Nomadic Networks: Dynamic Mobile Infrastructure

• Nomadic Networks are communication systems where infrastructure itself is mobile and relocatable, enabling dynamic adaptation to changing topologies.
• They employ diverse deployment models such as spawned, steered, and wandering RANs along with flexible core network function placements.
• NNs are pivotal for 6G, disaster recovery, and industrial applications by innovating spectrum management, backhaul adaptation, and seamless mobility control.

Searching arXiv for papers on Nomadic Networks and closely related 6G/NNPN work. Nomadic Networks (NNs) are a class of communication systems in which mobility is no longer confined to end devices; rather, network infrastructure itself becomes mobile, relocatable, intermittently active, or dynamically placed. Across recent 6G-oriented literature, the term denotes architectures in which radio access elements, and in some formulations even core-network functions, can be activated on moving or temporary platforms and remain operational under changing topology, intermittent backhaul, and heterogeneous regulatory conditions (Lindenschmitt et al., 18 Aug 2025). In this sense, NNs differ from conventional cellular systems, which assume static infrastructure and user mobility, and from fixed non-public networks (NPNs), which provide private cellular capability but are generally location-bound while operating (Lindenschmitt et al., 2024). The concept also appears in more specialized forms such as Nomadic Non-Public Networks (NNPNs), mobile private networks formed by vehicles themselves, and industrial subnetworks that appear temporarily inside a larger overlayer network (Lindenschmitt et al., 26 Aug 2025).

1. Definition and conceptual boundaries

The defining characteristic of an NN is infrastructure nomadicity. In the broadest formulation, base-station locations, activity states, and coverage footprints are no longer fixed or centrally planned; instead, stationary, mobile, and nomadic infrastructure elements coexist in a density-varying, topology-varying system (Onur et al., 2017). Recent 6G work sharpens this definition by describing NNs as systems composed of mobile and self-organizing infrastructure nodes that provide radio infrastructure capabilities while in motion, with topology, service reachability, and interface connectivity becoming dynamic properties of the infrastructure itself (Lindenschmitt et al., 18 Aug 2025).

This distinguishes NNs from several adjacent concepts. A conventional stationary mobile network supports user mobility over fixed infrastructure; an NN introduces mobility of infrastructure components themselves (Lindenschmitt et al., 2024). Temporary networks may be short-lived, but their defining property is not necessarily mobility; NNs may also be temporary, yet their central feature is nomadicity, understood as deployability and operation across changing locations (Lindenschmitt et al., 2024). Non-public networks provide localized, self-operated cellular capability, but are not inherently mobile; the combination of private-network operation with infrastructure mobility yields the more specific notion of NNPNs (Lindenschmitt et al., 18 Aug 2025). In industrial 6G Networks-in-Networks (NiN), this idea appears again at a smaller granularity: a nomadic entity is not merely a roaming user device, but a relocatable or intermittently present subnetwork with its own controller, spectrum needs, and service profile (Lindenschmitt et al., 26 Aug 2025).

A recurring distinction in the literature is between moving users and moving networks. Conventional public cellular systems already address the former. NNs address the latter by allowing radio access networks, and in some cases core functions, to move, migrate, or be instantiated dynamically in the field (Lindenschmitt et al., 2024). This is why several papers treat nomadicity as a 6G architectural issue rather than only a mobility-management refinement (Lindenschmitt et al., 18 Aug 2025). A plausible implication is that NN research sits at the intersection of mobile infrastructure, distributed systems, spectrum governance, and self-organizing network control rather than within any one of those areas alone.

2. Historical development and main lines of interpretation

One major line of interpretation emerged from densification research, where mobile and nomadic small cells were introduced to increase coverage and capacity under spatially and temporally varying demand. In that literature, nomadic infrastructure includes mobile and nomadic small cells, “cells on wheels,” and “cells on wings,” and the central argument is that network density must be treated as an optimization parameter rather than as a static descriptor (Onur et al., 2017). This strand frames nomadicity primarily as dynamic infrastructure deployment in response to traffic hotspots, disasters, blind spots, and green-operation strategies such as sleep scheduling.

