Internet of Flying Things (IoFT)

• Internet of Flying Things is a cyber-physical paradigm integrating airborne nodes like drones, balloons, and biological flyers into the IoT for dynamic sensing and edge computing.
• Research in IoFT emphasizes multi-tier architectures, trajectory optimization, and cross-layer protocols to enhance coverage, energy efficiency, and real-time analytics.
• IoFT supports critical applications including disaster response, precision agriculture, and industrial automation, leveraging AI-native networking and robust cybersecurity measures.

The Internet of Flying Things (IoFT) is a cyber-physical paradigm that integrates fleets of airborne nodes—ranging from UAVs and drones to high-altitude balloons and biologically powered flyers—into the broader Internet of Things (IoT) and Internet of Everything (IoE) fabric. IoFT enables these aerial elements to act as mobile sensing, communication, computing, and actuation platforms, pushing dynamic “edge intelligence” into 3D airspace. It encompasses architectural compositions, stochastic mobility, resource constraints, multi-modal communications, cross-layer protocols, and robust security requirements uniquely determined by the airborne context. IoFT supports applications such as on-demand aerial mesh connectivity, airborne fog computing, disaster response, precision agriculture, and security-critical surveillance, and is underpinned by AI-native networking and strict cyber-physical guarantees.

1. Architectural Foundations and System Models

IoFT system architecture is typically multi-tiered and heterogeneous. The core tiers comprise:

• UAV/Drone Layer: Deploys rotary/fixed-wing UAVs, autonomous drones, and, in a broader sense, living flyers (e.g., insects equipped with microelectronics (Iyer et al., 2018)). These nodes provide sensing, relay, and edge-processing capabilities with constraints on SWaP and endurance.
• Airborne Fog/Cloudlet Layer: Implements fog or cloudlet servers, often on larger UAVs or balloons, for local analytics, storage, aggregation, and orchestration (Loke, 2015). This intermediate tier reduces end-to-end latency and energy in client/server interactions.
• Ground Backend: Encompasses terrestrial 5G/6G stations, gateways, and cloud data centers offering compute backhaul, ML processing, and centralized control.

IoFT networks can be further structured into hierarchical architectures, integrating low-altitude UAVs, mid-altitude HAPs, and satellite platforms to scale coverage and robustness (Yang et al., 2021).

Formal system models emphasize:

• Nodes: Characterized by position $\mathbf{q}_i(t)\in\mathbb{R}^3$, velocity, energy budget, local compute/storage.
• Connectivity: Air-to-Ground (A2G), Air-to-Air (A2A), and Ground-to-Ground (G2G) links, with diverse physical-layer models (LoS/Non-LoS probabilities, path loss, Rician or Nakagami fading) (Liu et al., 2020, Saif et al., 2021).
• Protocols: Network stack combines customized MAC (3D TDMA/FDMA/CSMA), network-layer routing, SDN/NFV, and RESTful mission APIs (Loke, 2015, Liu et al., 2020).
• Analytical Models: Path loss, coverage probability, achievable rate $R = B\log_2(1+\mathrm{SINR})$, queueing theory for mesh relay, and mobility-aware optimization (Nikooroo et al., 2022, Abouzaid et al., 2020).

2. Mobility, Coverage, and Mesh Networking

Mobility is central to IoFT, dictating coverage, capacity, and robustness.

