Moving Networks (MN) in 5G and Beyond

• Moving Network (MN) is a wireless architecture where mobile infrastructure on vehicles, drones, or trains dynamically provides access alongside static cells.
• It complements fixed deployments by addressing issues like vehicular penetration loss, group handovers, and demand bursts with flexible, on-demand coverage.
• Studies show that integrating moving and static base stations can achieve up to 90% performance improvements through optimal resource and backhaul management.

Searching arXiv for recent and foundational papers on "Moving Network" in cellular systems, plus survey and optimization work. arXiv search query: all:"moving network" OR title:"moving network" OR abs:"moving base stations" Moving Network (MN) denotes a class of wireless-network architectures in which part of the access infrastructure is itself mobile. In the 5G and beyond-5G literature, MNs are implemented as mobile cells, moving relays, moving small cells, or moving base stations mounted on buses, trains, cars, or drones, and integrated as a mobile layer within a heterogeneous network. Rather than relying exclusively on fixed deployments, an MN repositions access supply in space and time to match non-uniform, bursty, or commuter-concentrated demand, thereby addressing vehicular penetration loss, group handovers, and static-network over-provisioning (Jaffry et al., 2020, Andreev et al., 2017, Finarelli et al., 2024, Finarelli et al., 21 Jul 2025).

1. Terminology and conceptual scope

In the cellular-systems literature, MN refers to a network whose infrastructure moves. The survey literature places MNs within 5G heterogeneous networks as a mobile layer that complements fixed macro cells and fixed small cells, especially for users inside public transport (Jaffry et al., 2020). A beyond-5G formulation generalizes the idea further: radio access points are no longer fixed only on buildings, but can be dynamically placed on moving, intelligent, capable devices such as cars and drones so that access supply is reconfigured on demand (Andreev et al., 2017). More recent optimization work instantiates the same paradigm through moving base stations mounted on ground vehicles and jointly dimensioned with static base stations over urban districts and time intervals (Finarelli et al., 2024, Finarelli et al., 21 Jul 2025).

The acronym is not uniform across networking research. In Mobile IP, OMNeT++ mobility, NetInf mobility, and related protocol work, MN commonly denotes the mobile node, namely the moving endpoint whose reachability or session continuity must be preserved [(Girgis et al., 2010); (Al-Rubaye et al., 2015); (Saleem et al., 2010)]. In Network Mobility (NEMO), a related but distinct usage appears: a mobile network is “a network whose attachment point to the Internet varies with time” (Dinakaran et al., 2010). This terminological collision is material, because Moving Network research in radio access is concerned primarily with mobile infrastructure, whereas Mobile IP and similar literatures are concerned primarily with mobile hosts and mobility anchors.

The central motivation for Moving Networks is the mismatch between static infrastructure and irregular demand. The beyond-5G MN literature explicitly argues that conventional cellular planning has relied on static radio access network deployments with gross over-provisioning, even though future traffic is spatially non-uniform, temporally bursty, and often concentrated around spontaneous events such as crowds, street performances, or large gatherings (Andreev et al., 2017). The survey literature frames the same problem in commuter settings: fixed deployments serve stationary or slow users reasonably well, but are poorly aligned with densely occupied, fast-moving vehicles and the resulting mass handovers and penetration losses (Jaffry et al., 2020).

2. Architectural forms and network composition

A canonical mobile-cell architecture for public transport separates in-vehicle and out-of-vehicle connectivity. The survey literature describes an Access Link (AL) antenna serving passengers inside the vehicle, an Out-of-Vehicle (OV) antenna providing backhaul to the macro network and sidehaul to neighboring mobile cells, and a Common Communication Unit (CCU) coordinating access, backhaul, and sidehaul resources (Jaffry et al., 2020). This architecture localizes user attachment within the vehicle, so passenger devices communicate with the on-board node rather than directly with the macro layer.

A broader beyond-5G architecture is the heterogeneous moving cell system. Here, access is provided by a mix of static mmWave access points mounted on buildings, vehicular MAPs (vMAPs) mounted on cars, and aerial MAPs (aMAPs) mounted on drones, with MAP denoting Moving Access Point (Andreev et al., 2017). These moving access points form a flexible overlay on top of ordinary infrastructure. They are not intended to replace static cells entirely; rather, they are activated, positioned, and reassociated dynamically where traffic is currently high. The role split is explicit: cars are abundant and useful when a modest amount of moving infrastructure is needed, whereas drones are less vulnerable to human-body blockage and street-level obstruction and become especially valuable when users are in the middle of a dense crowd (Andreev et al., 2017).

Recent analytical work translates this architectural idea into a two-tier urban deployment model comprising static base stations (SBSs) and moving base stations (MBSs) installed on vehicles (Finarelli et al., 21 Jul 2025). In a first-order variant, users and base stations are modeled as homogeneous planar Poisson point processes over a city partitioned into regions and time slots, and the design problem is to determine the optimal combination of moving and static BSs that minimizes the overall amount of deployed BSs while guaranteeing a target mean QoS for users (Finarelli et al., 2024). This suggests a convergence between the earlier mobile-cell concept for public transport and the later user-centric notion of demand-following moving infrastructure.

