This approach eliminates the need to transmit CP signals across the multi-hop IAB backhaul, which would introduce additional complexity. As such, the SRB0 and SRB1 radio bearers, which carry CP data, are directly terminated at the access IAB-node DU [3]. While assuming an ideal logical link does not allow us to capture the delay introduced by multi-hop control-plane signaling and handover procedures, maritime scenarios are generally characterized by low mobility and relatively stable topologies, which supports our assumption. Nevertheless, control signaling overhead is still captured through configurable control symbols in the NR frame structure. Moreover, notice that, while the AMF is a CP function, this abstraction permits focusing on the UP performance of the IAB network, which remains the primary objective of this work.
7
140 100 2,000
Modified 2-ray
180
Path loss [dB]
180
Path loss [dB]
Path loss [dB]
2-ray
140 100
4,000
2,000
Distance [m] (a) 5 GHz.
180 140 100
4,000
2,000
Distance [m]
4,000
Distance [m]
(b) 28 GHz.
(c) 60 GHz.
Fig. 3: Comparison of the path loss obtained using the 2-ray model and using the modified model proposed in [46], at different frequencies.
C. Maritime Channel Model The 3GPP 38.901 channel model [47], which is the de facto standard for system-level simulations of 5G and beyond systems, does not explicitly model maritime propagation scenarios. Rather, it supports five target environments, namely urban macro-cell (UMa), urban micro-cell (UMi), rural macrocell (RMa), indoor factory (InF), and indoor-office, primarily for terrestrial deployments. Among these, we identified the RMa scenario as the closest approximation to maritime conditions, due to the absence of large-scale obstructions and the reduced presence of scatterers. However, RMa does not fully capture the effects introduced by the sea surface (e.g., two-ray reflections and path loss variations with antenna height and sea state) [48]. Therefore, we extended the ns-3 implementation of the 3GPP TR 38.901 model in [49] with the following modifications to better represent coastal and offshore radio environments. 1) Path loss and propagation models: Electromagnetic propagation in maritime and coastal environments is characterized by strong first-order reflections from the sea surface3 , as well as by the formation of evaporation ducts. Accordingly, the 2-ray model [50] initially appeared as the most accurate theoretical model to represent the maritime scenario, as demonstrated through experimental results at 5 GHz [50], [51]. However, subsequent studies have shown that the 2-ray model becomes inaccurate at higher frequencies, such as in the 5G NR FR2 bands (e.g., n257 and n263) [46]. Therefore, in this work we employ the modified 2-ray model proposed in [46], which characterizes the path loss P L as [46, Eq. 5] α2π∆d λ P L [dB] = −20 log10 1 + R exp j , 4πd λ (1) where λ is the wavelength, d is the 3D distance between the transmitter and the receiver, and ∆d is the difference in length between the direct and first-order reflection paths. R and α are model parameters, which represent the reflection and unit-less frequency-dependent coefficients, respectively. In [46], R is assumed to be −1, while α is approximated via least-square minimization as α = 1.091 exp(−0.06256f )+0.06982, where 3 In most practical scenarios, the magnitude and phase of the reflection coefficient for vertically polarized signals are approximately 1 and 180◦ , respectively [50].
f is the carrier frequency in GHz. As depicted in Fig. 3, the modified 2-ray model leads to fewer path loss peaks for long (> 2 km) communication links than the classical 2-ray model [46]. Finally, in accordance with the study in [46], we consider deterministic LOS conditions, which is a common assumption for coastal and maritime areas. Although we focus on the maritime scenario, we did not implement the evaporation duct effect. In fact, this effect is primarily experienced for over-the-horizon propagation [52], while for short-range (up to 4 km) mostly-LOS links, such as those considered in this work, the duct contribution is negligible compared to other propagation phenomena, such as sea surface reflections and rain attenuation. 2) Rain attenuation model: In general, coastal areas are affected by more frequent precipitation than inland and oceanic areas, thus experiencing higher average rain rates [53]. In order to capture this peculiarity, we extended our simulator to account for rain-induced attenuation. Specifically, we implemented the ITU-R P.838-3 model [54], which characterizes the signal loss γr due to precipitation as [54] γr [dB/km] = kρα ,
(2)
where ρ is the rain rate in mm/h. k and α are model parameters which can be computed, respectively, as " 2 #! 4 X log10 f − bj log10 k = aj exp − cj j=1 + mk log10 f + ck ,
(3)
and α=
5 X j=1
" 2 #! log10 f − bj aj exp − cj
+ mα log10 f + cα ,
(4)
where f is the carrier frequency. The coefficients {aj , bj , cj }j=1,...,5 , mk , ck , and mα , cα depend on the signal polarization (vertical or horizontal) and are reported in [54, Tables 1–4]. To estimate ρ, the ITU provides historical rainfall data [55] which can be used to obtain a map of the average rain rates (mm/h) for each location on Earth. However, this map accounts for both rainy and non-rainy periods, resulting in an average
8
loss of only 0.2 dB/km, while severe storms can cause up to 15 dB/km of attenuation. Therefore, we treat ρ as a simulation parameter, which can be set at the beginning of the simulation based on the specific scenario and weather conditions.
