Balloon-Based Aerial Relays: Optimization & Applications

• Balloon-based aerial relays are elevated communication platforms—ranging from low-altitude tethered systems to stratospheric balloons—that provide access, backhaul, and optical relay functions in infrastructure-sparse environments.
• The literature documents diverse architectures, including Helikite-based LTE-A access nodes and optical quantum relay chains, each balancing persistence, payload, and line-of-sight constraints.
• Optimization studies highlight joint design challenges in power allocation, relay placement, and beam control, emphasizing that backhaul limitations and mechanical constraints are key to network performance.

Searching arXiv for recent and foundational work on balloon-based aerial relays to ground the encyclopedia entry. arXiv search query: "balloon aerial relay tethered balloon HAP quantum network aerial relays" Balloon-based aerial relays are communication nodes carried by tethered balloons, Helikites, or stratospheric balloons and used to provide access, backhaul, or optical relay functionality when terrestrial infrastructure is absent, damaged, congested, or insufficient. Across the literature, they occupy markedly different operating regimes: low-altitude tethered systems can carry RF access equipment or relay traffic for LTE-A and UAV-assisted networks, while stratospheric balloons at roughly $18$–$30$ km can support long line-of-sight free-space optical links and, in recent quantum-network proposals, form relay chains over continental or global distances (Chandrasekharan et al., 2016). A unifying feature is that balloon platforms trade strong persistence, elevated line of sight, and reduced dependence on ground infrastructure against platform drift, payload, power, and pointing constraints (Jha et al., 29 May 2025).

1. Architectural forms and operating regimes

The literature describes several distinct balloon-relay architectures. In infrastructure-sparse terrestrial wireless networks, tethered balloons (TBs) can act as persistent aerial backhaul nodes with fiber connectivity to the core network, while multiple UAVs operate below them as access relays serving ground users; in that architecture, TBs are fixed at higher altitude and a central controller at one TB manages resource assignment and control signaling (Alzidaneen et al., 2019). In integrated satellite–air–ground systems, TBs occupy the tropospheric layer beneath stratospheric HAPs and above GBSs, bridging ground users to HAP backhaul and complementing both terrestrial and satellite connectivity (Alsharoa et al., 2021).

A second architectural family uses tethered aerostats as access nodes. The ABSOLUTE Helikite implementation mounted the RF aerial segment aloft while keeping the eNB baseband unit, EPC functions, routing, and power-related infrastructure on the ground, linked through a twin optical fiber integrated with the tether (Chandrasekharan et al., 2016). In that system, the aerial segment consisted of a Remote Radio Head (RRH), antennas, batteries, a pendulum mount, and a waterproof case, while backhaul was provided through a deployable Ka-band satellite terminal, with Wi-Fi also considered as an alternative (Chandrasekharan et al., 2016).

A third family appears in optical and quantum networking. Stratospheric balloons at approximately $24$ km are modeled as passive optical relays that interconnect ground repeater stations through balloon-to-balloon free-space links, with all quantum memories, entangled photon sources, and single-photon detectors kept on the ground (Liu et al., 21 Jul 2025). A broader review places balloon-based relays within a layered quantum internet spanning ground fiber, free-space links, UAVs/HAPS, and LEO/GEO satellites, and states that balloons can serve either as trusted nodes or as trustless entanglement-swapping relays (Jha et al., 29 May 2025).

A compact taxonomy of roles follows directly from these studies.

Regime	Typical role	Representative source
Low-altitude tethered balloon / Helikite	RF access node or tethered relay with ground-connected baseband	(Chandrasekharan et al., 2016)
Tropospheric tethered balloon in integrated networks	Relay/access node between users, GBSs, and HAPs	(Alsharoa et al., 2021)
Higher-altitude tethered balloon with UAV layer	Persistent backhaul node for multiple UAV access relays	(Alzidaneen et al., 2019)
Stratospheric balloon chain	Passive optical relay backbone for quantum networking	(Liu et al., 21 Jul 2025)

This diversity explains why the term “balloon-based aerial relay” spans markedly different altitude, payload, and protocol assumptions. A plausible implication is that comparisons across papers are meaningful only when the altitude regime, relay function, and duplexing model are made explicit.

