What Is the Cost of a Handover Interruption? Rethinking 5G Mobility for Aerial Vehicles

arXiv:2610.17380v1 [cs.NI] 12 Oct 2026

What Is the Cost of a Handover Interruption? Rethinking 5G Mobility for Aerial Vehicles André Melo

Yueqi Chen

Aeronautics Institute of Technology, Brazil Brazil andre.melo@ita.br

Northwestern University, USA USA yueqichen@northwestern.edu

Igor Kadota

Lourenço Alves Pereira Jr.

Northwestern University, USA USA kadota@northwestern.edu

Aeronautics Institute of Technology, Brazil Brazil lourenco.junior@gp.ita.br

Abstract Low-altitude aviation is moving toward large-scale operations increasingly dependent on 5G cellular connectivity. However, 5G mobility remains blind to the operational risk of its decisions: the same interruption may be harmless in cruise over an empty field but critical near obstacles, crowds or other aircraft. Although 5G already provides mechanisms to obtain location, population density, and radio conditions, mobility decisions do not translate this information into an operational risk metric. We propose Risk-Aware Mobility (RAM), which maps predicted communication disruptions to an operational-risk proxy using information staleness, position uncertainty, and operational context, and exposes this risk to existing mobility mechanisms. In a dense São Paulo deployment with 188 licensed 5G sectors, risk-aware handover timing reduces the mean operational-risk proxy by 45% relative to conventional A3 without degrading servinglink quality. These results motivate shifting 5G mobility from optimizing handover performance to managing the operational risk of mobility decisions. Beyond handover timing, this perspective raises a broader question: which cells should serve aerial operations when connectivity is evaluated by operational risk rather than radio performance alone?

1

Introduction

Low-altitude airspace is undergoing a fundamental transformation. Aviation authorities worldwide are developing traffic-management frameworks for large-scale operations of unmanned and increasingly autonomous aircraft [13, 23, 25– 27]. The United States has established a national strategy for Advanced Air Mobility, while the United Kingdom targets routine beyond-visual-line-of-sight (BVLOS) drone operations by 2027. Commercial UAV delivery services are already emerging in China, the Middle East, and Brazil [12, 19, 24]. Scaling these operations requires timely exchange of aircraft

position, telemetry, and traffic information among aircraft, operators, and airspace-management services [13, 15, 23]. Commercial cellular networks are a promising component of this infrastructure, offering wide-area coverage, highcapacity communication, and integration with edge services [27, 28]. The question is shifting from whether 5G can connect aerial vehicles to whether its mechanisms can support their operational requirements. Unlike terrestrial users, UAVs experience aerial propagation that exposes them to multiple base stations, irregular antenna sidelobes, and rapid changes in serving-cell quality [1, 14, 16]. This creates a fundamental mobility trade-off: switching cells may interrupt communication, while staying connected may lead to radiolink failure. Either outcome can interrupt delivery of fresh aircraft state. Commercial measurements confirm frequent aerial handovers and second-scale data-plane outages even after nominally successful transitions [20]. Equal interruptions need not imply equal operational risk. Consider two UAVs experiencing the same 500-ms data-plane gap: one over sparsely populated ground, the other near a crowd or another aircraft. As position information becomes stale, uncertainty about the aircraft’s location grows; its consequences depend on surrounding traffic, operating constraints, and ground exposure. Consistent with this operationdependent view, the Specific Operations Risk Assessment (SORA) requires communication performance appropriate to the intended operation [17, 18]. Existing mobility mechanisms are designed to maintain connectivity and limit service disruption. Mobility Robustness Optimization (MRO) addresses too-early, too-late, and wrong-cell handovers; Conditional Handover (CHO) prepares targets in advance; and Dual Active Protocol Stack (DAPS) can reduce service interruption [3, 8]. UAV-specific approaches further exploit flight paths and radio conditions to improve mobility decisions [10, 11]. These mechanisms serve their intended purposes, but were not designed to

André Melo, Yueqi Chen, Igor Kadota, and Lourenço Alves Pereira Jr.