D. Slot Format and Control Symbols In our simulator, we define three slot types based on the traffic direction: (i) DL slots for DL transmissions; (ii) UL slots for UL transmissions; and (iii) switching (SW) slots in between DL and UL slots. In an SW slot, the last 4 OFDM symbols are allocated for UL transmissions, while the preceding symbols act as a guard period to accommodate propagation delays and switching times. This configuration reflects realistic TDD operations, even though the slot format remains fully configurable in the simulator. Based on the assumptions in the ns3-mmwave baseline module [41], we reserve the first and last symbols of each slot for control signals, while the rest is available for data transmissions. However, to enhance the accuracy and flexibility of the simulations, we introduced an option to allocate additional control symbols (DL or UL) at a custom periodicity and equally distributed across all the slots.
E. MT and DU Multiplexing 1) Time Division Multiplexing: To reduce interference, our simulator assumes that each IAB-node can activate only one interface at a time.4 As a result, the MT and DU resources of the same IAB-node must be orthogonal (in either time or frequency). Let “layer” refer to the depth of an IAB-node in the network topology, with the IAB-donor at the root (layer 0). In our setup, the resources allocated to an IAB-node of an even layer must be orthogonal to those allocated to an IABnode of an odd layer. To implement this system, we assign nos symbols to IAB-nodes of the odd layers, while the remaining are allocated to even layers. For example, using TDM, if there is a total of 12 OFDM symbols for data transmission, the first nos symbols can be scheduled to IAB-nodes of an odd layer, while the rest can be used by IAB-nodes of an even layer. 2) Frequency Division Multiplexing: In addition to TDM, where all transmissions share the same spectrum resources, and interference is mitigated by the fact that the MT and DU of each IAB-node operate on orthogonal OFDM symbols, we implemented FDM. Specifically, the MT and DU of each IAB-node can operate simultaneously, albeit in orthogonal frequency bands, through carrier aggregation. This approach allows a flexible bandwidth division, supporting both intra- and inter-band aggregation. Consequently, the DU of each IABnode can schedule both DL and UL transmissions over the entire slot, using all of the available OFDM symbols (i.e., 12 out of 14, since 2 OFDM symbols are used for control signals, as described in Sec. IV-D). 4 The
3GPP TS 38.300 specifications define IAB-nodes as half-duplex devices [45], while the 3GPP TS 38.340 specifications assume that, for inband IAB, IAB-MT and IAB-DU transmissions and receptions are mutually exclusive in time [33].
