5G Mobile Cell (MC): Architectures Overview

• 5G Mobile Cell (MC) is a mobile 5G access system installed on vehicles or nomadic nodes that extends coverage and mitigates penetration losses.
• MC architectures employ both overlay and integrated models with CU/DU splits and mobile relay components to optimize network performance.
• Experimental evidence reveals that MC performance relies on precise geometric placement, balancing backhaul conditions with user throughput and QoS.

5G Mobile Cell (MC) denotes a mobile 5G access system installed on a mobile platform and used to provide 5G wireless connectivity to User Equipment (UE) in areas with limited fixed 5G infrastructure or under adverse radio conditions. In public-transport literature, the term also denotes a cell installed inside a train, subway, or bus, with a wireless backhaul toward the macro network and an in-vehicle access link toward passengers. The concept is therefore architectural rather than purely radio-physical: it combines mobile radio access, upstream transport, mobility management, and often temporary or on-demand deployment. The terminology is not uniform across 5G research, however, because some papers use “MC” strictly for Multi-Connectivity rather than Mobile Cell (Coelho et al., 2024, Ruela et al., 7 Aug 2025, Jaffry et al., 2018, Pupiales et al., 2021).

1. Terminology and conceptual scope

In the narrowest sense, a 5G Mobile Cell is a cell installed on a mobile platform such as a train, subway, or bus, intended to aggregate users inside the vehicle and to shield them from vehicular penetration effects while maintaining a wireless backhaul to the serving macro network (Jaffry et al., 2018). In a broader and increasingly common sense, a Mobile Cell is a mobile base station or mobile base station relay hosted by a 5G network, including public and non-public networks, and used to extend, restore, or reinforce coverage and capacity on demand (Coelho et al., 2024).

The broader interpretation includes both fully mobile and quasi-static operation. A Mobile Cell may operate as a mobile node mounted on ground vehicles, towboats, or drones, but it may also be used as a nomadic node placed temporarily at a required location for an event, emergency, or dynamic work area (Coelho et al., 2024). This broader definition is especially prominent in private-network settings such as seaports, where fixed infrastructure may be non-existent, insufficient, damaged, not feasible, or not cost-effective (Coelho et al., 2024).

A recurrent misconception is that “MC” always refers to Mobile Cell. In 5G standards-oriented literature, “MC” can instead mean Multi-Connectivity, namely simultaneous use of radio resources from multiple base stations and possibly multiple RATs such as LTE, NR, and Wi‑Fi (Pupiales et al., 2021). That usage is distinct from the mobile-cell meaning. Another important distinction is between direct Mobile Cell studies and enabling-infrastructure work: some papers are not about vehicular or mobile base-station cells themselves, but about architectural mechanisms—such as coreless control, MEC-aware association, or mm-wave discovery—that can support Mobile Cell deployments.

2. Architectural realizations

The recent Mobile Cell literature organizes MC design around 5G RAN decomposition into Central Unit (CU), Distributed Unit (DU), and Radio Unit (RU). The CU/DU split follows 3GPP split option 2, with CU implementing RRC and PDCP and DU implementing the lower layers below PDCP; CU and DU communicate through the IP-based F1 interface. For DU/RU, the literature cites eCPRI based on split 7.2 as the standardized low-level split interface (Coelho et al., 2024).

Two overlay realizations are prominent. In the first, the mobile platform hosts a full mobile gNB containing CU, DU, and RU, plus a Mobile Termination (MT) function that behaves as a regular UE. The MT attaches to a 5G overlay network, establishes a PDU session toward a UPF, and that PDU session carries the control and user traffic between the onboard gNB and the home network (Coelho et al., 2024). In the second, the mobile platform hosts only DU and RU plus the MT, while the CU remains fixed in the infrastructure; the F1-C and F1-U interfaces are then transported over the overlay-network connectivity created by the MT (Ruela et al., 7 Aug 2025).