distinguish the operational consequences of the same communication disruption across different aerial contexts. The gap is not how 5G executes mobility, but what its decisions optimize. Radio measurements can guide when to switch or stay, but do not express what losing fresh aircraft information would mean at that point of the flight. Rather than replacing existing mobility procedures, we ask whether risk can become an additional decision metric. This gap extends beyond handover timing. Radio performance also determines which cells are preferred along an aerial route, without necessarily reflecting the operational consequences of their selection. We propose Risk-Aware Mobility (RAM), which maps communication induced information staleness into position uncertainty and an aircraft-specific risk proxy. RAM uses this quantity to inform existing mobility decisions, without requiring new air-interface procedures or UE modifications. Its inputs can be derived from network measurements, 5G positioning and analytics functions, and geographic exposure data [4, 7]; 3GPP aerial-UE flight-path reporting could further refine the estimate [2, 6, 8]. Can an operational-risk proxy improve 5G mobility decisions? We investigate this question with commercial 5G traces, OpenAirInterface (OAI) experiments, and a dense São Paulo deployment reconstructed from 188 licensed 5G sectors [1, 5, 9]. Our evaluation reveals a non-monotonic trade-off: suppressing handovers initially lowers operational risk, but excessive deferral increases it as the serving link deteriorates. Risk-aware handover timing reduces the mean risk proxy from 0.289 to 0.160 (45%) relative to a conventional 3-dB A3 configuration while maintaining serving-link quality. This is a reduction in a modeled risk proxy, not a measured reduction in accident probability. Our contributions are threefold: (i) We characterize the competing costs of switching and staying using commercial 5G traces and OAI experiments. (ii) We introduce RAM, which extends an existing operational-risk abstraction to 5G mobility decisions [21], enabling risk-aware handover timing. (iii) We show that risk-aware handover timing lowers modeled operational risk beyond conventional A3 tuning in a dense urban deployment.

2

When should a UAV switch cells, and when should it stay?

A mobility controller must balance the immediate disruption of switching cells against the increasing risk of staying on a deteriorating link. We characterize this trade-off using commercial 5G traces and controlled OAI experiments.

2.1

Observation 1: Switching Can Interrupt Data Delivery

We analyze 547 traces from the 5G SA measurements of Liu et al.[20], spanning four operators, several altitudes, and rural and urban environments. We identify a post-handover data-plane outage when a handover recorded as successful is followed by at least one second without L1 throughput within a two-second observation window. This trace-derived criterion is not itself a 3GPP handover-failure classification. Successful handover does not guarantee continuous data delivery. At 300 m, aerial UEs experience 14.6 handovers/km, compared with about 9.2/km on the ground. Of these aerial transitions, 16.9% are followed by a posthandover data-plane outage, with a median duration of about 2.2 s. Along the urban–rural route, the measured fraction varies from 0.14 in its rural segment to 0.22 near the urban– rural transition. Its duration, however, does not depend on where it happens: the median is 1.9–2.2 s at every altitude from 80 to 300 m. What varies along a flight is how likely the outage is and what lies beneath. Liu et al. identify a related mechanism, termed pseudosuccessful handover, in which control-plane signaling succeeds but target uplink access fails [20]. The trace-derived outages can include events consistent with this mechanism, but do not individually establish their cause. These measurements reveal a cost of switching that conventional handover-success indicators may not capture: a completed mobility procedure can still leave the aircraft temporarily unable to deliver fresh information. However, avoiding handovers altogether is not a solution, as the serving link may deteriorate while the aircraft continues moving.

2.2

Observation 2: Staying Can Destroy Connectivity

We investigate the competing costs of switching and staying in a controlled OAI testbed [22], with one CU, two DUs on different carriers, and one UE in the RF simulator. Per-link attenuation is varied along a flight trajectory while a UDP probe measures data-plane continuity. Switching has an immediate cost. A regular inter-frequency handover introduces a 377-ms data-plane gap in our setup. Deliberately blocking target uplink access causes repeated random-access failures and a multi-second interruption. These results illustrate why postponing a handover may initially appear beneficial for continuous data delivery. Staying has an increasing cost. The serving link withstands approximately 23 dB of additional attenuation before deteriorating sharply, with radio-link failure between 30 and 35 dB. In one controlled run, excessive handover deferral lost 43% of probe packets, compared with 1.9% when deferral was

What Is the Cost of a Handover Interruption? Rethinking 5G Mobility for Aerial Vehicles

constrained by the serving-link limit. These are results for the tested configuration, not universal 5G thresholds. Thus, postponing a handover can avoid an immediate interruption, but only while the serving link remains capable of carrying data. Beyond that point, the accumulated cost of staying can exceed the cost of switching. Takeaway. Switching and staying impose competing communication costs, but radio measurements alone cannot determine their operational consequences. Comparing these alternatives requires a risk metric that accounts for the aircraft and its operating context.