TABLE I: Simulation parameters. Parameter
Value
DL source rate (rDL ) [Mb/s] Inter-packet interval [µs] Carrier frequency (f ) [GHz] Transmit power [dBm] Bandwidth [MHz] Rain rate (ρ) [mm/h] Beamforming technique IAB-node velocity [m/s] Simulation time [s] Simulation runs Antenna radiation pattern Channel model 5G NR numerology index TDD slot pattern Symbols for odd-layer nodes (nos )
{60, 80, 100, 120, 140} 50 26 30 400 {0, 15, 30} Codebook-based analog 5 2 50 3GPP model [47], 13 dBi gain Maritime model (see Sec. IV-C) 3 4DS2U 6
V. P ERFORMANCE E VALUATION OF A M ARITIME IAB S CENARIO To test and showcase the proposed 5G NR IAB simulator, we run a full-stack system-level simulation campaign of a maritime IAB deployment for remote connectivity. In particular, we consider a scenario where coastal base stations, with a fiber connection to the CN and the DN, play the role of IABdonors, and provide a wired endpoint to 8 IAB-nodes deployed on moving vessels. The IAB-nodes, located near the coastline and moving parallel to it at a velocity of 5 m/s5 , operate as UEs and transmit (receive) non-PDU UDP data packets originating from (destined to) their LAN interfaces. To evaluate the impact of the IAB network topology on the end-to-end performance, we consider the 4 different deployment options depicted in Fig. 4, which are defined based on the maximum depths of IAB-nodes and the number of IABdonors. In Topology 1, all IAB-nodes are directly connected to a single IAB-donor. In Topologies 2 and 3, the maximum IAB-node depth is 2 and 3, respectively. Topology 4 features two IAB-donors, and the IAB-nodes are inter-connected to obtain a maximum depth of 3. The IAB network implements in-band wireless backhauling, where the access and backhaul resources are allocated using either a static TDM (unless specified otherwise), or an FDM scheme. Both IAB-nodes and IAB-donors are equipped with a 64-element Uniform Planar Array (UPA) antenna, realizing codebook-based analog beamforming. Each element has a maximum gain of 13 dBi, and the radiation pattern is modeled as in the 3GPP specifications [47]. The UPAs of the IABdonors and all the DUs of the IAB-nodes are oriented such that their boresight direction is parallel to the positive X-axis. In contrast, for the MTs of the IAB-nodes, the orientation is toward their respective parent IAB-node.6 Both IAB-nodes and IAB-donors operate with a transmission power of 30 dBm, a 5 We assume a vessel speed of 5 m/s, which is compatible with coastal navigation. Nevertheless, in LOS conditions, higher speeds are not expected to significantly affect SINR, latency, or the overall performance trends. 6 The impact of the antenna orientation is negligible as long as the target node remains reachable through dynamic beamforming.
9
IAB-node
IAB-donor
Active MT-DU link 3,000 Y [m]
Y [m]
3,000
Speed vector
2,000 1,000 0
2,000 1,000 0
0
200
400
600
800
0
1,000 1,200
200
400
X [m]
800
1,000 1,200
800
1,000 1,200
X [m]
(a) Topology 1.
(b) Topology 2.
3,000 Y [m]
3,000 Y [m]
600
2,000 1,000 0
2,000 1,000 0
0
200
400
600
800
1,000 1,200
X [m] (c) Topology 3.
0
200
400
600 X [m]
(d) Topology 4.
Fig. 4: IAB network topologies. The arrows represent the speed vectors of IAB-nodes.
receiver noise figure of 5 dB, and a bandwidth of 400 MHz. The full list of simulation parameters is provided in Table I. In the following, we numerically validate the new features of our IAB simulator described in Sec. IV, specifically the impact of the rain rate (Sec. V-A), the slot pattern (Sec. V-B), and the MT and DU multiplexing (Sec. V-C). A. Rain Rate First, we assess how the rain rate affects the IAB network performance for different topologies. In Figs. 5a and 5c, we compare the Signal-to-Noise Ratio (SNR) and the SINR, respectively, under different meteorological conditions, ranging from clear weather to rain rates of up to ρ = 30 mm/h. In Topology 1, all IAB-nodes are directly connected to the IAB-donor. This configuration results in relatively long (i.e., up to approximately 3 km) child-to-parent links, which have a negative effect on the SNR. On the contrary, it guarantees an interference-free deployment since all transmissions are centrally coordinated by the IAB-donor through a TDMA-based scheduling scheme. In contrast, Topologies 2 and 3 support multi-hop connections, where IAB-nodes can act either as a parent or as a child, thus reducing the length of the link and improving the average backhaul SNR compared to Topology 1 (up to +15 dB). However, as described in Sec. IV-E, this configuration can create concurrent transmissions among IAB-nodes of the same layer (even or odd). Therefore, the resulting median interference is approximately 25 dB under clear weather conditions (see Fig. 5b, for ρ = 0 mm/h). Therefore, despite the longer links, Topology 1 achieves a higher SINR than Topologies 2 and 3. Finally, in Topology