3. Control, association, and mobility management

Moving Networks require dynamic association control because attachment conditions vary with user mobility, blockage, interference, and backhaul availability. In the heterogeneous MAP literature, two association styles are distinguished. In user-controlled association, a user chooses the access point with the best signal power, provided it has enough resources, and may switch during a session using multi-connectivity for seamless transfer; this is characterized as a locally optimal, greedy approach. In network-controlled association, the network evaluates feasible beam configurations, optimizes a target metric, and instructs users and MAPs how to associate; this is treated as an achievable bound because the network has better global knowledge (Andreev et al., 2017).

MN research also extends control beyond supply-side adaptation. The beyond-5G formulation explicitly includes user-in-the-loop (UIL) concepts, socially aware cooperation, incentive-aware mechanisms, and possible penalties or rewards tied to using specific MAP services (Andreev et al., 2017). The underlying premise is that user-owned or user-carried devices are increasingly capable enough to participate in network functions, which blurs the classical distinction between infrastructure and user equipment. In this formulation, a Moving Network includes both mobile infrastructure and demand shaping.

For commuter-oriented mobile cells, mobility management is organized around decoupling passengers from the core network. The survey literature emphasizes that in-vehicle users connect only to the on-board access point, while the vehicle itself maintains the external backhaul connection, so the core network sees the vehicle as one moving entity rather than dozens or hundreds of separate moving UEs (Jaffry et al., 2020). This sharply reduces group handovers. In beyond-5G mmWave settings, session continuity is further linked to multi-connectivity, predictive handover, and probabilistic LoS/nLoS awareness, because short-range directional links are fragile under blockage and crowd dynamics (Andreev et al., 2017).

The hardest part of this control architecture is typically backhaul. The survey literature treats backhaul as the main bottleneck of MN design, with wired or optical solutions favored for deterministic rail deployments and wireless backhaul favored for buses and general urban scenarios (Jaffry et al., 2020). The newer urban MBS optimization literature assumes wireless backhauling, but either does not model its detailed performance (Finarelli et al., 2024) or models it explicitly as in-band SBS-to-MBS backhaul sharing the same bandwidth used for access, with weighted processor-sharing between broadband users and MBS traffic (Finarelli et al., 21 Jul 2025).

4. Modeling frameworks and performance metrics

The heterogeneous MAP vision has been evaluated through a simulation-based urban case study around central London. The scenario uses an area of about 0.5 km², 10,000 pedestrians, 1,500 vehicles, 30 drones, 10% AR glasses penetration, and 28 GHz mmWave access with 1 GHz bandwidth; static APs are mounted at 10 m and their density is varied from 20 to 250 APs/km² (Andreev et al., 2017). Pedestrians and vehicles follow a Manhattan mobility pattern, drones follow random straight trajectories, drones fly at 60 m, and speeds are 20 km/h for vehicles, 40 km/h for drones, and 3 km/h for pedestrians. Traffic spikes are induced by street performances that occur on average 10 times per hour, last 15 minutes, and attract 50% of pedestrians within 70 m. The radio model is a 3GPP UMi street-canyon mmWave model with line-of-sight and non-line-of-sight conditions, blockage by buildings, additional blockage by humans and vehicles, and 20 dB signal degradation when blocked (Andreev et al., 2017).

The principal metrics in that study are outage probability, defined as the probability that a session starts in outage when

$\mathrm{SINR} < 3 \text{ dB},$

and individual capacity share, defined as the proportion of total cell capacity received by an average user (Andreev et al., 2017). These are compared across static infrastructure only, static plus vMAPs, static plus aMAPs, and heterogeneous combinations of both.

Analytical work on moving base stations adopts a different abstraction. In the 2024 first-order model, the city is partitioned into $Z$ regions and $J$ observation slots, the performance metric is mean per-bit delay, and a closed-system assumption fixes the total number of moving BSs:

$$\forall j,\quad \sum_{z=1}^Z \lambda^{z}_{m,j} A_z = M.$$

The optimization seeks the optimal mix of moving and static BSs by minimizing

$$\underset{\Lambda_m,\mathbf{\lambda}_b}{\text{minimize} \quad M + \sum_{z=1}^Z \lambda_b^z A_z }$$

subject to QoS and utilization constraints (Finarelli et al., 2024). A 2025 extension retains the two-tier SBS/MBS structure but adds explicit wireless backhaul, strongest-SINR association, Shannon-capacity-based rate modeling,

$$C(r,P,I) = (B/k)\log_{2}(1 + P r^{-\alpha}(N_0+I(r,k))^{-1}),$$

weighted processor-sharing, Palm-expectation delay constraints, and a backhaul violation probability constraint (Finarelli et al., 21 Jul 2025). The resulting resource-optimal deployment problem is nonlinear and nonconvex, and is addressed through a three-step heuristic built around the Hippopotamus Optimization Algorithm (Finarelli et al., 21 Jul 2025).