2. Wireless channel structure, relay models, and bottlenecks

In RF relay architectures, the central design problem is not merely coverage but coupled access–backhaul performance. In the TB–UAV system, UAV–user access uses OFDMA resource blocks with $B = 180$ kHz and orthogonal allocation, while TB–UAV backhaul operates on orthogonal spectrum to avoid loop interference (Alzidaneen et al., 2019). The access gain is modeled from an air-to-ground LoS/NLoS composite, with LoS probability

$$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$

and the end-to-end throughput of UAV $l$ is constrained by

$$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$

That minimum coupling is the paper’s central bottleneck: increasing access power improves throughput only until the backhaul bound is reached, and increasing backhaul bandwidth helps only until access becomes limiting (Alzidaneen et al., 2019).

Related work on cooperative satellite–aerial–terrestrial links models the balloon or HAP relay as a decode-and-forward node between a satellite transmitter and terrestrial receivers, with Rician fading on the relay–destination hop and Shadowed-Rician fading on the satellite–relay hop (Pan et al., 2020). There the relay’s ground coverage is a disk of radius $L$, the relay altitude is $H_1$, and the relay–destination slant range is $d=\sqrt{H_1^2+r^2}$. Coverage probability degrades as $30$0 or $30$1 increases, while the satellite–relay outage is comparatively insensitive to $30$2 because satellite-range variation dominates the geometry (Pan et al., 2020).

Optical and quantum studies adopt a different channel decomposition. The free-space balloon-network model factors channel efficiency as

$30$3

with turbulence represented through the Hufnagel–Valley model for $30$4, Rytov variance, beam wander, and AO-corrected fiber coupling (Karakosta-Amarantidou et al., 2024). In the global quantum-network proposal, Gaussian propagation, aperture clipping, adaptive optics, and residual wind drift are combined segment by segment, and total transmittance multiplies across the balloon chain as

$30$5

These optical models make clear that balloon-based relay performance depends as much on beam geometry and wavefront control as on nominal path length (Liu et al., 21 Jul 2025).

A recurrent misconception is that elevated relays are automatically “coverage-limited” but not “backhaul-limited.” The cited wireless and optical works do not support that view. In the TB–UAV RF architecture, end-to-end rate saturates because of backhaul; in the integrated HAP–TB system, throughput saturates due to limited backhaul capacity; and in quantum architectures, long-distance rates are dominated by accumulated channel loss and BSM success rather than by geometric visibility alone (Alzidaneen et al., 2019).

3. Optimization, placement, and relay signal processing

Balloon-based relay design is repeatedly cast as a joint optimization problem. In the TB–UAV network, association is first solved as an integer linear program over binary variables $30$6 and $30$7, maximizing $30$8 under access, backhaul, and association constraints (Alzidaneen et al., 2019). With association fixed, UAV transmit powers are optimized through a convex formulation because each access rate

$30$9

is concave in power. KKT stationarity yields the water-filling-like solution

$24$0

Placement is then handled with a shrink-and-realign modified recursive random search, which empirically converges within approximately $24$1–$24$2 while-loop iterations (Alzidaneen et al., 2019).

The vertically integrated satellite–HAP–TB–GBS architecture also emphasizes joint optimization, though at a more abstract level. Access links are RF-only; backhaul uses hybrid FSO/RF; and the resource-management problem includes user associations, power and bandwidth allocation, HAP and TB placement, FSO alignment, and energy constraints for renewable-energy-powered TBs (Alsharoa et al., 2021). The paper reports that throughput gains are ultimately controlled by backhaul bandwidth, particularly HAP–gateway/satellite and HAP–TB/GBS links, so adding FSO capacity alleviates the dominant bottleneck (Alsharoa et al., 2021).