3

From Communication Disruption to Operational Risk

We now map communication-induced information staleness into an operational-risk proxy for comparing candidate handover decisions. From staleness to uncertainty. Consider a UAV transmitting aircraft status and, when required by its operation, video for remote situational awareness. Let Δ𝑠 (𝑡) denote the age of the latest status update received by the network. During an interruption, the network continues relying on that update even as the UAV moves. Following the conservative containment abstraction used in[17, 18], we bound its position uncertainty by 𝑟 (𝑡) = 𝑟 min + 𝑣 max Δ𝑠 (𝑡),

(1)

where 𝑟 min is baseline localization uncertainty and 𝑣 max bounds aircraft speed. The corresponding region B (𝑡) is a disk centered at the last reported position. This is a bound on the network’s knowledge of the aircraft, not an estimate of navigation error or loss of onboard control. From uncertainty to consequence. We retain three operational consequences [21] from the risk abstraction in [17, 18]: consumed separation margin, ground exposure, and loss of video coherence. Let 𝑑ℎ (𝑡) denote the distance from the last reported position to each relevant hazard ℎ (another aircraft, a restricted volume, an obstacle) and let 𝑟ℎ (𝑡) include its protection distance and positional uncertainty. We define 1 ∑︁ 𝑅 𝑁 (𝑡) = 𝑁 [𝑟 (𝑡) + 𝑟ℎ (𝑡) − 𝑑ℎ (𝑡)] +, (2) 𝑅¯ ℎ ∫ 1 𝑅𝐺 (𝑡) = 𝐺 𝜌 (x, 𝑡) 𝑑x, (3) 𝑅¯ B (𝑡 ) 𝑅𝑉 (𝑡) =

[𝑧 (𝑡) − 𝐻 nom ] + . 𝑅¯𝑉

(4)

Here, 𝜌 (x, 𝑡) is population density, [𝑥] + = max{𝑥, 0}, and Í Í (𝑖 ) 𝑧(𝑡) = 𝑚 𝑖=1 𝛼𝑖 Δ𝑣 (𝑡), with 𝑖 𝛼𝑖 = 1, is the weighted age of the 𝑚 most recently received video frames, where Δ𝑣(𝑖 ) (𝑡) is the age of the 𝑖-th most recent frame.

(b)

(a)

(c) Most recent

UAV u

UAV v

ru(t)

h(1)(t) α1

Bv(t)

Bu(t) x̂ u(t)

duv(t)

Older

…

Bu(t)

x̂ v(t)

h(2)(t) α2

…

h(w)(t) αw

Weighted coherence-window staleness rv(t)

zu(t)

Population density Low

High

0

Hnom

Higher risk

Figure 1: The same staleness can induce different operational consequences. (A) Navigation risk increases as position uncertainty consumes the separation margin between neighboring UAVs. (B) Ground risk depends on the population exposed within the UAV’s uncertainty region, so the same AoI can imply different risk across locations. (C) Video-coherence risk captures the weighted staleness of the most recent completed frames and increases once it exceeds the nominal coherence threshold. The three components are normalized at their corresponding operational boundaries.

The navigation component measures a separation deficit, while the ground component measures population exposure within the uncertainty region; neither is a collision or casualty probability. The video measures excess staleness relative to a task-specific coherence requirement. It is relevant only when timely video supports the intended operation. Operational reference levels. The three components have different physical units. Following the normalization approach used for risk-aware UAV communications, we scale them by positive reference levels derived from declared operational limits. For example, the ground reference can be defined as 𝑅¯𝐺 = 𝜌 max 𝜋𝑟𝜏2 , where 𝑟𝜏 = 𝑟 min + 𝑣 max𝜏𝑠max and 𝜏𝑠max is the declared maximum status-update age. Similarly, 𝑅¯𝑉 = 𝜏𝑣max − 𝐻 nom defines the video reference at its maximum tolerable coherence age. The navigation reference is determined by the declared separation margin. The resulting composite proxy is 𝑅(𝑡) = 𝑤 𝑁 𝑅 𝑁 (𝑡) + 𝑤𝐺 𝑅𝐺 (𝑡) + 𝑤𝑉 𝑅𝑉 (𝑡)

(5)

where 𝑤 𝑁 + 𝑤𝐺 + 𝑤𝑉 = 1, with nonnegative weights. Reference levels and weights are fixed when comparing mobility decisions. The level 𝑅 = 1 defines a declared composite reference boundary, not a certified safety threshold. The metric quantifies communication-induced operational consequences rather than accident probability. From risk to handover timing. A handover affects operational risk through the freshness of the information available to the network. The interruption duration depends on whether the transition succeeds normally, experiences a prolonged data-plane outage, or is preceded by failure of the serving link.