A second line developed around private and edge-enabled networks. The AMMCOA work describes groups of off-road vehicles that “span a mobile network by themselves,” with one vehicle acting as a network master and with local instantiation of core-network functions such as authentication when infrastructure is unavailable (Kochems et al., 2018). This interpretation is close to a mobile private network or nomadic 5G private network: the access point, gateway, and local network anchor are carried by the workgroup rather than fixed in space (Kochems et al., 2018). Its emphasis is not only on moving radio access, but on partial self-sufficiency through local edge/core capability.

A third line, now dominant in 6G discussions, treats NNs as an architectural class in which both RAN and, optionally, CN functions can be mobile, relocatable, or dynamically rehosted (Lindenschmitt et al., 2024). This line formalizes deployment modes such as spawned, steered, and wandering RANs, and integrated, migrated, and donated CNs, thereby generalizing beyond a single moving base station toward mobile service instances and Nomadic-Network-as-a-Service (NNaaS) (Lindenschmitt et al., 2024). Closely related work on architecture, standardization, and regulation extends this view by arguing that existing interfaces such as N2, N3, X2, and NG-C/NG-U were not designed for mobile infrastructure, dynamic topology discovery, or peer coordination (Lindenschmitt et al., 18 Aug 2025).

A fourth line arises from industrial 6G NiN research, where nomadicity is treated as a first-class systems property inside a shared industrial wireless system. Here a nomadic object is a temporary, movable subnetwork that must be discovered, connected, registered, allocated spectrum, and reconfigured as it arrives, moves, and departs (Lindenschmitt et al., 26 Aug 2025). This reframes nomadic networking from “device roaming” to “dynamic integration of subnetworks,” which is especially significant in modular manufacturing and logistics.

3. Architectural principles and deployment models

The recent 6G literature converges on the idea that NN architecture must be stateful, elastic, and topology-adaptive (Lindenschmitt et al., 18 Aug 2025). A central architectural question is the placement of RAN and CN functions. In the 6G architectural taxonomy, nomadic RANs can be classified as spawned, steered, or wandering. A spawned RAN is geographically fixed during operation but can be turned on and off dynamically for a constrained period or location; a steered RAN moves along a predefined route; a wandering RAN moves without a predefined route, or the route is unknown to the network (Lindenschmitt et al., 2024). These categories reflect different levels of predictability and correspond to different expectations for spectrum coordination, route planning, and orchestration.

Core-network deployment is classified separately. An integrated CN places a complete set of CN network functions within the nomadic infrastructure itself, using dedicated resources. A migrated CN preserves logical independence but redeploys CN functions on static infrastructure as the nomadic RAN moves. A donated CN leaves CN functions outside the nomadic system and attaches the nomadic RAN to an already existing CN (Lindenschmitt et al., 2024). The trade-off is explicit: integrated CN provides the highest independence, flexibility, security, and reliability, but increases CAPEX, OPEX, and onboard compute burden; donated or migrated CN reduces local burden but increases dependence on external infrastructure and orchestration (Lindenschmitt et al., 2024).

Work focused on interfaces refines these architectural ideas at the protocol boundary. In 5G, N2 connects the gNB to the AMF and N3 connects the gNB to the UPF. In nomadic operation, a gNB mounted on a vehicle, drone, or vessel may be disconnected from the central AMF for extended periods, making a purely centralized control model non-functional (Lindenschmitt et al., 18 Aug 2025). The proposed response is dual-mode operation of N2, with lightweight AMF-like capability embedded locally so that session management, mobility handling, and access control can continue autonomously during separation, followed by synchronization and reconciliation when connectivity returns (Lindenschmitt et al., 18 Aug 2025). On the user plane, the same logic leads to proxy-UPF functionality, buffering, delay-tolerant forwarding, and multi-hop relaying across mobile infrastructure nodes rather than a fixed tunnel to a remote UPF (Lindenschmitt et al., 18 Aug 2025).