A different optimization objective appears in the HAP-drone MIMO X-network with a tethered balloon relay. There, the balloon relay uses decode-and-forward and half-duplex operation to realize interference alignment without CSIT at the HAPs, provided the relay has $24$3 antennas (Sudheesh et al., 2017). The resulting sum-rate is

$24$4

and simulations show the existence of an optimal balloon altitude that balances the two hops. When $24$5 and $24$6, the paper states that maximum capacity occurs when

$24$7

which places the optimal relay at the geometric midpoint if $24$8 (Sudheesh et al., 2017).

These results collectively indicate that balloon-relay optimization is rarely separable by layer. A plausible implication is that relay altitude, power, association, antenna configuration, and even beam-control order should be co-designed rather than tuned independently.

4. Platform embodiments and terrestrial broadband deployments

The most concrete terrestrial implementation in the cited material is the Helikite-based LTE-A aerial base station. A Helikite combines a helium-filled aerostat with a kite aerofoil, using both wind lift and helium lift; in the ABSOLUTE implementation, the AeNB was engineered to operate at altitudes up to $24$9 m, with approximately $B = 180$0 hours autonomy at $B = 180$1 m determined by battery capacity and RRH consumption (Chandrasekharan et al., 2016). The demonstrator used a $B = 180$2 Desert Star Helikite, a compact SDR-based RRH supporting up to $B = 180$3 GHz with $B = 180$4–$B = 180$5 MHz signal bandwidth, and lightweight helix antennas shaped for quasi-uniform ground illumination at $B = 180$6 GHz (Chandrasekharan et al., 2016).

Field measurements were performed with RRH transmit power set to $B = 180$7 dBm and Helikite altitude $B = 180$8 m for initial validation. Reported outcomes include smartphone maximum RSRP at the LAP site of $B = 180$9 dBm, maximum distance with ping of $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$0 m, and dongle maximum distance with ping of $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$1 m with minimum RSRP of $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$2 dBm and minimum INIR of $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$3 dB (Chandrasekharan et al., 2016). The paper also presents the standard geometric and path-loss relations

$$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$4

and emphasizes the familiar altitude trade-off: increasing $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$5 raises elevation angle and LoS probability but also increases slant range and free-space loss (Chandrasekharan et al., 2016).

A more recent balloon-derived robotic platform, BEAVIS, addresses a different terrestrial relay problem: how to combine loiter time and maneuverability. BEAVIS merges passive lift with multirotor-like planar and yaw control, using four horizontally oriented rotors and a nonlinear controller that exploits pressure regulation beneath the balloon envelope to induce vertical motion without a vertical propulsor (Sharma et al., 2023). The platform measured $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$6 m/s planar speed along axes, up to $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$7 m/s diagonally, yaw up to $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$8 deg/s, and up to $$p^{\mathrm{LoS}}_{lu} = \frac{1}{1 + c_1 \exp(-c_2[\theta_{lu}-c_1])},$$9 increased lifetime relative to a Crazyflie 2.1 under duty-cycling, while supporting rotor fault detection and controlled descent under single-rotor failure (Sharma et al., 2023).

The disaster-relief literature also includes a visible-light variant. LiBNet proposes lightweight balloons carrying Philips Li-Fi hardware, infrared uplink receivers, D-Link wireless access points for inter-balloon communications, and motion/position sensors, with balloons modeled as a homogeneous Poisson point process and mean co-channel interference derived in both $l$0D and $l$1D (Surampudi et al., 2017). In that system, the ground-coverage radius under field-of-view constraint is

$l$2

and the interference results explicitly show linear scaling with balloon density $l$3 and dependence on Lambertian order $l$4, altitude $l$5, user offset $l$6, and PD field of view $l$7 (Surampudi et al., 2017).