The overlay design is attractive because it can exploit ordinary 5G coverage as transport, including the home network itself when within range or an external PLMN when outside the private-network footprint (Coelho et al., 2024). It is also modular. The 2025 architectural comparison formalizes two overlay subtypes—mobile gNB and mobile gNB-DU relay—and notes that the overlay solution is agnostic to the architecture of the overlay gNB because it only requires establishment of the MT PDU session (Ruela et al., 7 Aug 2025).

The integrated alternative is based on Integrated Access and Backhaul (IAB). In this model, the MC is a mobile IAB-node, specifically analogous to a Mobile Base Station Relay (MBSR), with an IAB-DU on the access side and an IAB-MT on the backhaul side. Instead of tunneling over an overlay UPF, the backhaul is native IAB backhaul toward an IAB-donor, with BAP support on the integrated backhaul path (Ruela et al., 7 Aug 2025). The integrated model is architecturally cleaner and supports end-to-end QoS natively, but it is less flexible because it requires an IAB-capable serving network and donor. Overlay models are more deployable where ordinary 5G coverage exists, but they insert additional transport overhead and expose timing sensitivity, especially when F1 must be carried over the overlay network (Ruela et al., 7 Aug 2025).

A further architectural distinction concerns function placement. A full mobile gNB is more self-contained and can, according to the later comparative analysis, host a UPF onboard, making localized services possible. A DU-relay architecture is lighter on the mobile platform but depends more strongly on overlay latency and throughput characteristics because it carries F1 over the wireless backhaul (Ruela et al., 7 Aug 2025).

3. Radio-link structure, backhaul, and spectrum reuse

In vehicular Mobile Cell literature, the canonical architecture contains two radio links. The first is the downlink backhaul (DL-BH) from the serving macro eNB to the mobile cell’s external backhaul antenna. The second is the downlink access link (DL-AL) from an in-vehicle directional antenna to the passenger UE inside the vehicle (Jaffry et al., 2018). The BH antenna is mounted externally, while the AL antenna is mounted inside the vehicle under the roof and designed to create a strong LOS component toward in-vehicle users.

This architecture is motivated by public-transport hotspot behavior and by vehicular penetration effect (VPE), also described as vehicular penetration loss (VPL), which the literature cites as reaching up to $25$ dB (Jaffry et al., 2018). Without an onboard cell, passengers connect directly to outdoor infrastructure through the vehicle body and suffer both poor QoS and repeated handover burden. With a mobile cell, users are unified behind the onboard node.

The principal spectrum problem is that an MC must maintain both a wireless backhaul and an access link, which increases pressure on scarce spectrum. One proposed solution is to reuse the same downlink subchannel for both DL-BH and DL-AL. In that model, the serving macro eNB allocates a subchannel $\omega$ for DL-BH, and the in-vehicle DL-AL reuses the same $\omega$ simultaneously (Jaffry et al., 2018). The feasibility argument relies on two physical effects: vehicular penetration attenuation, modeled by a penetration factor $0 < \varepsilon \le 1$, and the strong short-range LOS created by the directional in-vehicle antenna.

The corresponding SIR models are: $$\Upsilon_1(\omega,c\to m) = \frac{P_c r_c^{-\alpha_i}h^\omega_{c,m}}{I_C + I_o\varepsilon}$$ for the backhaul link, and

$$\Upsilon_2(\omega,o \to b) = \frac{P_o l^{-\alpha_o}h^\omega_{o,b}}{I_c\varepsilon}$$

for the in-vehicle access link (Jaffry et al., 2018). The BH side uses quasi-static Rayleigh fading and NLOS path loss, whereas the AL uses Rician fading and LOS path loss. The paper’s analysis and simulations show that both links can maintain high success probability, especially in low-SIR regions, when isolation is good and the AL is directional (Jaffry et al., 2018).

The trade-offs are structurally important. Better isolation, i.e. lower $\varepsilon$, improves both links. Higher macrocell density helps the backhaul because the serving macrocell tends to be closer, but can hurt the in-vehicle access link because outdoor interferers also become closer (Jaffry et al., 2018). The article therefore ties mobile-cell design not only to radio scheduling but also to vehicle construction, antenna placement, and the physical separation $X_d$ between AL and BH antennas.