André Melo, Yueqi Chen, Igor Kadota, and Lourenço Alves Pereira Jr.

For a candidate execution time 𝜏, let 𝑅(𝑠 | 𝜏) denote the operational-risk trajectory resulting from these possible communication outcomes. We define ∫ 𝑡 +𝐻  𝐶 (𝜏) = E 𝑅(𝑠 | 𝜏) 𝑑𝑠 , (6) 𝑡

where 𝐻 is a common prediction horizon and the expectation is taken over possible outage durations and link-failure outcomes. Candidate decisions are evaluated using the same operational context and outage model. RAM selects

Evento

(b) Alphaville

4000

2000 Shopping Iguatemi ver Tietê Ri 0 0

2000

4000

6000

East [m] 1.0

(c)

10

0.5 0 −10 0.0

0

1000

2000

3000

4000

5000

Position along route A (river) [m] 1.0

(d)

route B: 33 HOs, 8 over exposed ground 20

10

0.5

4 Evaluation 4.1 Does Risk Awareness Improve Mobility Decisions? Experimental setting. We reconstruct a UAV corridor around Alphaville, São Paulo, using licensed 5G NR deployments from Anatel’s Mosaico database [9]. As shown in Fig. 2(a)– (b), the deployment contains 59 base stations and 188 sectors operating at 3.55 GHz within 5 km of the corridor; the simulation uses the 72 sectors within 1.5 km of the routes (59 after pruning those never within 20 dB of the best server). We model aerial propagation using 3GPP urban-macro path loss, aerial line-of-sight probability, antenna azimuth, tilt, gain, correlated shadowing, and inter-cell interference [1, 5]. Route A follows the Tietê river for 5.4 km and exposes little population; route B is the 3.9-km straight line from the mall and crosses residential areas. Six UAVs fly both routes in both directions at 60 m and 14 m/s, with departures spread over 300 s, for three seeds. Ground exposure 𝜌 is the population level set of the corridor; “exposed ground” denotes normalized exposure above 0.3. We compare conventional A3 with hysteresis values of 3, 6, 9, and 12 dB against RAM. All policies experience the same channel realizations and handover-outage outcomes. Outage probabilities and durations are derived from the

SINR [dB]

route A: 50 HOs, 0 over exposed ground 20

0

SINR [dB]

where Tfeasible contains execution times satisfying the existing radio conditions and serving-link viability constraints. RAM does not change the handover procedure or select a different target cell. It retains the A3-selected target and uses operational risk to determine when the transition should occur. The required context may be supplied by aviation services and combined with network measurements, positioning, and analytics [4, 7]. Evaluating Eq. (6) requires predicting where the aircraft will be and what radio conditions it will see; Section 4.1 compares a predictor that knows the trajectory with one that uses only what the network observes.

North [m]

(7)

ρ(s) (norm.)

𝜏 ∈ Tfeasible

ρ(s) (norm.)

𝜏 ∗ = arg min 𝐶 (𝜏),

−10 0.0

0

500

1000

1500

2000

2500

3000

3500

Position along route B (straight) [m] Route A (along the Tietê) Route B Start End

TIM 5G NR site (24 used) sector (azimuth, HPBW) cell boundary, 60 m (model)

Pop. density SINR (model) A3 handover

Figure 2: The Alphaville delivery corridor. (a) Geographic overview showing the two routes, licensed TIM 5G NR sites, and ground-exposure levels. (b) Local view of the licensed sectors, including antenna azimuths, beamwidths, and best-server cell boundaries at 60 m predicted by the channel model. (c)–(d) Ground exposure and predicted SINR along the river route A and straight route B, with serving-cell transitions under a 3-dB A3 trigger.

commercial 5G traces examined in Section 2: a regular handover interrupts delivery for 21.4 ms and, with probability 0.081, becomes a post-handover outage drawn from the 133 episodes measured at 80 m (median 2.2 s). The proxy uses equal weights, 𝑣 max = 14 m/s, the other aircraft as hazards and 𝑅¯𝑉 = 𝐻 nom ; RAM uses a 30-s horizon on a 1-s grid. RAM retains the A3-selected target cell, but shifts handover execution within a short prediction horizon to reduce

