Source-linked AI summary
UAV Relay-Assisted Emergency Communications in IoT Networks: Resource Allocation and Trajectory Optimization
Dinh-Hieu Tran, Van-Dinh Nguyen, Sumit Gautam, Symeon Chatzinotas, Thang X. Vu, Bjorn Ottersten
TL;DR
Emergency IoT communications require timely data collection despite device latency requirements and limited UAV storage. The paper jointly optimizes FD/HD relaying resources and UAV trajectory using relaxation and inner approximation, with numerical results showing improved served-device counts and throughput over benchmarks.
Problem
Emergency IoT data can become unreliable when stale, while device and UAV storage limitations constrain timely collection.
Method
The paper jointly optimizes bandwidth, device/UAV power, trajectory, and service decisions using continuous relaxation, penalties, and iterative inner approximation.
Results
20 and 13 served GUs are obtained by the proposed FD and HD methods, respectively, while proposed schemes improve served devices and collected throughput over benchmarks.
Takeaways & Limitations
FD relaying can serve more IoT devices than HD in the reported setting, and the proposed schemes improve both service count and throughput over benchmarks.
Takeaways & Limitations
When U is large, the UAV should operate in HD mode for simpler implementation.
Abstract
from arXiv · showhide
Unmanned aerial vehicle (UAV) communication has emerged as a prominent technology for emergency communications (e.g., natural disaster) in the Internet of Things (IoT) networks to enhance the ability of disaster prediction, damage assessment, and rescue operations promptly. A UAV can be deployed as a flying base station (BS) to collect data from time-constrained IoT devices and then transfer it to a ground gateway (GW). In general, the latency constraint at IoT devices and UAV's limited storage capacity highly hinder practical applications of UAV-assisted IoT networks. In this paper, {full-duplex (FD) radio} is adopted at the UAV to overcome these challenges. In addition, half-duplex (HD) scheme for UAV-based relaying is also considered to provide a comparative study between two modes (viz., FD and HD). {Herein, a device is considered to be successfully served iff its data is collected by the UAV and conveyed to GW timely during flight time}. In this context, we aim to maximize the number of served IoT devices by jointly optimizing bandwidth, power allocation, and the UAV trajectory while satisfying each device's requirement and the UAV's limited storage capacity. The formulated optimization problem is troublesome to solve due to its non-convexity and combinatorial nature. {Towards appealing applications, we first relax binary variables into continuous ones and transform the original problem into a more computationally tractable form.} By leveraging inner approximation framework, we derive newly approximated functions for non-convex parts and then develop a simple yet efficient iterative algorithm for its solutions. Next, we attempt to maximize the total throughput subject to the number of served IoT devices. Finally, numerical results show that the proposed algorithms significantly outperform benchmark approaches in terms of the number of served IoT devices and system throughput.
I. INTRODUCTION
UAV relaying is positioned for emergency IoT communications because fixed infrastructure, stale data, latency requirements, and limited device storage constrain timely collection. The paper jointly considers FD and HD relaying with resource allocation and trajectory optimization, using iterative algorithms whose numerical evaluation improves served devices and throughput over benchmarks.
- Fixed terrestrial BSs cannot rapidly shift resources, motivating UAVs as aerial platforms for emergency IoT communications.
- Out-of-date data can undermine emergency decisions, while limited IoT storage makes timely UAV collection necessary.
- Unlike prior HD studies focused on UL or DL timing, this work jointly addresses RT constraints for both UL and DL transmissions.
- The proposed model jointly optimizes bandwidth, device and UAV power, trajectory, storage, and latency while maximizing served IoT devices.
- FD and HD UAV relaying are both investigated, with FD intended to improve served devices, throughput, latency, and storage utilization.
- Numerical results show significant improvements in served IoT devices and collected throughput over benchmarks using fixed allocation or fixed trajectories.
B. Caching Model
The UAV cache is explicitly limited, and its occupancy accounts for data collected from devices but not yet transmitted to the gateway. This enables storage feasibility to be tracked over time and motivates FD operation for releasing cached data.
- The UAV cache has storage capacity C, so the total cached files must remain within that capacity.
- At time slot n, stored data equals files collected through n minus files transmitted to the GW through n−1, preserving space for future streams.
C. Problem Formulation
The formulation maximizes served IoT devices by jointly selecting trajectory and UL/DL resources under operational, storage, and timing constraints. Its mixed-integer non-linear structure and non-convex constraints make direct solution difficult.
- The objective maximizes served IoT devices through joint optimization of trajectory, UL/DL bandwidth, and transmit power under storage and timeout requirements.