Industrial NiN architectures instantiate similar principles in a different form. There, multiple independent application-specific subnetworks coexist under an overlayer network and are coordinated by a centralized spectrum manager. Each subnetwork has a sub-network controller, and control-plane connectivity is provided not by static, manually configured IP reachability, but by a self-organizing routing architecture supporting zero-touch operation, node mobility, moving network partitions, and DHT-based service discovery (Lindenschmitt et al., 26 Aug 2025). This suggests that NN architecture often combines two layers of adaptation: resilient control-plane discovery and continuity on the one hand, and dynamic radio-resource or spectrum reallocation on the other.

Architectural dimension	Main categories	Representative significance
Nomadic RAN mode	Spawned, steered, wandering	Encodes predictability of movement and deployment
CN placement	Integrated, migrated, donated	Encodes autonomy versus dependence on external infrastructure
Interface behavior	Dual-mode N2, proxy-UPF, transport-aware signaling	Encodes survival under disconnection and dynamic topology

A plausible implication is that the most useful architectural abstractions for NNs are not fixed node types but operational modes: how much of the stack moves, how much remains local during separation, and how state is reconciled after reintegration.

4. Mobility, control continuity, and resource adaptation

NN mobility management differs from classical user-centric mobility because the infrastructure itself moves. One consequence, emphasized in density-aware cellular work, is that even stationary users may need handovers when cells move (Onur et al., 2017). The literature therefore points toward new dynamic location-area concepts, virtual cells, motion or deployment planning, and registration or tracking of base-station locations (Onur et al., 2017). This is not a minor extension of existing mobility frameworks; it changes the object being managed from a moving terminal to a moving access system.

In industrial NiN environments, control continuity is achieved procedurally rather than through cellular-style handoff. A nomadic subnetwork arrives or is expected to arrive, obtains immediate control-plane reachability through self-organizing routing, discovers the central spectrum manager via service discovery, registers its presence and frequency requirements, and receives a new allocation after a negotiation process (Lindenschmitt et al., 26 Aug 2025). Reconfiguration may be planned or reactive. The paper repeatedly describes this as ad-hoc reconfiguration and seamless discovery and reconfiguration, with continuity arising from the combination of resilient control-plane routing and real-time spectrum reassignment (Lindenschmitt et al., 26 Aug 2025).

A more explicit handover-centric treatment appears in work on NNPNs for mobile healthcare. There the argument is that once a private network becomes mobile, the classical separation between handover and spectrum management becomes inadequate, because network attachment and spectrum access may both need to change simultaneously when moving across heterogeneous infrastructures (Lindenschmitt et al., 17 Jun 2026). The proposed architecture introduces an edge-based Spectrum Broker together with a Cognitive Spectrum Manager, and handover is made atomic in the sense that network selection and spectrum allocation are performed as a single coordinated procedure (Lindenschmitt et al., 17 Jun 2026). The decision process is based on periodic measurement collection, candidate-network evaluation, spectrum feasibility checks, ranking, unified RRC reconfiguration, and execution of the atomic handover (Lindenschmitt et al., 17 Jun 2026). The decision criterion is expressed as a weighted utility

$$U_i = w_1 \cdot \text{RSRP}_i + w_2 \cdot \text{SINR}_i + w_3 \cdot S_i$$

where candidate networks are jointly optimized over radio quality and spectrum feasibility (Lindenschmitt et al., 17 Jun 2026).