What Is the Cost of a Handover Interruption? Rethinking 5G Mobility for Aerial Vehicles

time with R > 1 [%]

12 dB

3.0 2.5

A3, 3 dB

2.0 9 dB 6 dB

1.5 RAM

0

2

4

6

8

10

handovers per km

Figure 3: Time above the risk boundary vs. handovers per km. Dashed: the A3 hysteresis sweep (3–12 dB); black circle: RAM-T; diamond: RAM-R. Fewer handovers first reduce and then increase risk. Error bars span the three seeds. Table 1: Mobility decisions in the São Paulo deployment. 𝑅 > 1 denotes time above the declared operational-risk boundary. Results use three simulation seeds. Policy A3, 3 dB A3, 6 dB A3, 9 dB A3, 12 dB RAM-T RAM-R

HO/km

𝑅¯ 𝑅 > 1 SNR

9.7 0.289 5.2 0.178 3.6 0.171 2.6 0.221 2.9 0.160 3.2 0.198

2.6% 1.4% 1.6% 3.1% 1.3% 1.4%

3.8 3.4 2.7 1.9 3.4 3.3

expected operational risk while preserving serving-link viability. We evaluate two predictors of 𝑅(𝑠 | 𝜏). RAM-T (trajectory) uses the planned path and the simulated future channel: an upper bound on what context can deliver. RAM-R (reachability) uses only what the network observes: the last reported position and heading, exposure averaged over a 20◦ cone of reachable positions, and a shadow-free radio map corrected by the current measurement with a 5-s decay. Finding 1: Fewer handovers are not necessarily safer. Table 1 and Fig. 3 reveal a non-monotonic relationship between handover frequency and operational risk. Increasing A3 hysteresis from 3 to 9 dB reduces handovers from 9.7 to 3.6 per km and lowers mean operational-risk proxy from 0.289 to 0.171. However, increasing hysteresis further to 12 dB raises risk to 0.221, despite reducing handovers to only 2.6 per km. The reason is the mobility trade-off identified in Section 2: suppressing unnecessary transitions reduces interruption exposure, but holding the serving connection too long allows radio quality to deteriorate. At 12 dB, average serving SNR falls to 1.9 dB, and time above the declared risk boundary increases to 3.1%. Finding 2: Risk awareness changes which handovers are executed. RAM-T achieves the lowest mean operationalrisk proxy (0.160), a 45% reduction relative to A3 at 3 dB,

in each of the three seeds and not only on average, while keeping the serving SNR within 0.4 dB of the 3-dB trigger. More importantly, the improvement does not come simply from suppressing handovers. RAM-T performs 2.9 handovers/km, close to the 12-dB A3 configuration, but achieves lower risk and substantially better serving-link quality. The difference is which transitions are executed. Along the crowd-exposed route B, conventional A3 at 3 dB executes 94 handovers over exposed ground per run, compared with 30 under RAM-T; the associated data-plane mute time over exposed ground decreases from approximately 23 s to 12 s. These are handovers RAM-T drops, not moves: the share of transitions over exposed ground is 49% for both policies, and no executed handover is shifted by more than 0.4 s. Hysteresis suppresses transitions by signal margin; RAM drops those whose predicted advantage does not outlast the horizon, including all 110 ping-pong transitions per run of the 3-dB trigger, and keeps the rest at the serving quality of the 3-dB trigger. Finding 3: The gain does not require knowing the trajectory. RAM-R reproduces the mechanism with network information only: 3.2 handovers/km, no ping-pong, the same 2.7 outages over exposed ground per run as RAM-T, 1.4% of time above the boundary against 1.3%, and serving SNR within 0.1 dB. Its mean proxy (0.198) differs from RAM-T only in one seed, where a single long outage shared by all policies falls worse for RAM-R; the other two seeds coincide. Not knowing the trajectory costs little here. What does this tell us? Conventional mobility optimization already balances the radio costs of switching and staying. Our results suggest that operational risk provides an additional dimension for navigating that trade-off. Rather than making every handover faster or simply reducing their number, RAM prioritizes when interruptions can be better tolerated by the aerial operation.