• Trajectory Optimization: UAVs must be dynamically positioned to maximize throughput, minimize energy, and ensure robustness under constraints (battery, airspace, regulatory). Optimization formulations are typically non-convex, involving trajectory $\mathbf{q}(t)$, transmission power, and capacity guarantees (Nikooroo et al., 2022). Alternating geometric-convex algorithms—combining sphere intersection, local quadratic approximation, and convex solvers—efficiently compute feasible 3D placements and power allocations achieving 15–46% sum-capacity improvement over benchmarks.
• Coverage Probability: Air-to-ground LoS probability $P_{\mathrm{LoS}}(\theta)$ is captured by S-curve models parameterized by elevation angle and environment. Optimal coverage requires carefully selecting UAV altitude and ground-user horizontal range to maximize $P_{\mathrm{cov}}$, e.g., optimal $\theta^\star\approx 20^\circ$ (suburban) or $\approx 45^\circ$ (urban), with $r_o\lesssim 100\,\mathrm{m}$ (Saif et al., 2021).
• Mesh Networking: IoFT supports multi-hop aerial meshes, as in the 802.11s-based “meshing of the sky,” where UAVs sequentially relay packets to a cloud-connected gateway UAV (Abouzaid et al., 2020). Queueing models (G/G/1 or G/G/2 per node), MAC contention, and WFQ (Weighted Fair Queuing) arbitrate flows, with explicit cross-layer performance tuning of queue loads, throughput, and delay given PHY and MAC parameters. Cross-layer design is critical for stability and high throughput, with optimal uplink forwarding probabilities $f_U^A\approx 0.3$–$0.4$.

3. Computation, Edge Intelligence, and Energy Constraints

IoFT nodes function as edge servers supporting real-time analytics and ML inference.

• Joint Trajectory and Computation/Offloading: Mobile edge computing resources deployed on UAVs require co-optimization of flight trajectory, offloading strategy, CPU scheduling, and energy allocation, especially under uncertainty (e.g., UAV jittering). Chance-constrained energy minimization, reformulated by Bernstein-type inequalities, admits convex surrogates solved via alternated SCA (successive convex approximation) methods (Tang et al., 2023). Robust plans consistently respect probabilistic outage constraints (unlike non-robust baselines), maintaining a small energy penalty (2–5%) and can halve total energy versus all-local schemes.
• Airborne Fog Computing: Processing-intensive tasks (e.g., video analytics) are offloaded from resource-limited UAVs to airborne fog nodes. Task assignment and placement optimization minimize weighted latency and energy under compute and energy budgets and airspace constraints (Loke, 2015). Heuristics such as clustering and predictive scheduling are used alongside mathematical optimization to adapt to dynamic workloads.
• Biological Flyers: Living-IoT demonstrates a 102 mg “backpack” that transforms live insects into fully mobile, location-aware sensor/communication nodes through ultra-low-power microelectronics, backscatter uplink, ambient RF self-localization, and micro-battery/solar energy harvesting (Iyer et al., 2018).

4. Security and Privacy Mechanisms

Security is a first-order design criterion, particularly due to the threat surface of airborne communications and the high-value data flows in IoFT.

• Physical Layer Security: Integrating UAV relays between terrestrial base stations and IoT end nodes substantially elevates the achievable secrecy rate, maintaining secure links even under adversarial presence. Time-averaged secrecy can be maximized by optimal UAV trajectory selection and speed/altitude adaptation, with instant recovery possible even near eavesdropper proximity (Abdalla et al., 2020).
• Authentication in Cross-Domain IoFT: Lightweight, mutual authentication schemes leverage dual root-of-trust: radio-frequency fingerprints (RFF) for device identification and PUFs (Physical Unclonable Functions) for ephemeral key agreement and one-time-pad encryption. These decentralized, over-the-air enrollable protocols eliminate the need for stored key material and permit highly efficient, two-message authenticated key exchange with perfect forward secrecy, outperforming classical PUF or RFFI-only schemes in latency, communication, and storage overhead (Chen et al., 26 Dec 2025).
• Intrusion Detection and Data Privacy: IoFT intrusion detection faces severe class imbalance and privacy risk. Multi-stage pipelines such as PrivFly integrate self-supervised pretraining (VIME-style masked feature reconstruction), rare-class augmentation (SMOTE/CTGAN), and privacy-preserving DNN classifiers trained under DP-SGD. SHAP-based explainability quantifies privacy’s effect on model behavior; the composition maintains 98–99% accuracy/F1 while providing $(\varepsilon,\delta)$-differential privacy guarantees (Menssouri et al., 19 Jan 2026).

5. Algorithm Development, Simulation, and System Verification

Practical IoFT deployments demand robust, validated distributed algorithms.