4. Mobility, handover, security, and privacy

Dense small-cell 5G environments increase handover frequency because HeNB or HgNB coverage areas are small. That basic SCN effect is directly relevant to Mobile Cell scenarios, since moving users or moving platforms can traverse many small cells rapidly. The security literature therefore treats handover as an authenticated key-exchange problem whose repeated execution can dominate latency (Fan et al., 2018, Alnashwan et al., 2022).

One class of solutions is region-based fast handover. ReHand defines a region as one macro eNB together with its associated HeNBs. Entering a new region triggers an initial handover involving HSS/AuC and MME, but subsequent movement among HeNBs in the same region is handled locally with a region secret

$D_{ij}=H(GK_j,rID_i,T_{ex}),$

where $GK_j$ is the region group key, $\omega$0 an anonymous user identity, and $\omega$1 a warrant expiration time (Fan et al., 2018). ReHand also uses Nyberg’s one-way accumulator for active and passive revocation. Under the assumptions in that paper, the claimed total-cost reduction ranges from $\omega$2 to $\omega$3 relative to three comparison schemes (Fan et al., 2018).

A later privacy-aware region-based design reformulates the same idea in a gNB/HgNB setting and explicitly separates initial authentication, intra-region handover, and inter-region handover. Its distinctive mechanisms are sanitizable signatures for region-specific credential transformation, dynamic universal accumulators for revocation via non-membership witnesses, and ephemeral Diffie–Hellman for perfect forward secrecy (Alnashwan et al., 2022). In that scheme, intra-region handover is handled locally by the HgNB, whereas inter-region handover requires the target gNB to sanitize the modifiable region-specific component of the UE certificate. The paper positions this as the first scheme to address both intra-region and inter-region handovers with privacy and security in 5G communication (Alnashwan et al., 2022).

These handover schemes were proposed for dense SCNs rather than vehicular mobile cells in the strict sense. Even so, they target the same structural problem: repeated secure handovers caused by movement across many small cells. For Mobile Cell architectures, especially those used on moving platforms or in dense urban corridors, the key takeaway is the value of region-scoped trust, local re-authentication, pseudonymous identifiers, and efficient revocation.

5. Enabling 5G infrastructure for Mobile Cells

Several adjacent 5G architecture strands are not direct Mobile Cell studies, but they define the control, edge, and discovery mechanisms that make MC deployments viable. One example is the coreless mobile-network proposal, which replaces dedicated EPC nodes such as MME, S-GW, P-GW, HSS, and PCRF with open-source software functions running on general-purpose commodity hardware. In that architecture, a generalized wireless access point (WAP) may be a cellular base station, Wi‑Fi access point, UAV, or satellite, and the Software Defined Access (SoDA) model keeps BBU and RFU together at the WAP while centralizing higher-level radio-resource intelligence (Khan, 2015). The same work cites $\omega$4 Gb/s fronthaul for one $\omega$5 MHz $\omega$6 MIMO LTE channel in C-RAN, with 5G fronthaul possibly increasing by $\omega$7 or more, and notes that dedicated SoCs can offer up to $\omega$8 the performance per Watt of general-purpose processors for intensive PHY processing (Khan, 2015). These constraints explain why compact, locally self-contained mobile or distributed access nodes are favored.

Another enabling strand is MEC-aware cell association in 5G HetNets. For uplink task offloading, the relevant latency is the Extended Packet Delay Budget

$\omega$9

which combines radio transmission and execution time at the MEC host (Emara et al., 2017). That work compares conventional DL-RSRP-based association with a computational-proximity rule based on $\omega$0 and reports nearly $\omega$1 latency reduction at the 50th percentile when the radio/compute disparity ratio is $\omega$2 (Emara et al., 2017). For Mobile Cells, this is significant because a “best” serving node is not determined solely by radio strength when edge compute is part of the service path.