4, the presence of two IAB-donors effectively creates two separate yet potentially interfering IAB networks, which may be a source of strong interference in certain conditions, as demonstrated by the high upper whiskers in Fig. 5b (up to more than 40 dB for ρ = 0 mm/h). As discussed in Sec. IV-C2, rain attenuation is directly proportional to the link length. In Fig. 5c we see that Topology 1 achieves the highest SINR without rain given the absence of interference. However, at ρ = 15 mm/h, the minimum SINR decreases by more than 10 dB, and at ρ = 30 mm/h the median SINR is even lower than in Topologies 2 and 3 due to the effect of the long links. Interestingly, for Topologies 2, 3, and 4, rainfall can successfully mitigate interference, as interfering paths are generally longer than the intended communication links, and thus experience greater attenuation. For example, the median SINR in Topology 3 improves by around 5 dB from ρ = 0 mm/h to ρ = 30 mm/h. Figs. 5d to 5f show the Packet Delivery Ratio (PDR) as a function of the DL source rate rDL and rain rate ρ. The traffic is mainly DL, where each IAB-node receives data with a DL source rate rDL , and transmits data with a UL source rate rUL = rDL /5. As expected, the PDR decreases as the source rate increases due to possible network congestion. Without rain, Topology 1 outperforms Topologies 2 and 3. Nevertheless, Topology 4 achieves the best performance, with a PDR of 1 with up to rDL = 100 Mbps, given the presence of an additional IAB-donor and additional available resources. As previously discussed, rain has a negative impact on Topology 1, while for Topologies 2, 3, and 4, rain attenuation actually helps reduce the effect of interference.
10
40 20 0
60
30 20 10
40 20
0
0 0 15 30 Rain rate (ρ) [mm/h]
(a) DL SNR.
0 15 30 Rain rate (ρ) [mm/h]
(b) DL interference.
(c) DL SINR.
1
1
0.8
0.8
0.8
0.6 0.4 0.2
PDR
1 PDR
PDR
Topology 4
40
0 15 30 Rain rate (ρ) [mm/h]
0.6 0.4 0.2
60 80 100 120 140 DL source rates (rDL ) [Mbps] (d) ρ = 0 mm/h.
60 80 100 120 140 DL source rate (rDL ) [Mbps] (e) ρ = 15 mm/h.
103 102 101
60
80 100 120 DL source rate (rDL ) [Mbps]
0.6 0.4 0.2
UL latency [ms]
DL latency [ms]
Topology 3
DL SINR [dB]
60 DL SNR [dB]
Topology 2 DL interference [dB]
Topology 1
140
60 80 100 120 140 DL source rate (rDL ) [Mbps] (f) ρ = 30 mm/h.
102 101
12
(g) DL latency, ρ = 0 mm/h.
16 20 24 UL source rate (rUL ) [Mbps]
28
(h) UL latency, ρ = 0 mm/h.
Fig. 5: DL SNR, DL interference, DL SINR, PDR, and DL/UL latency, as functions of the source rate for different IAB topologies.
Finally, in Figs. 5g and 5h, we evaluate the DL and UL latency under clear weather conditions using a 4DS2U slot pattern, meaning that each scheduling period consists of four consecutive DL slots, followed by a switch slot, and then two UL slots. The latency is defined as the time from when a packet is generated at the application layer of the local host to when it is received at the remote host (i.e., from the IABdonor to the IAB-node), and accounts for both transmission and queuing delays. We use a Round Robin (RR) scheduler, ensuring fair and balanced distribution of OFDM symbols across all IAB-nodes. We observe that the latency trends are similar to those of the PDR in Fig. 5d. As the source rate increases, the system approaches its capacity limits, and the endto-end latency is dominated by the resulting queuing delays. Although the 4DS2U pattern allocates fewer symbols to UL transmissions, the UL source rate is five times lower than the DL source rate. Consequently, UL queues are less congested than in DL, resulting in lower UL latency. Traffic accumulation
occurs primarily at IAB-nodes closer to the donor, where data traffic from multiple upstream nodes converges as it propagates toward the donor, and therefore experiences higher queuing delay as the source rate increases. For example, in Topology 3, the median DL latency increases from approximately 100 ms at rDL = 60 Mbps to 500 ms at rDL = 140 Mbps. The introduction of a second IAB-donor in Topology 4 increases the network capacity, reducing both DL and UL delays. In this configuration, the DL latency remains below 10 ms for source rates up to 100 Mbps. B. Slot Pattern In this section, we focus on Topology 3, and evaluate the impact of two different slot patterns vs. the source rate. Specifically, the 4DS2U slot pattern allocates 4 DL slots and 2 UL slots within each set of 7 consecutive slots, while the 3DS2U slot pattern allocates 3 DL slots and 2 UL slots within each set of 6 consecutive slots. In Figs. 6a and 6b, we set
11
3DS2U
1
1
0.8
0.9
UL PDR
DL PDR
4DS2U
0.6 0.4
0.8 0.7
0.2 0.6 60
80 100 120 DL source rate (rDL ) [Mbps]
140
6
14
(b) UL PDR, for rDL = 10rUL .