Together, these terrestrial studies show that balloon-based aerial relays are not restricted to a single hardware philosophy. Tethered aerostats favor persistence and simple integration with ground baseband, while balloon–rotor hybrids favor local agility and precise placement. Neither eliminates the altitude–coverage–payload trade-off; they operationalize it differently.

5. Balloon chains and quantum networking

Quantum-network research has made balloon-based aerial relays a distinct architectural class rather than a variant of conventional HAPS. A review of aerial quantum networks places stratospheric balloons at approximately $l$8–$l$9 km, above most clouds and much of the dense aerosol and turbulence affecting near-ground free-space links, and highlights their long endurance, lower turbulence than the boundary layer, and shorter atmospheric path lengths relative to many satellite or ground-only geometries (Jha et al., 29 May 2025). That review also distinguishes trusted-node use from trustless entanglement-swapping use, and gives practical design ranges such as $$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$0–$$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$1 km balloon-to-balloon line-of-sight paths, $$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$2–$$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$3 cm practical apertures, and pointing stability driven to $$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$4–$$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$5 RMS via gimbals and fine-steering mirrors (Jha et al., 29 May 2025).

The “free-space model for a balloon-based quantum network” develops a hardware-level loss model for balloon-to-ground, ground-to-balloon, and balloon-to-balloon channels at $$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$6 nm, using LOWTRAN for atmospheric transmission, HV turbulence, pointing and tracking efficiency, and AO-corrected single-mode-fiber coupling (Karakosta-Amarantidou et al., 2024). The model reports that for trusted-node QKD sub-links in the studied Italian scenario, the vertical downlink Balloon $$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$7 Qonnector yields $$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$8 kbit/s, the slant balloon $$R_l = \min\!\left(\sum_{u,n}\epsilon_{lu,n}R_{lu,n},\;\sum_m \vartheta_{ml}R_{ml}\right).$$9 ground link at roughly $L$0 yields $L$1 kbit/s, and the horizontal balloon-to-balloon link yields $L$2 kbit/s (Karakosta-Amarantidou et al., 2024). In that analysis, a two-balloon architecture with vertical downlinks plus one horizontal balloon-to-balloon segment outperforms a single midpoint balloon because it avoids large zenith-angle losses (Karakosta-Amarantidou et al., 2024).

The global-scale proposal goes further by replacing satellite backbone segments with a chain of balloons acting as passive optical relays between ground repeater servers (Liu et al., 21 Jul 2025). Its core optical design rules are unusually specific: balloon altitude $L$3 km, optimal spacing approximately $L$4 km, roughly $L$5 balloons per $L$6 km leg, Gaussian waist placement at $L$7 toward the receiver, and AO correcting Zernike modes up to radial order $L$8 at every relay (Liu et al., 21 Jul 2025). Under those assumptions, the optimized chain yields $L$9 or approximately $H_1$0 dB over $H_1$1 km, outperforming satellite-based relays by approximately $H_1$2 dB under the same device parameters (Liu et al., 21 Jul 2025).

The same work embeds the optical backbone into the H4QR hybrid repeater architecture, where metropolitan fiber links connect clients to local servers and balloon-based free-space links interconnect servers via a central node (Liu et al., 21 Jul 2025). Using Eu$H_1$3:Y$H_1$4SiO$H_1$5 memories with $H_1$6, $H_1$7 s, $H_1$8, $H_1$9, SPDC sources at $d=\sqrt{H_1^2+r^2}$0 MHz and $d=\sqrt{H_1^2+r^2}$1, and SNSPDs with $d=\sqrt{H_1^2+r^2}$2, the paper reports sub-Hz entanglement distribution rates over $d=\sqrt{H_1^2+r^2}$3 km and distribution times $d=\sqrt{H_1^2+r^2}$4 s across $d=\sqrt{H_1^2+r^2}$5 km (Liu et al., 21 Jul 2025).