• Development Toolchains: Multi-environment frameworks (e.g., GrADyS-SIM NextGen) enable seamless prototyping, realistic network-physics emulation, and hardware-in-the-loop testing within a unified Python toolchain (Lamenza et al., 2024). Iterative development cycles, where protocols are advanced from abstract event-driven models to full OMNeT++/INET simulation and then to SITL, uncover bugs undetectable in lower fidelity settings and ensure code portability to real vehicles.
• Simulation Techniques: Modular OMNeT++/INET architectures with parameterized mobility (MAVLink-style waypoints), 802.11 Friis-model radio, and emergent coordination protocols (e.g., DADCA, ZigZag) allow instrumented benchmarking of swarm behavior, data collection efficiency, and resilience to failure (Lamenza et al., 2022). Metrics include coverage ratio, handover latency, and throughput, with experimental tuning of mission parameters and adaptation mechanisms for realistic deployment contexts.

6. Applications, Use Cases, and Scalability Considerations

IoFT systems support a spectrum of applications requiring coordinated mobility, adaptive resource management, and robust communications:

• Disaster Response and Emergency Networks: UAVs rapidly reconstitute wireless coverage in post-disaster zones via dynamic mesh networks, achieving coverage probabilities $\ge0.9$ in suburban and $\ge0.7$ in dense urban environments at optimal deployment parameters (Saif et al., 2021).
• Precision Agriculture and Environmental Monitoring: Swarms of drones (and, biologically, sensorized insects) deliver wide-area, granular sensing for agricultural analytics, environmental hazard mapping, and pollination tracking. Ultra-light platforms extend this reach at minimal energy cost (Iyer et al., 2018).
• Event Coverage and On-demand Services: Airborne fog nodes (balloons, larger UAVs) form temporary event networks, supporting high-throughput media delivery and customized user services (drones-as-a-service, fly-in/out infrastructure) (Loke, 2015).
• Industrial Automation and Edge Computing: UAV-mounted MEC servers support latency-critical IIoT workloads, with robust trajectory and task allocation accounting for physics uncertainties, computation/energy trade-offs, and probabilistic channel constraints (Tang et al., 2023).

Scalability is conditioned by the interplay between air-link quality, processing bottlenecks, and ground/cloud backhaul, mandating adaptive clustering, SDN/NFV orchestration, and federated/online learning-based adaptation (Liu et al., 2020, Yang et al., 2021).

7. Open Problems and Future Research Directions

Several research avenues remain open in the IoFT domain:

• Integrated, Hierarchical Networking: Multi-layer ACNs (UAV+HAP+satellite) with unified cross-layer optimization, AI-native protocol stacks, and self-organizing cluster formation for dynamic scenario adaptation (Yang et al., 2021).
• Lightweight Onboard Intelligence: Realization of TinyML, federated learning, and meta-learning models compatible with SWaP limits, enabling distributed perception, adaptation, and collaborative sensemaking.
• Secure, Resilient Orchestration: Physical-layer security through dynamic trajectory/jammer design, decentralized authentication (PUF+RFFI), blockchain-backed UAV ID, and adversarial-robust learning (Chen et al., 26 Dec 2025).
• Simulation/Experiment Harmonization: “What-you-test-is-what-you-fly” toolchains with seamless integration of emulation, physics-based simulation, and hardware-in-the-loop support, plus automated parameter sweeping across environmental and adversarial scenarios (Lamenza et al., 2024).
• Regulatory, Energy, and Human-in-the-Loop Challenges: Airspace management, in-flight energy harvesting, universal open APIs, hybrid autonomy/human control, and context-driven orchestration languages remain as practical and theoretical frontiers.

Collectively, current research demonstrates that IoFT is maturing toward a rigorously engineered, scalable, robust, and AI-native system architecture—one capable of supporting mission-critical applications under adversarial conditions and operational uncertainties, while catalyzing new domains of airborne cyber-physical computation and control (Loke, 2015, Liu et al., 2020, Yang et al., 2021).