1
1
0.8
0.9
UL PDR
DL PDR
(a) DL PDR, for rUL = rDL /10.
8 10 12 UL source rate (rUL ) [Mbps]
0.6 0.4
0.8 0.7
0.2 0.6 60
80 100 120 DL source rate (rDL ) [Mbps]
140
(c) DL PDR, for rUL = rDL /5.
12
16 20 24 UL source rate (rUL ) [Mbps]
28
(d) UL PDR, for rDL = 5rUL .
Fig. 6: DL and UL PDR as a function of the source rate for different slot patterns in Topology 3.
rUL = rDL /10, and observe that the 4DS2U configuration improves the DL PDR compared to the 3DS2U configuration (0.75 vs. 0.65 for rDL = 60 Mbps), while for the UL PDR it is the opposite, given that the system configures more DL slots and transmission opportunities. In Figs. 6c and 6d, we increase the UL source rate to rUL = rDL /5, and see that the UL PDR further decreases (by up to around 20%), given the insufficient number of UL resources to handle the additional UL traffic. C. MT and DU Multiplexing In this section, we compare different MT and DU multiplexing mechanisms in Topology 3, considering a UL source rate equal to rUL = rDL /5. We compare TDM (for different values of nos ) vs. FDM with a balanced bandwidth partition between DU and MT. As described in Sec. IV-E, in TDM, nos denotes the number of OFDM symbols reserved to odd-layer IABnodes. In Topology 3, this includes layer-1 and layer-3 IABnodes. The remaining 12 − nos OFDM symbols are assigned to even-layer IAB-nodes, which in Topology 3 corresponds to the IAB-donor and layer-2 IAB-nodes. In Topology 3, backhaul links from the IAB-donor to layer-1 IAB-nodes need to carry the whole aggregated traffic generated and requested by all IAB-nodes in the network. Therefore, in Fig. 7 we observe that reducing nos (thus increasing the number of OFDM symbols that the IAB-donor can allocate) improves the overall network performance in terms of both PDR and latency. On the contrary, setting nos = 8 increases the capacity between layer-1 and layer-2 IAB-nodes, but reduces
the capacity of the IAB-donor, which becomes the bottleneck. For example, in Figs. 7b and 7d, we focus on the DL and UL latency as a function of the source rate. With rDL = 60 Mbps, rUL = 12 Mbps, and nos = 4, the median DL latency is below 10 ms, but it rises sharply to approximately 500 ms when nos = 8. The same conclusions can be drawn for the UL PDR and latency, which are consistently close to 1 and lower than 10 ms, respectively, for nos = 4 even when the source rate increases. These results demonstrate the critical role of resource allocation between MTs and DUs in IAB. Finally, FDM performs similarly to TDM with nos = 6, given that the DU and MT bandwidth is partitioned equally. VI. C ONCLUSIONS AND F UTURE W ORK In this paper, we presented a comprehensive study of 5G IAB networks, focusing on their performance in maritime environments. To this end, we developed and released an open-source, ns-3-based simulator fully aligned with the 3GPP Release 16 IAB specifications [42]. The framework integrates a dedicated maritime channel model capturing sea-surface reflections and a rain attenuation model, and supports flexible configuration of the control overhead, slot formats, and multiplexing mechanisms including both TDM and FDM. Through an extensive system-level simulation campaign, we observed several key insights. The performance of maritime IAB networks is strongly influenced by the deployment topology, weather conditions, and resource allocation strategies. Single-hop deployments offer lower latency and higher PDR under clear weather conditions, but are highly susceptible
12
TDM
nos = 4
nos = 6 DL latency [ms]
DL PDR
1
nos = 8
0.8 0.6 0.4 0.2
FDM
103 102 101 100
60
80 100 120 DL source rate (rDL ) [Mbps]
60
140
(a) DL PDR.
140
(b) DL latency.