- The model includes flight-time transmission assumptions, UAV movement and power limits, device data requirements, and fixed start and end locations.
- PFD is a mixed-integer non-linear program that is generally NP-hard because of binary and non-convex constraints.
III. PROPOSED ITERATIVE ALGORITHM FOR SOLVING PFD
The proposed solution relaxes binary service variables, penalizes non-binary values, and applies inner approximation to obtain successive convex programs. Each iteration solves an approximation and updates the feasible point until convergence.
- Inner approximation replaces non-convex differentiable functions with convex approximations around a feasible iterate.
- The iterative procedure initializes a feasible point, solves the approximate convex program, updates the iterate, and repeats until convergence.
- Binary variables λ_k are relaxed to 0≤λ_k≤1 and a penalty function encourages near-exact binary solutions.
- When relaxed variables are binary at optimum, the relaxation is tight and the resulting solution is feasible for the original problem.
- The relaxed problem remains difficult because of non-concave objectives, non-convex constraints, and strong coupling among optimization variables.
B. Proposed IA-based Algorithm
The proposed inner-approximation (IA) method convexifies the non-convex design problem through successive bounds, slack variables, and first-order approximations, then solves iterative convex programs until convergence.
- B. Proposed IA-based Algorithm: The IA method exposes non-convex components and convexifies them using lower bounds, slack variables, and first-order Taylor approximations.The construction includes difference-of-convex reformulations and approximated rate-related constraints.
- B. Proposed IA-based Algorithm: The objective's convex penalty function is iteratively replaced by a linear approximation in the relaxed served-device variables.This approximation supports the iterative convex-program formulation.
- B. Proposed IA-based Algorithm: At each iteration, the algorithm solves an approximate convex program, updates the design variables, and repeats until convergence.The variable set includes trajectory, allocation, power, service, slack, and rate-related variables.
- B. Proposed IA-based Algorithm: An initial feasible point is obtained by solving an auxiliary problem that maintains feasibility before successive convex programs are solved.The starting point must satisfy the relevant feasibility condition and other constraints.
C. Convergence and Complexity Analysis
The convergence analysis establishes that the IA-based algorithm generates improved solutions that converge to at least a local optimum of the relaxed problem.
- C. Convergence and Complexity Analysis: Algorithm 1 is based on inner approximation, whose convergence is established through the cited IA framework.The section introduces a proposition to make the convergence argument self-contained.
- C. Convergence and Complexity Analysis: The iterative procedure repeatedly solves the designated optimization problem, updates the variables, and stops when convergence is reached.Algorithm 2 provides the corresponding iterative structure for its convex subproblem.
- C. Convergence and Complexity Analysis: Algorithm 1 generates a sequence of improved solutions converging to at least a local optimum of the relaxed problem PFD_relaxed.The proposition states the convergence result directly.
2) Complexity Analysis:
The HD formulation adapts the relay-rate model and applies the same IA strategy to a mixed-integer non-convex served-device problem, with HD-specific transmission sequencing and convex subproblems.
- 2) Complexity Analysis:: The FD and HD formulations use analogous IA transformations, while the HD model modifies the rate expressions to reflect half-duplex operation.The HD derivation reuses developments from the FD formulation.
- 2) Complexity Analysis:: In HD mode, the UAV transmits data to the GW only after collecting data from all GUs, so the residual self-interference term disappears.The achievable rates are then defined for the UAV-device and UAV-GW links under this sequencing.
- 2) Complexity Analysis:: The HD served-device problem remains mixed-integer non-convex because of binary and non-convex constraints.The formulation relaxes the binary constraint and introduces auxiliary variables before applying IA.
- 2) Complexity Analysis:: The HD rate expressions are reformulated with slack variables and first-order approximations, producing convex constraints for iterative solution.The construction uses lower bounds and Taylor approximations for the relevant rate terms.
- 2) Complexity Analysis:: The resulting HD problems are solved by successively solving simpler convex programs and updating the optimization variables until convergence.A penalty function is used to promote exact binary values of the service variables.
1) Complexity Analysis:
The proposed IA-based algorithms iteratively solve convex approximations and update the optimization variables until convergence. The complexity analysis counts the constraints and scalar variables of the convex subproblem and incorporates the iteration count needed to reach a local solution.
- The convex problem (58) contains N(7 + 8K) + 4K linear and quadratic constraints and 3N(1 + 4K) + K scalar real variables.
- The per-iteration complexity for solving problem (58) depends on its constraint and variable counts, while the overall complexity also includes N_i, the iterations required to reach a local solution.