5

Discussion and Open Questions

What should mobility optimize? Existing mobility mechanisms are designed to maintain connectivity and reduce service disruption. Our results do not challenge these objectives. Instead, they expose a limitation of decisions based only on radio performance: two handovers with similar communication costs may create different levels of operational risk. Adjusting A3 hysteresis changes how frequently handovers occur, but cannot explicitly distinguish where an interruption is safer to tolerate. RAM introduces operational risk as an additional decision metric, without changing how handovers are executed. Can a more intelligent RAN become risk-aware? Advances in 5G positioning, integrated sensing and communication (ISAC), and network analytics are expanding the

André Melo, Yueqi Chen, Igor Kadota, and Lourenço Alves Pereira Jr.

information available to cellular networks. Combined with flight information and airspace-management services, these capabilities could allow the RAN to estimate the operational context of individual aircraft. However, knowing more about the environment does not necessarily lead to safer mobility decisions. The network still needs a way to translate this information into a metric its mobility mechanisms can use. RAM provides such a connection by mapping information staleness, position uncertainty, and operational context into a common risk proxy. This suggests a path toward risk-aware mobility in which existing radio measurements determine feasible transitions, while operational risk informs their timing. Such decisions could be integrated into radio resource management or supported by an external controller, without introducing new handover procedures. How to expose and update this risk information within the RAN remains open. What are the limits of risk awareness? Our metric is an operational-risk proxy, not an estimate of accident probability. Its weights and reference levels depend on the operation, and its accuracy depends on the freshness and quality of the available context. Likewise, our evaluation uses simulated radio conditions and commercial trace-derived outages in one urban deployment. These results show that risk awareness can improve decisions under the modeled assumptions, but do not establish a measured improvement in flight safety or performance in a commercial RAN. In this corridor the exposure varies little within the 30-s horizon, so RAM changes which handovers happen rather than where; moving them requires an exposure gradient the aircraft can reach in time. Who should define acceptable risk? Cellular operators can characterize radio reliability, but aviation services and aircraft operators define the safety requirements of their operations. Should the RAN independently estimate operational risk, or should aviation services expose risk limits that mobility mechanisms can directly use? How should the network react when this information is uncertain, unavailable, or changes during flight? These questions extend beyond aerial vehicles to other autonomous systems that depend on cellular connectivity. As networks become increasingly capable of sensing and understanding their environment, mobility decisions may need to evolve accordingly. The question is no longer only when a handover can succeed, but whether the network can account for the operational risk of executing it. Can operational risk inform which cells should serve UAVs? Risk awareness need not be limited to deciding when a handover occurs. The same operational-risk metric could inform which cells are eligible to serve aerial vehicles along a corridor or within a geographic region. A cell offering stronger radio coverage may still lead to frequent interruptions in operationally sensitive areas, while a different

serving-cell configuration may provide lower overall risk. This raises a broader question: should aerial connectivity be planned solely around radio coverage, or should network operators and aviation services jointly identify serving-cell configurations that satisfy operational requirements? Such an approach could extend risk awareness from individual mobility decisions to aerial service planning, without requiring changes to the underlying handover procedures.

