High-Altitude Platforms (HAPs)

• High-Altitude Platforms (HAPs) are stratospheric stations operating at 17–25 km, providing communication, computing, and sensing services across diverse regions.
• They integrate unmanned airships, balloons, and solar-powered aircraft with advanced avionics, phased-array antennas, and onboard data centers to ensure resilient connectivity.
• HAPs enable scalable, sustainable networks by leveraging green computing through passive cooling and solar harvesting, significantly reducing energy costs.

A High-Altitude Platform (HAP) is a stratospheric airborne station—typically an unmanned airship, balloon, or fixed-wing vehicle—operating at altitudes from 17 km to 25 km, designed to deliver communication, computing, sensing, or data services over a wide terrestrial footprint. HAPs bridge the gap between terrestrial networks and satellite systems, enabling low-latency, high-capacity, energy-efficient service delivery across urban, rural, and remote regions. Recent advances leverage the stratosphere’s cold, high-solar-irradiance environment to enable green computing, resilient connectivity, and rapid deployment in both routine and exigent scenarios.

1. Physical Principles and System Architecture

HAPs operate above the troposphere (typically at 17–22 km), exploiting near-constant, low ambient temperatures (≈218 K), weak horizontal winds, and a clear line-of-sight to ground and air nodes. Principal HAP architectures include aerostatic airships (helium/hydrogen lift, payloads up to several hundred kg), super-pressure or zero-pressure balloons (passive drift, lower payload), and fixed-wing solar aircraft (tight orbits, moderate payload) (Kurt et al., 2020, Çoğay et al., 19 Feb 2026, Elkhazraji et al., 8 Nov 2025).

A canonical HAP architecture integrates:

• Flight Control and Power: Autonomous avionics, station-keeping through electric propellers and solar PV arrays (8,000 m² area, η_PV ~ 0.4), Li–S battery storage (~2 kWh/kg), and efficient propulsion (η_prop ~ 0.8) (Abderrahim et al., 2023).
• Communications Payload: Phased-array antennas (256–1024 elements for mMIMO), massive-MIMO for multi-beam 3D coverage, onboard baseband units, and high-gain FSO/mmWave/THz transceivers for air-to-ground, air-to-satellite, and inter-HAP links (Kement et al., 2022, Elkhazraji et al., 8 Nov 2025, Abbasi et al., 2023).
• Edge Computing/Data Center: Airborne rack-scale servers (each ≈11 kg; total payload ≤450 kg per platform), passive thermal management using stratospheric cold (no active chillers), onboard AI/GPU for edge inference, caching, and control (Abderrahim et al., 2023, Mershad et al., 2021, Huang et al., 23 Feb 2026).
• Auxiliary Subsystems: Real-time TT&C, energy management (MPPT, SOC), and structural modules (helium envelope, gondola, redundancy for fault tolerance) (Hjaiji et al., 16 Nov 2025).

For multi-mode operational flexibility, HAPs may dynamically switch between super macro base station (HAPS-SMBS), relay (HAPS-RS), and reconfigurable intelligent surface (HAPS-RIS) modes to trade off between coverage, energy, and connectivity requirements (Alfattani et al., 2022).

2. Stratospheric Environmental Constraints and Energy Model

The stratosphere’s thermal regime is nearly isothermal (T(h) ≈ 218 K for 11–20 km altitude), making passive cooling feasible. Energy harvesting is dominated by PV output:

• Instantaneous solar power, P_solar(t) = η_PV·A_PV·G(t), where G(t) is solar irradiance depending on latitude and day-of-year.
• Total daily harvested energy,

$$E_{\text{solar,day}} = η_{\text{PV}}\,A_{\text{PV}}\,\int_0^{t_{\text{sunset}}} G(t)\,dt$$

(Abderrahim et al., 2023). Station-keeping and payload operation are conditioned on the “flying condition” (E_{solar,day} ≥ E_{propulsion} + E_{payload} + E_{comm} over 24 h).

Energy consumption is partitioned as E_{total} = E_{propulsion} + E_{payload} + E_{comm} (Mershad et al., 2021).

• Propulsion powers (flight speed V, altitude H, circular bank angle ζ) are optimized to minimize night-time draw, e.g., a closed-form minimum of P_{prop} with typical values ≲1 kW for optimized trajectories (Javed et al., 2022).
• Communication payloads may consume up to ≈1 kW in active (SMBS) mode but only tens of watts in RIS mode, enabling major loiter time extension under passive operation (Alfattani et al., 2022).

The number of deployed servers/active payload is capped by the lift constraint: N·W_server ≤ W_payload_max (e.g., N ≤ 40 servers for W_server = 11 kg, W_payload_max = 450 kg) (Abderrahim et al., 2023).