In density-aware mobile networks, the role of mobility management is framed more macroscopically. The literature introduces base-station density $\lambda$ and a critical density $\lambda_c$, distinguishing sparse and dense regimes by

$\lambda < \lambda_c$

and

$\lambda > \lambda_c$

with a phase transition near $\lambda_c$ where isolated coverage clusters merge into a giant connected component (Onur et al., 2017). This is not a handover procedure, but it provides a conceptual basis for why static mobility settings and resource configurations become inefficient in networks containing nomadic infrastructure. The main message is that density and mobility must be co-optimized rather than treated independently (Onur et al., 2017).

5. Spectrum, backhaul, and control-plane organization

Spectrum coordination is central to NN operation because nomadic infrastructure perturbs coexistence conditions and may move between administrative domains. In industrial NiN, Dynamic Spectrum Management (DSM) is the controller for the NiN concept: it centrally allocates frequency resources among subnetworks according to their requirements and current coexistence conditions, adapts spectrum allocation in real time, and supports both planned and unplanned reconfigurations while preserving application-level QoS requirements (Lindenschmitt et al., 26 Aug 2025). The control loop is qualitative rather than mathematical, but its operational logic is clear: preserve the most critical subnetwork, accommodate the nomadic subnetwork when it appears, and flex the less critical bandwidth-intensive subnetwork when necessary (Lindenschmitt et al., 26 Aug 2025).

At a broader cellular scale, nomadic base stations create local but sometimes non-local interference consequences. Work on transmit-power optimization in cellular networks with nomadic eNodeBs studies the problem of neighborhood selection when a temporary nomadic base station is added, because global re-optimization can reconfigure practically all cells (Grochla et al., 2019). The optimization objective is total network throughput, derived from SINR-dependent spectral efficiency and Round Robin sharing:

$$\eta_i = \log_2 \left(1+\frac{SINR_i}{\Gamma}\right), \qquad \Gamma = -\frac{2}{3}\ln(5 \times BER), \qquad BER = 0.00005$$

with total throughput written as the sum over all users (Grochla et al., 2019). The study concludes that a sampling-based local TX-power reconfiguration method can usually limit changes to fewer than 10 neighboring cells while keeping total network throughput within less than 1% of global optimization (Grochla et al., 2019). In NN terms, this shows that temporary infrastructure need not force network-wide replanning if local interference impact is identified correctly.

Backhaul is treated as both bottleneck and enabler. The 6G architectural literature emphasizes that moving RAN elements may require wireless backhaul, midhaul, fronthaul, or N6 connectivity, and that if hard latency requirements exist between RAN and CN, then the CN software may need to be dynamically migrated to remain close to the moving RAN (Lindenschmitt et al., 2024). Work on architecture and regulation further argues that N2 and N3 should be treated as logical constructs over adaptable transport layers rather than fixed bearers, because nomadic systems may use heterogeneous terrestrial and non-terrestrial links with different latency, jitter, and reliability (Lindenschmitt et al., 18 Aug 2025). In NNPN standardization-oriented work, NTN backhaul is presented as a major enabler, but with explicit constraints: LEO links can offer roughly 100–200 Mbps per individual link, HAPs can provide aggregate capacities around 1–5 Gbps for localized areas, while GEO one-way propagation delay of around 250 ms makes GEO unsuitable for uRLLC (Lindenschmitt et al., 14 Nov 2025).

In weak-satellite-coverage scenarios, UAV-assisted nomadic communications provide a more physical-layer view of the same problem. There the system is modeled as a cooperative satellite–aerial–terrestrial network in which UAVs act as nomadic aerial relays for a finite-size cluster of terrestrial users (Dong et al., 2024). The terrestrial terminals are modeled by a BPP inside a finite circular region, while UAVs follow a type-II Matérn hard-core point process with minimum separation $D_{\min}$, yielding aerial density

$$\lambda_{\textrm{A}} = \frac{1-\exp\big(-\pi D_{\mathrm{min}}^{2}\lambda_1\big)}{\pi D_{\mathrm{min}}^{2}}$$

(Dong et al., 2024). The paper shows that finite-region geometry, UAV spacing, altitude, coverage radius, density, and beamforming all materially affect access and backhaul reliability, thereby grounding nomadic relay design in explicit stochastic-geometry analysis (Dong et al., 2024).