- Algorithm 3 repeatedly solves problem (58), updates q, a, p, λ, z, Φ, and r, and stops upon convergence.
- Algorithm 4 applies the same IA-based iterative structure to problem (59), solving (60), updating the optimization variables, incrementing j, and terminating at convergence.
B. Throughput Maximization
The paper evaluates throughput-oriented FD and HD UAV relaying through joint resource allocation and trajectory designs under latency, power, cache, and flight constraints. Numerical results examine how QoS, cache size, data size, network size, transmit power, and operating mode affect served devices and throughput.
- Optimization procedure: The HD throughput problem is transformed into a convex optimization problem and solved iteratively using the same approximation steps as the preceding formulation.The procedure reuses slack variables and obtains solutions through Algorithm 4.
- Experimental setup: The designed UAV trajectory and joint bandwidth-power allocation are evaluated against benchmarks using fixed resource allocation or a fixed linear trajectory.The experiments use randomly distributed IoT devices and compare proposed FD and HD schemes with BFD1, BFD2, BHD1, and BHD2.
- FD–HD comparison: 20 versus 13 served GUs are obtained by the proposed FD and HD methods, respectively, because FD transfers data to the GW immediately after collecting it.The resulting additional flight time toward GUs and the GW helps FD satisfy more latency requirements than HD.
- Benchmark comparison: 85% versus less than 15% of GUs are served by the HD algorithm and BHD1, respectively, while BHD2 serves 35% under the reported setting.The proposed FD and HD algorithms outperform the corresponding benchmarks in Figs. 4 and 5.
- Cache and packet size: 85% versus 40% of IoT users are served by FD and HD at S_k ∈[10, 30] Mbits and C = 400 Mbits, while increasing packet size degrades performance.The FD advantage persists across cache sizes; smaller packets require less time and resources to serve.
- Power and self-interference: FD outperforms HD below P^max_U = 22 dBm, whereas both modes serve the same number of users at or above 22 dBm because FD suffers increased residual self-interference.At higher UAV power, FD’s residual self-interference increases noise power, while HD avoids that interference.
B. Throughput Maximization:
The proposed FD and HD algorithms improve throughput over benchmark schemes across network sizes and bandwidth settings, with FD achieving the strongest reported throughput.
- At x = 700 m, the FD algorithm obtains 788 Mbits, compared with less than 131 Mbits for BFD1.BFD2, HD, BHD1, and BHD2 achieve 230, 537, 372, and 140 Mbits, respectively.
- The proposed FD and HD algorithms significantly improve throughput over BFD1, BFD2, BHD1, and BHD2 across all evaluated network sizes.The evaluation defines throughput as data transferred from GUs to the GW for successfully served GUs.
- HD outperforms BFD2 in the reported comparison, underscoring the benefit of optimized resource allocation.
- All schemes achieve higher maximum throughput as total bandwidth increases.The paper attributes this trend to greater transmission enabled by higher bandwidth allocation.
VI. CONCLUSION AND FUTURE DIRECTIONS
The paper formulates and solves resource-allocation and trajectory-design problems for emergency UAV-assisted FD IoT communications under latency and storage constraints. It also studies throughput maximization subject to a minimum served-device threshold and identifies settings favoring HD operation or future system extensions.
- Conclusion: The main problem jointly optimizes UAV trajectory, bandwidth, and IoT-device and UAV transmit power to maximize served IoT devices under timeout and storage constraints.The formulation addresses emergency communications and the UAV’s limited storage capacity.
- Conclusion: An iterative algorithm solves the transformed problem with polynomial computational complexity per iteration.The original non-convex problem is first transformed into a tractable form.
- Conclusion: A second optimization problem maximizes total collected data while requiring a minimum number of served IoT devices.
- Conclusion: When UAV transmit power is large, HD operation is recommended for simpler implementation.
- Future Directions: Future work includes multi-antenna UAV systems and low-complexity machine-learning methods for predicting LoS probability.The multi-antenna extension may increase complexity while potentially improving network performance.
APPENDIX C
The appendix establishes bounds and convergence properties for the iterative approximation used in the optimization framework. It shows that successive approximations improve the objective and converge to a locally optimal solution under the stated constraints.
- APPENDIX C: The approximation sequence improves the relaxed objective because each successive iterate is a better solution of the relaxed problem.
- APPENDIX C: The objective sequence converges, and every accumulation point is a Karush-Kuhn-Tucker point.
- APPENDIX C: The feasible set is convex, connected, closed, and bounded under the power, bandwidth, and flight-time constraints.
- APPENDIX C: The iterative procedure therefore obtains a locally optimal solution to the relaxed optimization problem.