References [1] 3GPP. 2017. TR 36.777: Enhanced LTE Support for Aerial Vehicles. [2] 3GPP. 2019. TS 36.331: E-UTRA; Radio Resource Control (RRC); Protocol Specification, Release 15. Technical Report V15.7.0. [3] 3GPP. 2021. TS 38.300: NR and NG-RAN. Technical Report. [4] 3GPP. 2024. 5G System Location Services. TS 23.273. [5] 3GPP. 2024. TR 38.901: Study on Channel Model for Frequencies from 0.5 to 100 GHz. Technical Report. [6] 3GPP. 2024. TS 38.331: NR; Radio Resource Control (RRC); Protocol Specification, Release 18. Technical Report V18.3.0. [7] 3GPP. 2025. Architecture enhancements for 5G System (5GS) to support network data analytics services. TS 23.288. Release 19. [8] 3GPP. 2026. TS 38.300: NR and NG-RAN Overall Description; Stage-2, Release 18. Technical Report V18.8.0. [9] Agência Nacional de Telecomunicações (Anatel). 2026. Mosaico: Public Database of Licensed Telecommunications Stations. Brazil. [10] Amin Azari, Fayezeh Ghavimi, Mustafa Ozger, Riku Jäntti, and Cicek Cavdar. 2020. Machine Learning Assisted Handover and Resource Management for Cellular Connected Drones. In 2020 IEEE VTC. [11] Jingwen Bai, Shu-Ping Yeh, Feng Xue, and Shilpa Talwar. 2019. RouteAware Handover Enhancement for Drones in Cellular Networks. In 2019 IEEE GLOBECOM. Meituan First Nationwide Low-Altitude [12] China Daily. 2025. Operating License. https://www.chinadaily.com.cn/a/202504/22/ WS6807752fa3104d9fd3820e14.html. [13] DECEA. 2022. Concepção Operacional UTM Nacional. Technical Report DCA 351-6. Departamento de Controle do Espaço Aéreo. https: //publicacoes.decea.mil.br/publicacao/dca-351-6 [14] Sebastian Euler, Helka-Liina Määttänen, Xingqin Lin, Zhenhua Zou, Mattias Bergström, and Jonas Sedin. 2019. Mobility Support for Cellular Connected Unmanned Aerial Vehicles: Performance and Analysis. In 2019 IEEE WCNC. doi:10.1109/WCNC.2019.8885820 [15] Federal Aviation Administration. 2020. Unmanned Aircraft System (UAS) Traffic Management (UTM) Concept of Operations, Version 2.0. Technical Report. FAA. [16] Jingqi Huang, Yanbing Liu, and Chunyi Peng. 2024. Uncovering Missed 5G Performance in the Sky. In HotMobile. [17] JARUS. 2024. Guidelines on Specific Operations Risk Assessment. Technical Report. Joint Authorities for Rulemaking on Unmanned Systems. [18] Joint Authorities for Rulemaking on Unmanned Systems (JARUS). 2024. Guidelines on SORA: Annex F. Technical Report. https://jarus-rpas.org/wp-content/uploads/2024/06/SORA-v2.5Annex-F-Release.JAR_doc_29pdf.pdf [19] KrASIA. 2024. Meituan’s Drone Delivery Debuts in Dubai, Eyeing a Future in Last-Mile Logistics. https://kr-asia.com/meituans-dronedelivery-debuts-in-dubai-eyeing-a-future-in-last-mile-logistics. [20] Xinzhe Liu, Jianer Zhou, Xiaoyong Ni, Ke Luo, Zhenyu Li, Xiaofeng Tao, and Weichao Li. 2026. Unveiling Low-Altitude 5G Performance: Linking Key Influencing Factors with UAV Flight Parameters. In Proceedings of the ACM SIGCOMM 2026 Conference. ACM, 317–333.

What Is the Cost of a Handover Interruption? Rethinking 5G Mobility for Aerial Vehicles

[21] André Melo, Yueqi Chen, Lourenço Pereira Jr., and Igor Kadota. 2026. Who Can Safely Wait? Risk-Aware 5G Scheduling for Low-Altitude UAV Operation. arXiv:2610.07101 [cs.NI] doi:10.48550/arXiv.2610. 07101 [22] OpenAirInterface Software Alliance. 2026. OAI 5G Software. [23] SESAR Joint Undertaking. 2023. U-space Concept of Operations (ConOps), Fourth Edition. Technical Report. SESAR Joint Undertaking. [24] sUAS News. 2026. Delivery Drones Cleared for Flight over Populated Cities in Brazil. https://www.suasnews.com/2026/03/delivery-dronescleared-for-flight-over-populated-cities-in-brazil/.

[25] UK Civil Aviation Authority. 2025. Future of Flight: BVLOS Roadmap. Civil Aviation Publication CAP 3182. UK Civil Aviation Authority. https://www.caa.co.uk/publication/download/26211 [26] UK Civil Aviation Authority. 2026. Conops: Future Air Traffic Management. Technical Report. https://www.caa.co.uk/publication/ download/29467 [27] U.S. Department of Transportation. 2025. Advanced Air Mobility National Strategy 2025. Technical Report. [28] U.S. DoT. 2025. Advanced Air Mobility Comprehensive Plan 2025. Technical Report. U.S. Department of Transportation.

Related documents

• What Is the Cost of a Handover Interruption? Rethinking 5G Mobility for Aerial Vehicles#1342599
• Learning the LoS Skyline from LEO Satellite Observations for Proactive Handover#423916
• Sensing Assisted Satellite Backhaul with FBL UL and Broadcast DL for Massive IoT#1076244
• Memory reactivation levels remain unaffected by anticipated interference#976805
• Memory reactivation levels remain unaffected by anticipated interference#976736