6. Application domains and operational taxonomies

NNs are motivated by environments in which fixed infrastructures are unavailable, uneconomic, too rigid, or damaged. The 6G architectural literature repeatedly names large outdoor events, emergency situations, mobile industrial applications, agriculture, transportation, ships, truck platoons, UAV swarms, police, and military networks (Lindenschmitt et al., 2024). The architecture-and-regulation literature expands this to vehicular, drone, vessel, tactical, industrial, and non-terrestrial contexts (Lindenschmitt et al., 18 Aug 2025). The common thread is that connectivity must move with the operation rather than remain tied to a pre-existing site.

Industrial use is one of the clearest settings. In the NiN demonstrator, three subnetworks coexist under a common overlayer network: SN-1 supports mission-critical control between a machine tool and a virtualized CNC controller and requires deterministic xURLLC; SN-2 supports high-fidelity IIoT sensing for analytics and digital twin applications; SN-3 is a nomadic logistics and material-handling network associated with an AGV (Lindenschmitt et al., 26 Aug 2025). The AGV is virtually called to a machine to deliver or collect a workpiece, and its subnetwork introduces spatially and temporally varying spectrum demand. This framing is significant because it ties nomadicity to modular production and logistics rather than to generic mobile broadband (Lindenschmitt et al., 26 Aug 2025).

Private and non-public use cases have been systematized through KPI-oriented clustering. One survey-oriented 6G standardization paper distinguishes UUS-NNPN, SS-NNPN, and PC-NNPN: respectively, unpredictable, unscheduled, and safety-relevant NNPNs; scheduled, single-operator NNPNs; and predictable, cross-border NNPNs (Lindenschmitt et al., 14 Nov 2025). The associated use cases are PPDR and PMSE for UUS-NNPN, agriculture and construction for SS-NNPN, and transport for PC-NNPN (Lindenschmitt et al., 14 Nov 2025). Another closely related study defines KPIs such as duration, schedulability, motion predictability, maximum coverage, cross-border interaction, number of active operator, uplink to public network, safety-relevant communication, QoS aspect, and key aspect according to ITU-R M.2160-0, using them to cluster NNPN applications (Lindenschmitt et al., 2024). These taxonomies do not provide a formal evaluation function, but they standardize the descriptive dimensions along which nomadic deployments differ.

Urban augmentation is another application line. A Kathmandu case study examines LTE Nomadic Nodes as portable, self-contained cellular networking units placed at scale to densify an existing LTE network in a dense urban environment (Mavromatis et al., 2024). Using a Digital Twin based on the DRIVE framework, the study models buildings, roads, incumbent BS sites, user mobility, and traffic demand, and then optimizes NN placement through Delaunay triangulation and low-RSS candidate selection (Mavromatis et al., 2024). This use case interprets nomadicity pragmatically as flexible, lower-cost deployability rather than necessarily as in-motion operation, but it still belongs to the NN family because the nodes are not permanently fixed and can be added or relocated as needed (Mavromatis et al., 2024).

Off-road vehicle networks, such as AMMCOA for agriculture and construction, form yet another class. There a group of vehicles forms a local network, one becomes network master and acts as a base station, and the role can migrate to the vehicle with the best infrastructure link (Kochems et al., 2018). The local network may continue operating without public-network connectivity and may instantiate certain core functions locally, which makes it a mobile, self-contained private network rather than a mere V2X system (Kochems et al., 2018).