UL latency [ms]
1 UL PDR
80 100 120 DL source rate (rDL ) [Mbps]
0.8 0.6
103 102 101 100
12
16 20 24 UL source rate (rUL ) [Mbps]
28
(c) UL PDR.
12
16 20 24 UL source rate (rUL ) [Mbps]
28
(d) UL latency.
Fig. 7: DL and UL PDR and latency as a function of the source rate for different multiplexing schemes in Topology 3.
to rain attenuation. Multi-hop topologies, despite the resulting interference, are characterized by shorter link lengths and therefore experience a higher SNR. Interestingly, heavy rain can attenuate interfering signals more than the desired ones, which may improve the network performance. These complementary behaviors indicate that no single topology is globally optimal, suggesting that dynamically reconfiguring the topology based on external factors such as weather and channel conditions could enable the network to maintain or even enhance the overall performance. Resource allocation also plays a central role. The configuration of DL and UL slots, and the distribution of resources between DUs and MTs, significantly affect the network performance. In multi-hop topologies, prioritizing resource allocation to the IAB-donor is essential to avoid bottleneck effects, as the donor must carry the aggregated traffic from all connected IAB-nodes. Moreover, our results show that, under equal bandwidth partition, FDM achieves similar throughput to TDM. This is expected, as the total amount of resources allocated to each node is the same in both schemes. However, FDM offers greater flexibility by enabling simultaneous parent-child operations, which can reduce latency and simplify scheduling in multi-hop topologies. The proposed simulator opens several future research directions. It can be used to validate theoretical models through endto-end simulations, and support IAB network optimization. The simulator can also be adapted to other non-terrestrial environments, such as satellite- or aerial-based networks.
R EFERENCES [1] N. Tafintsev et al., “Analysis of duplexing patterns in multi-hop mmWave integrated access and backhaul systems,” IEEE Open Journal of the Communications Society, vol. 5, pp. 5392–5407, Aug. 2024. [2] D. López-Pérez, M. Ding, H. Claussen, and A. H. Jafari, “Towards 1 Gbps/UE in cellular systems: Understanding ultra-dense small cell deployments,” IEEE Communications Surveys & Tutorials, vol. 17, no. 4, pp. 2078–2101, Fourthquarter 2015. [3] 3GPP, “NR; Study on integrated access and backhaul,” TS 38.874, 2019. [4] M. Polese et al., “Integrated access and backhaul in 5G mmWave networks: Potential and challenges,” IEEE Communications Magazine, vol. 58, no. 3, pp. 62–68, Mar. 2020. [5] Y. Zhang, M. A. Kishk, and M.-S. Alouini, “A survey on integrated access and backhaul networks,” Frontiers in Communications and Networks, vol. 2, pp. 1–17, Jun. 2021. [6] Y. Sadovaya et al., “Integrated access and backhaul in Millimeter-Wave cellular: Benefits and Challenges,” IEEE Communications Magazine, vol. 60, no. 9, pp. 81–86, Sep. 2022. [7] A. Topcu, S. A. R. Zaidi, and A. Q. Lawey, “Optimal resource allocation for O-RAN with integrated access and backhaul,” IEEE Open Journal of the Communications Society, pp. 10 061–10 078, Oct. 2025. [8] V. F. Monteiro et al., “Paving the way toward mobile IAB: Problems, solutions and challenges,” IEEE Open Journal of the Communications Society, vol. 3, pp. 2347–2379, Nov. 2022. [9] V. F. Monteiro et al., “TDD frame design for interference handling in mobile IAB networks,” in IEEE Global Communications Conference, 2022. [10] M. Choi, B. Yoon, D. Kim, and D. Sung, “IAB-based railway communication method for stable service provision,” in 12th IEEE International Conference on Ubiquitous and Future Networks (ICUFN), 2021. [11] Q. Li, J.-C. Sibel, M. Berbineau, I. Dayoub, F. Gallée, and H. Bonneville, “Physical layer enhancement for next-generation railway communication systems,” IEEE Access, vol. 10, pp. 83 152–83 175, Jul. 2022. [12] S. Horsmanheimo, L. Tuomimäki, A. Gentili, J. Scholliers, A. Sadam, and K. K. Kaare, “Maritime multi-hop communication over 5G networks: An indoor testbed implementation and evaluation,” in IEEE 100th Vehicular Technology Conference (VTC2024-Fall), 2024.