5. Reported gains, applications, and operating regimes

The main system-level claim of the beyond-5G MAP literature is that moving networks provide non-incremental gains relative to static densification. As the MAP involvement factor (MIF) increases, outage probability drops substantially; at low MIF, vMAPs are more effective because cars are easier to exploit, whereas at higher MIF, aMAPs become preferable because they are less affected by blockage and interference. When outage requirements are stringent, drones are often necessary to serve users in dense crowds (Andreev et al., 2017). Heterogeneous use of vMAPs and aMAPs produces the strongest gains, with aggregate improvement reported in the range

$20\% \text{ to } 90\%$

depending on static mmWave AP density (Andreev et al., 2017). The same study argues that similar outage or capacity performance can be achieved with much lower static AP density if MAPs are deployed and maintained.

In the public-transport literature, the most frequently cited benefits are the circumvention of vehicular penetration loss (VPL), mitigation of Doppler effects on the passenger-facing link, and the reduction of group handovers (Jaffry et al., 2020). The survey reports VPL of as much as 25 dB in sub-6 GHz bands, webpage load time reduced from 7 s to 2.5 s, maximum load time reduced by 30 s, throughput improved from 0.1 Mbps to 0.25 Mbps in challenging train positions, VoIP latency increasing from 25 ms to 37 ms while remaining below the ITU-recommended 150 ms, control signaling reduced by 66% in a train scenario with user/core decoupling, and urban throughput increasing from about 200 Mbps in static macro/small-cell deployments to as much as 700 Mbps with vehicle-mounted infrastructure (Jaffry et al., 2020). The survey also treats mobile caching, mobile relaying, and adaptive radio access or on-demand densification for disasters, black spots, sports events, festivals, and urban hotspots as major application families.

Optimization studies sharpen these benefits into operating-regime statements. In the two-district 2024 study using measurement-based traffic profiles, the optimal hybrid solution yields an average reduction of about 17% in total deployed BSs compared with the fully static case, with savings reaching 21% in favorable conditions and persisting across different user-density ratios (Finarelli et al., 2024). The same work reports that the reduction in SBSs relative to the static case ranges from about 95% for low user-density ratios to about 70% for higher ratios (Finarelli et al., 2024). The 2025 stochastic-geometry analysis is more restrictive: in most cases the optimal solution is SBS-only, and MN solutions become attractive only when the relative unitary cost of MBSs is sufficiently low, especially when

$\mu \approx 0.1$

or lower, and for a middle range of user densities (Finarelli et al., 21 Jul 2025). This indicates that the benefits of movement are conditional rather than universal.

6. Constraints, misconceptions, and research directions

A common misconception is to treat a Moving Network as a replacement for static infrastructure. The primary MN papers instead describe mobile infrastructure as a complement or flexible overlay on top of ordinary infrastructure (Andreev et al., 2017, Jaffry et al., 2020). Static cells remain necessary for baseline coverage and capacity, while moving cells or moving base stations are introduced to supply temporary densification, mitigate commuter-specific impairments, or exploit temporal complementarity between districts.

The main technical limitation identified across the literature is backhaul. The survey treats the external link from vehicle to network as the principal bottleneck, especially in wireless deployments, and also highlights interference, mobility complexity, lack of strong analytical models, security and privacy concerns, energy efficiency, practical deployment, multi-operator sharing, and standardization as open problems (Jaffry et al., 2020). The 2025 optimization paper reaches a closely related conclusion from a different angle: the infrastructure savings from reuse across districts must outweigh the resource cost of in-band wireless backhauling, and this balance fails at very low or very high user densities (Finarelli et al., 21 Jul 2025). The 2024 first-order model is explicit that MBSs are assumed to move freely across the whole scenario and that wireless backhauling is assumed but not modeled in detail, which limits the realism of its savings estimates (Finarelli et al., 2024).

The literature also identifies increasingly specific next steps. The beyond-5G MAP work points toward ultra-dense moving cells with multi-connectivity, predictive handover, probabilistic LoS/nLoS awareness, and incentive-aware user involvement (Andreev et al., 2017). The survey emphasizes AI/ML-based mobility prediction and resource allocation, integrated access/backhaul/sidehaul optimization, mmWave and sub-THz backhaul for high-speed rail, secure and context-aware authentication, energy harvesting and low-power operation, UAV-assisted backhaul, and better handover design for ultra-high-mobility scenarios (Jaffry et al., 2020). The 2025 urban MBS study states more narrowly that a natural next step is to incorporate more realistic mobility patterns and the urban road grid geometry (Finarelli et al., 21 Jul 2025).

Taken together, these strands define Moving Network research as the study of mobile infrastructure as a controllable resource. Its unifying problem is the space-time matching of access supply to user demand. The commuter-oriented mobile cell, the beyond-5G heterogeneous MAP overlay, and the urban moving-base-station optimization framework are different realizations of the same principle: a network can trade fixed overbuilding for controlled mobility, but only subject to backhaul, interference, control, and cost constraints that remain active research problems (Andreev et al., 2017, Finarelli et al., 2024, Finarelli et al., 21 Jul 2025).