7. Performance evidence, open challenges, and contested assumptions

The NN literature is rich in architectural and systems arguments but comparatively uneven in rigorous quantitative evaluation. Some papers are demonstrator-oriented rather than benchmark-driven. In industrial NiN, the strongest stated quantitative result is that mission-critical control is supported with latency between 2–3 ms and time-deterministic behavior in a demonstrator with three subnetworks, Raspberry Pi-based master nodes, custom radio hardware, KIRA in the backbone, and a dashboard visualizing network state (Lindenschmitt et al., 26 Aug 2025). The paper explicitly does not provide comparative baselines, confidence intervals, packet error rates, reliability percentages, handover interruption times, throughput traces, or control-loop stability measures (Lindenschmitt et al., 26 Aug 2025). This supports feasibility rather than exhaustive characterization.

By contrast, the atomic handover study offers a concise KPI comparison. In a MATLAB-based simulation of a mobile healthcare scenario, baseline handover interruption time is reported as 420 ms versus 280 ms for atomic handover; signaling steps decrease from 6 to 4; spectrum reconfiguration delay falls from 150 ms to 0 ms because it is integrated into the handover; packet loss during handover drops from 3.5% to 1.8%; and handover success rate improves from 96.2% to 98.7% (Lindenschmitt et al., 17 Jun 2026). The paper also reports 180 spectrum allocation decisions over the simulation and claims sub-10 ms added latency for local Spectrum Broker decisions under the assumed edge-processing model (Lindenschmitt et al., 17 Jun 2026). Even here, however, the evaluation is limited to a single mobile network, a small number of handover events, and no multi-user environment (Lindenschmitt et al., 17 Jun 2026).

The Kathmandu Digital Twin study provides another quantitative perspective. Augmenting 49 incumbent LTE BSs with up to 20 nomadic nodes shifts median RSS from about $-45$ dBm to about $\lambda$0 dBm, with all two-sample Kolmogorov–Smirnov tests for 5, 10, 15, and 20 added NNs yielding $\lambda$1 (Mavromatis et al., 2024). Average user datarate rises from 6.79 Mbps in the LTE-only baseline to 11.18 Mbps in the combined LTE+20-NN system, roughly a 65% increase; the NN-only system reaches 5.86 Mbps, implying that the principal value of the nomadic layer lies in augmentation rather than replacement (Mavromatis et al., 2024). These figures suggest that NN concepts can also be productive in dense urban enhancement, not only in isolated or private deployments.

Several recurring limitations and controversies remain. First is centralization versus adaptability. A centralized spectrum manager or CN simplifies coordination, but may create dependencies on central decision points and stable control reachability; papers routinely note that redundancy or failover for such central entities is under-specified (Lindenschmitt et al., 26 Aug 2025). Second is the gap between conceptual architecture and formal models. Many of the most influential papers explicitly contain no mathematical formulations for routing, spectrum allocation, utility maximization, or delay/reliability constraints, and describe optimization logic entirely in prose (Lindenschmitt et al., 26 Aug 2025, Lindenschmitt et al., 18 Aug 2025, Lindenschmitt et al., 2024). Third is regulation. Multiple papers argue that static, geographically fixed spectrum licensing is fundamentally misaligned with mobile infrastructure, and that manual processes taking weeks or months are incompatible with dynamic deployment (Lindenschmitt et al., 18 Aug 2025, Lindenschmitt et al., 14 Nov 2025). Fourth is trust and security. The literature often identifies local authentication, dynamic trust establishment, secure signaling, and decentralized trust anchors as essential, but detailed protocols remain sparse (Lindenschmitt et al., 18 Aug 2025, Lindenschmitt et al., 14 Nov 2025).

This suggests that NN research is currently strongest in problem formulation, system decomposition, and demonstrator-level integration, while remaining comparatively immature in standardized control procedures, cross-domain trust frameworks, and large-scale empirical validation. Even so, the field has converged on a durable core insight: nomadicity is not a special case of user mobility, but a distinct networking condition in which infrastructure, interfaces, spectrum rights, and management authority all become mobile variables.