13
[13] G. Capela, W. Low, and L. Bastos, “5G for deployable and maritime communications,” in International Conference on Military Communication and Information Systems (ICMCIS), 2021. [14] G. Rafiq, P. Bose, and P. Orten, “5G cellular communication for maritime applications,” IEEE Access, vol. 13, pp. 109 451–109 472, Jun. 2025. [15] M. Polese, M. Giordani, A. Roy, S. Goyal, D. Castor, and M. Zorzi, “End-to-end simulation of integrated access and backhaul at mmWaves,” in IEEE 23rd International Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), 2018. [16] S. Zhang, X. Xu, M. Sun, X. Tao, and C. Liu, “Joint spectrum and power allocation in 5G integrated access and backhaul networks at mmWave band,” in IEEE 31st Annual International Symposium on Personal, Indoor and Mobile Radio Communications, 2020. [17] J. Y. Lai, W.-H. Wu, and Y. T. Su, “Resource allocation and node placement in multi-hop heterogeneous integrated-access-and-backhaul networks,” IEEE Access, vol. 8, pp. 122 937–122 958, Jul. 2020. [18] M. Pagin, T. Zugno, M. Polese, and M. Zorzi, “Resource management for 5G NR integrated access and backhaul: A semi-centralized approach,” IEEE Transactions on Wireless Communications, vol. 21, no. 2, pp. 753–767, Feb. 2022. [19] Y. Liu, A. Tang, and X. Wang, “Joint incentive and resource allocation design for user-provided network under 5G integrated access and backhaul networks,” IEEE Transactions on Network Science and Engineering, vol. 7, no. 2, pp. 673–685, Apr.–Jun. 2020. [20] C. Saha and H. S. Dhillon, “Millimeter wave integrated access and backhaul in 5G: Performance analysis and design insights,” IEEE Journal on Selected Areas in Communications, vol. 37, no. 12, pp. 2669– 2684, Dec. 2019. [21] W. Lei, Y. Ye, and M. Xiao, “Deep Reinforcement Learning-Based Spectrum Allocation in Integrated Access and Backhaul Networks,” IEEE Transactions on Cognitive Communications and Networking, vol. 6, no. 3, pp. 970–979, Sep. 2020. [22] T. Tian et al., “Field trial on millimeter wave integrated access and backhaul,” in IEEE 89th Vehicular Technology Conference (VTC2019Spring), 2019. [23] J.-H. Kwon, B. Lim, and Y.-C. Ko, “Resource allocation and system design of out-band based integrated access and backhaul network at mmWave band,” IEEE Transactions on Vehicular Technology, vol. 71, no. 6, pp. 6503–6517, June 2022. [24] H. Alghafari and M. S. Haghighi, “Decentralized joint resource allocation and path selection in multi-hop integrated access backhaul 5G networks,” Computer Networks, vol. 207, p. 108837, Apr. 2022. [25] C. Fang, C. Madapatha, B. Makki, and T. Svensson, “Joint scheduling and throughput maximization in self-backhauled millimeter wave cellular networks,” in 17th IEEE International Symposium on Wireless Communication Systems (ISWCS), 2021. [26] M. Polese, M. Giordani, A. Roy, D. Castor, and M. Zorzi, “Distributed path selection strategies for integrated access and backhaul at mmWaves,” in IEEE Global Communications Conference (GLOBECOM), 2018. [27] M. N. Islam, N. Abedini, G. Hampel, S. Subramanian, and J. Li, “Investigation of performance in integrated access and backhaul networks,” in IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), 2018. [28] I. Bor-Yaliniz, M. Salem, G. Senerath, and H. Yanikomeroglu, “Is 5G ready for drones: A look into contemporary and prospective wireless networks from a standardization perspective,” IEEE Wireless Communications, vol. 26, no. 1, pp. 18–27, Feb. 2019. [29] L. Zhang and N. Ansari, “A framework for 5G networks with inband full-duplex enabled drone-mounted base-stations,” IEEE Wireless Communications, vol. 26, no. 5, pp. 121–127, Oct. 2019. [30] M. Gapeyenko, V. Petrov, D. Moltchanov, S. Andreev, N. Himayat, and Y. Koucheryavy, “Flexible and reliable UAV-assisted backhaul operation in 5G mmWave cellular networks,” IEEE Journal on Selected Areas in Communications, vol. 36, no. 11, pp. 2486–2496, Nov. 2018.