3. Energy Efficiency, Sustainability, and Green Computing

HAPs achieve compelling energy and sustainability metrics by:

• Eliminating active cooling: Stratospheric ambient ≈–55°C eliminates the need for compression-based cooling; per-server P_cool,HAP ≈ 0 (Abderrahim et al., 2023).
• Solar harvesting: Up to ≈8 MWh/day per HAP, adequate for server farm plus propulsion (Abderrahim et al., 2023).
• Electricity cost reduction: For a typical scenario, one HAP yields a ≈12% electricity-cost reduction (C_1 ≈ 0.88·C₀); four-HAP constellations reach ≈36% savings, driven by both cooling elimination and solar self-sufficiency (Abderrahim et al., 2023).
• Urban offload: In dense urban settings, deploying a HAP-SMBS to supplement (and partially replace) terrestrial densification substantially increases capacity utilization (up to ~71% from 31%) while halving power demand (≈140 kW vs. 314 kW) (Kement et al., 2022).

Cell-switching frameworks that leverage HAPS to dynamically deactivate terrestrial BSs while meeting outage-based QoS constraints demonstrate up to 77% energy savings at low load (and ~40% at high load) in vertical HetNet scenarios (Salamatmoghadasi et al., 15 Jan 2026).

4. Scalability, Network Integration, and Performance Bottlenecks

Scalability constraints arise from both wireless link capacity and scheduling:

• Air-to-Ground Link: C = BW·log₂(1 + SNR). With BW = 100 MHz and f_c = 31 GHz, the practical offload rate is capped by link SNR and bottlenecked at high load (Abderrahim et al., 2023).
• Task Queuing: Each airborne server can be modeled as M/M/1 with utilization ρ = λ/μ, implying queueing delay D = 1/(μ–λ). High utilization (ρ→1) minimizes solar energy wastage but increases delay (Abderrahim et al., 2023).
• Multi-HAP Constellation: Aggregates capacity, distributes workload, and mitigates congestion and single-point failure risks. End-to-end delay comprises queueing, transmission (RTT_ground–HAP ≈ 2–5 ms), and (if relevant) inter-HAP relay hops (Abderrahim et al., 2023).
• Management Agility: Effective offloading policies dynamically allocate workloads between HAP and terrestrial DC, balance server utilization in the 0.7–0.9 range, and adjust for available PV power and admission thresholds (Abderrahim et al., 2023).

Performance evaluation indicates that for short tasks (≤1 ms), offloading to HAP results in much lower delay than terrestrial queueing; for longer tasks, overhead may outweigh advantages unless careful admission control is employed (Abderrahim et al., 2023). Seasonally, cost savings are lowest at winter solstice (minimum G(t)), highest in summer.

5. Management, Orchestration, and Reliability

To maximize HAP utility as a flying data center:

• Workload Offloading: Pseudocode-driven adaptive split based on instantaneous power surplus and average incoming load, with dynamic adjustment for target server utilization (Abderrahim et al., 2023).
• Resource Allocation: Multi-objective optimization balancing total energy and end-to-end delay, subject to air-to-ground link capacity and PV generation constraints:

$$\min_{x_t}\,\alpha\,E_{total}(x) + \beta\,D_{total}(x)\quad \text{s.t.}\, E_{solar,day} \geq E_{consumed}(x),\, \lambda_{offload}(t) \leq C_{link}$$

• Redundancy and Failover: Fault-tolerance is achieved by dual-replication of critical data between paired HAPs, with automatic load shifting upon HAP failure (Abderrahim et al., 2023).
• Distributed Learning: Each HAP locally trains models of solar and traffic patterns, exchanging updates to maintain globally optimized policies (Abderrahim et al., 2023).

6. Sustainability, Implementation Outcomes, and Outlook

Numerical results demonstrate that HAP-enabled data centers and network nodes deliver measurable environmental and operational benefits:

• Electricity cost savings: 12% for a single HAP, 36% for four-node constellations (Abderrahim et al., 2023).
• Operational feasibility: With 40 airborne servers, the solar energy budget supports ρ ≈ 0.8 utilization most days; over-provisioning introduces only modest risk (Abderrahim et al., 2023).
• Scalability: Multi-HAP deployment achieves sustainable green operation as both computation load and communication demand scale (Abderrahim et al., 2023).
• Management flexibility: Adaptive algorithms ensure performance is robust to task size, daily solar input, and link variability.

Sustainability is underpinned by leveraging the stratosphere’s cold to eliminate cooling overhead, while large-scale PV arrays (e.g., 8,000 m²) at altitude deliver high energy yield. As a consequence, HAP-based platforms provide grid-independent, scalable, and green computational services (Abderrahim et al., 2023). The technology roadmap emphasizes integration of HAPs in three-tier network architectures (terrestrial–stratospheric–satellite) for broader regional/global impact, further amplifying energy, cost, and sustainability advantages.

---

References:

• (Abderrahim et al., 2023) How to Leverage High Altitude Platforms in Green Computing?
• (Kement et al., 2022) Sustaining Dynamic Traffic in Dense Urban Areas with High Altitude Platform Stations (HAPS)
• (Salamatmoghadasi et al., 15 Jan 2026) Sustainable Vertical Heterogeneous Networks: A Cell Switching Approach with High Altitude Platform Station
• (Mershad et al., 2021) Cloud-Enabled High-Altitude Platform Systems: Challenges and Opportunities
• (Alfattani et al., 2022) Multi-Mode High Altitude Platform Stations (HAPS) for Next Generation Wireless Networks
• (Javed et al., 2022) An Interdisciplinary Approach to Optimal Communication and Flight Operation of High-Altitude Long-Endurance Platforms