A third enabling issue is mm-wave initial access. In a control-plane/user-plane split architecture with legacy macro overlay and mm-wave small cells, directional cell discovery becomes a beam-search problem. Context-aware discovery procedures that use estimated UE position can substantially outperform random directional scanning; the paper reporting this states that random discovery has an average rendezvous time of $\omega$3 beam switches, while the Enhanced Discovery Procedure can minimize average delay around a $\omega$4 sector width for far-user distributions with $\omega$5 m location error (Capone et al., 2015). Such mechanisms are directly relevant when a Mobile Cell employs directional mm-wave access.

Finally, collaborative MEC is relevant because Mobile Cells often aggregate localized and delay-sensitive demand. A collaborative MEC framework spanning devices and MEC servers reports about a $\omega$6 gain in execution time for collaborative MEC over a single MEC server in a Canny-edge-detection task, while collaborative caching and processing reduce backhaul traffic for adaptive video delivery (Tran et al., 2016). This suggests that MC performance depends not only on radio placement but also on nearby compute and cache cooperation.

6. Deployment environments, experimental evidence, and open issues

The most direct experimental validation in the supplied literature is the modular evaluation of Mobile Cell architectures based on overlay and integrated models. That work emulates a mobile gNB-DU relay using OpenAirInterface, with OAI 5GCN, CU, DU, UE, and RF Simulator, all deployed as namespaces on a single Ubuntu 22.04 system. The testbed uses carrier frequency $\omega$7 GHz, bandwidth $\omega$8 MHz, transmit power $\omega$9 dBm, noise $0 < \varepsilon \le 1$0 dBm, and 3GPP UMi path loss. The fixed overlay gNB is placed at $0 < \varepsilon \le 1$1 m, while the Mobile Cell is evaluated at $0 < \varepsilon \le 1$2 positions from $0 < \varepsilon \le 1$3 m to $0 < \varepsilon \le 1$4 m, with three stationary UEs and downlink TCP throughput as the main metric (Ruela et al., 7 Aug 2025).

The central experimental conclusion is that MC positioning significantly influences performance. Across the $0 < \varepsilon \le 1$5 tested positions, UE1’s RSRP gradually decreases, UE2’s first increases and then decreases, and UE3’s steadily increases; throughput follows the same broad trend, with deviations explained by CQI estimation, MCS selection, retransmissions, and Proportional Fair scheduling effects when multiple UEs are active (Ruela et al., 7 Aug 2025). The experiment therefore validates not merely that an MC can function, but that its usefulness depends strongly on geometric placement relative to both upstream backhaul conditions and downstream UE positions.

Adjacent experimental evidence comes from device-centric and multi-hop cellular networking. Field trials using 4G LTE and IEEE 802.11 report cellular spectral-efficiency gains of up to $0 < \varepsilon \le 1$6 in outdoor pedestrian scenarios and $0 < \varepsilon \le 1$7 in vehicular NLOS scenarios when demand-driven opportunistic networking is combined with device-centric relaying. That literature explicitly cites vehicles acting as moving cells or moving relay nodes to connect passengers to the donor eNB, making it directly relevant to Mobile Cell thinking beyond fixed infrastructure (Coll-Perales et al., 2018).

The main open issue is that the evidence base remains uneven. Some MC papers are architectural and contain no equations or performance measurements, as in the on-demand private-network proposal for seaports (Coelho et al., 2024). Others provide detailed protocol-stack analysis and emulation, but only for one architecture subtype, namely the overlay-based mobile DU relay (Ruela et al., 7 Aug 2025). Moreover, overlay approaches still face unresolved end-to-end QoS questions, while integrated IAB-based approaches require IAB-capable infrastructure and are therefore less generally deployable (Ruela et al., 7 Aug 2025). This suggests that 5G Mobile Cell is best understood not as a single fixed design, but as a family of mobile RAN realizations spanning vehicular onboard cells, nomadic private-network extensions, and modular overlay or integrated backhaul architectures.