[31] C. Madapatha et al., “On integrated access and backhaul networks: Current status and potentials,” IEEE Open Journal of the Communications Society, vol. 1, pp. 1374–1389, Sep. 2020. [32] 3GPP, “Study on new radio access technology: Radio access architecture and interfaces,” TR 38.801, 2017. [33] ——, “5G; NR; Backhaul Adaptation Protocol (BAP) Specification,” TS 38.340, 2020. [34] ——, “Release 17 Description; Summary of Rel-17 Work Items,” TR 21.917, 2023. [35] ——, “NR; Physical layer procedures for control,” TR 38.213, 2024. [36] ——, “NR; Integrated access and backhaul radio transmission and reception,” TR 38.174, 2024. [37] ——, “TSG RAN WG1 Meeting #99: RAN1 agreements for Rel-16 IAB,” R1-1913600, 2024. [38] X. Lin, “An overview of 5G advanced evolution in 3GPP release 18,” IEEE Communications Standards Magazine, vol. 6, no. 3, pp. 77–83, Sep. 2022. [39] C.-K. Wen, L.-S. Tsai, A. Shojaeifard, P.-K. Liao, K.-K. Wong, and C.B. Chae, “Shaping a smarter electromagnetic landscape: IAB, NCR, and RIS in 5G standard and future 6G,” IEEE Communications Standards Magazine, vol. 8, no. 1, pp. 72–78, Mar. 2024. [40] 3GPP, “Study on Architecture Enhancements for Vehicle-Mounted Relays,” TR 23.700, 2022. [41] M. Mezzavilla et al., “End-to-End Simulation of 5G mmWave Networks,” IEEE Communications Surveys Tutorials, vol. 20, no. 3, pp. 2237–2263, Third quarter 2018. [42] 3GPP, “Study on Integrated Access and Backhaul for NR,” RP-171880, 2017. [43] ——, “Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; 5G; Mobile radio interface Layer 3 specification; Core network protocols; Stage 3,” TS 24.008, 2021. [44] G. Piro, N. Baldo, and M. Miozzo, “An LTE module for the ns-3 network simulator.” in SimuTools, 2011, pp. 415–422. [45] 3GPP, “NR; NR and NG-RAN Overall Description; Stage 2,” TS 38.300, 2024. [46] N. Mehrnia and M. K. Ozdemir, “Novel maritime channel models for millimeter radiowaves,” in 24th IEEE International Conference on Software, Telecommunications and Computer Networks (SoftCOM), 2016. [47] 3GPP, “Study on channel model for frequencies from 0.5 to 100 GHz,” 2020. [48] M. Brata and I. Zakia, “Path loss estimation of 5G millimeter wave propagation channel–literature survey,” in 7th IEEE International Conference on Wireless and Telematics (ICWT), 2021. [49] T. Zugno, M. Polese, N. Patriciello, B. Bojović, S. Lagen, and M. Zorzi, “Implementation of a spatial channel model for ns-3,” in ACM Workshop on ns-3 (WNS3), 2020. [50] J. Reyes-Guerrero, M. Bruno, L. A. Mariscal, and A. Medouri, “Buoyto-ship experimental measurements over sea at 5.8 GHz near urban environments,” in 11th IEEE Mediterranean Microwave Symposium (MMS), 2011. [51] L. Yee Hui, F. Dong, and Y. S. Meng, “Near sea-surface mobile radiowave propagation at 5 GHz: measurements and modeling,” Radioengineering, vol. 23, no. 3, pp. 824–830, Sep. 2014. [52] G. S. Woods, A. Ruxton, C. Huddlestone-Holmes, and G. Gigan, “Highcapacity, long-range, over ocean microwave link using the evaporation duct,” IEEE Journal of Oceanic Engineering, vol. 34, no. 3, pp. 323– 330, Jul. 2009. [53] S.-Y. Ogino, M. D. Yamanaka, S. Mori, and J. Matsumoto, “How much is the precipitation amount over the tropical coastal region?” Journal of Climate, vol. 29, no. 3, pp. 1231–1236, Feb. 2016. [54] ITU-R, “P.838-3. Specific attenuation model for rain for use in prediction methods,” 2005. [55] ——, “P.837-7. Characteristics of precipitation for propagation modelling,” 2017.