A common misconception is that balloons in quantum networking are simply “cheap satellites.” The cited work does not reduce them to that role. Instead, the claimed advantages arise from a specific combination of lower slant paths, lower turbulence in the stratosphere, larger feasible apertures, continuous line of sight without orbital dynamics, and server-centric tracking in which balloons only track fixed local servers rather than geographically dispersed clients (Liu et al., 21 Jul 2025).

6. Performance limits, operational constraints, and open problems

Across RF and optical contexts, the major operational constraints recur with striking consistency. Station-keeping is weaker for free balloons than for powered HAPS, and stratospheric winds of approximately $d=\sqrt{H_1^2+r^2}$6–$d=\sqrt{H_1^2+r^2}$7 m/s can drive drift footprints; superpressure balloons, payload gimbals, and fine-steering mirrors are presented as mitigation mechanisms rather than complete solutions (Jha et al., 29 May 2025). In the global quantum-relay chain, random position jitter introduces additional penalties of approximately $d=\sqrt{H_1^2+r^2}$8 dB to approximately $d=\sqrt{H_1^2+r^2}$9 dB at $30$00 km relative to the ideal evenly spaced chain (Liu et al., 21 Jul 2025). In the sparse-satellite downlink model, lateral displacement is formalized through a random zenith angle $30$01, shrinking the effective spherical cap and degrading connectivity, SNR coverage, and association delay (Choi, 2024).

Power and payload remain a second dominant constraint. The Helikite implementation identified RF power efficiency and power delivery as the main determinant of endurance, noting future potential for power-over-fiber to improve reliability and efficiency (Chandrasekharan et al., 2016). The quantum-network review points out that SNSPDs require approximately $30$02–$30$03 W cryocoolers and materially increase mass and power budgets; APD-based payloads are lighter but reduce efficiency and daytime margin (Jha et al., 29 May 2025). BEAVIS addresses the endurance side by exploiting passive buoyancy and duty-cycled propulsion, but its reported wind envelope and payload are still those of a light low-altitude platform rather than a stratospheric communications node (Sharma et al., 2023).

A third class of limits concerns interference, fairness, and model idealization. The TB–UAV optimization assumes static users, orthogonal access without intercell interference, and uniform TB backhaul allocation (Alzidaneen et al., 2019). The cooperative satellite–aerial–terrestrial model considers one dominant interferer rather than a PPP field of interferers (Pan et al., 2020). LiBNet derives mean co-channel interference under reuse-1 and explicitly identifies blockage modeling as future work (Surampudi et al., 2017). The integrated satellite–HAP–TB framework discusses linear/convex optimization over OFDMA resources, associations, and energy constraints, but does not provide explicit closed-form RF or FSO link-budget formulas (Alsharoa et al., 2021). In the optical quantum models, daytime background-light treatment is repeatedly deferred or simplified relative to the full loss model (Karakosta-Amarantidou et al., 2024).

These limitations also frame the most credible future directions already identified in the literature. In RF relay systems, natural extensions include fairness, delay, mobility robustness, interference-aware resource allocation, and multi-channel spectrum reuse (Alzidaneen et al., 2019). In hybrid optical/RF architectures, improved FSO alignment, RF/GPS-assisted discovery, and better weather-aware operation are emphasized (Alsharoa et al., 2021). In quantum networking, open problems include background-noise integration, richer collection-efficiency distributions beyond weak-wander approximations, adaptive optics design on constrained platforms, and coupling balloon backbones with memories and repeaters beyond current sub-Hz global regimes (Karakosta-Amarantidou et al., 2024).

The broad picture is therefore technically specific rather than generic: balloon-based aerial relays are best understood as a family of elevated relay architectures whose usefulness depends on precise coupling between platform physics, channel regime, and network layer. The literature does not support a single canonical design; it supports a set of well-defined operating points, from $30$04 m tethered LTE-A access cells and UAV-backhauled TB networks to $30$05 km optical relay chains intended for global entanglement distribution (Chandrasekharan et al., 2016).