Source-linked AI summary
Cell-Free Satellite-UAV Networks for 6G Wide-Area Internet of Things
Chengxiao Liu, Wei Feng, Yunfei Chen, Cheng-Xiang Wang, Ning Ge
TL;DR
Wide-area IoT devices often lie outside terrestrial cellular coverage, while conventional cellular deployment and satellite service face important limitations. This paper proposes process-oriented, iterative allocation of subchannels, transmit power, and hovering times for cell-free satellite-UAV networks, with simulations showing improved efficiency, convergence, and adaptive coverage patterns.
Problem
Wide-area IoT devices outside terrestrial cellular coverage are difficult to serve because conventional deployment is cost-ineffective and satellite networks have limited communication rate and high latency.
Method
The paper jointly allocates subchannels, transmit power, and hovering times across the whole UAV flight process using slowly-varying large-scale CSI and iterative decomposition of allocation subproblems.
Results
Simulation results show beneficial performance from the proposed methods, with Algorithm 3 converging within 2 iterations and Algorithm 6 within 6 iterations or fewer.
Takeaways & Limitations
The observed irregular, subchannel-dependent coverage areas indicate that cell-free CSUNs can adapt coverage while coordinated UAVs improve efficiency in varying large-scale channel environments.
Abstract
from arXiv · showhide
In fifth generation (5G) and beyond Internet of Things (IoT), it becomes increasingly important to serve a massive number of IoT devices outside the coverage of terrestrial cellular networks. Due to their own limitations, unmanned aerial vehicles (UAVs) and satellites need to coordinate with each other in the coverage holes of 5G, leading to a cognitive satellite-UAV network (CSUN). In this paper, we investigate multi-domain resource allocation for CSUNs consisting of a satellite and a swarm of UAVs, so as to improve the efficiency of massive access in wide areas. Particularly, the cell-free on-demand coverage is established to overcome the cost-ineffectiveness of conventional cellular architecture. Opportunistic spectrum sharing is also implemented to cope with the spectrum scarcity problem. To this end, a process-oriented optimization framework is proposed for jointly allocating subchannels, transmit power and hovering times, which considers the whole flight process of UAVs and uses only the slowly-varying large-scale channel state information (CSI). Under the on-board energy constraints of UAVs and interference temperature constraints from UAV swarm to satellite users, we present iterative multi-domain resource allocation algorithms to improve network efficiency with guaranteed user fairness. Simulation results demonstrate the superiority of the proposed algorithms. Moreover, the adaptive cell-free coverage pattern is observed, which implies a promising way to efficiently serve wide-area IoT devices in the upcoming sixth generation (6G) era.
I. INTRODUCTION
Wide-area IoT devices often lie beyond terrestrial coverage, while satellite and UAV limitations make massive access difficult. The paper therefore develops a cell-free, process-oriented CSUN resource-allocation approach using large-scale CSI and coordinated satellite-UAV operation.
- Motivation: Wide-area IoT devices outside terrestrial coverage are difficult to serve because conventional IoT technologies and satellite networks face deployment, rate, and latency limitations.Examples include ocean buoys and remote-area sensors, while satellite networks have limited communication rate and inherent large latency.
- Motivation: Conventional cellular coverage is cost-ineffective for sparsely and unevenly distributed devices, and wide-area spectrum scarcity prevents straightforward local spectrum reuse.These challenges motivate cell-free on-demand coverage and opportunistic spectrum sharing in CSUNs.
- Approach: The proposed CSUN jointly allocates subchannels, transmit power, and hovering time across a satellite and UAV swarm to establish spectrum-efficient cell-free coverage.The framework targets wide-area IoT access while incorporating the whole UAV flight process.
- Approach: Resource allocation considers the whole UAV flight process and uses slowly-varying large-scale CSI under UAV energy and satellite-user interference-temperature constraints.The optimization operates on a larger time scale than channel coherence time, enabling predictive use of large-scale CSI.
- Approach: The algorithms iteratively allocate subchannels, transmit power, and hovering times, while a fairness-oriented formulation promotes service to IoT devices as equally as possible.The efficiency objective is decomposed into subproblems and solved iteratively with low complexity; a further formulation targets minimum data transmission efficiency.
- Findings: Simulations report improved network efficiency and user fairness, together with an adaptive cell-free coverage pattern when only large-scale CSI is used.The coverage illustration concerns a satellite and coordinated UAV swarm serving wide-area IoT users outside terrestrial cellular coverage.
III. PROCESS-ORIENTED DATA TRANSMISSION EFFICIENCY MAXIMIZATION
The paper formulates data transmission efficiency maximization for CSUN resource allocation. The resulting problem is a constrained MINLP that is simplified for iterative solution.
- Problem formulation: The data transmission efficiency maximization problem optimizes transmit power, hovering time, and subchannel allocation jointly.The objective is written as max over P, T, and x of D_e(P,T,x).
- Constraints: The formulation includes an interference temperature threshold, practical constraints, and exclusive subchannel assignment to avoid harmful interference.Each subchannel can be used by only one UAV user under the stated assignment constraint.
- Solution strategy: Because the formulation is a non-convex mixed-integer nonlinear program, the paper simplifies it and solves it iteratively.The direct problem is described as hard to solve directly.
A. Problem Transformation
The paper replaces the original expected-efficiency objective with an approximation involving slack variables, then decomposes the resulting problem into tractable resource-allocation subproblems.
- Approximation: A new objective D_a(P,T,w,x) approximates D_e(P,T,x) without an expectation, using slack variables w.The slack variables satisfy an additional relation that couples them to transmit power.
- Approximation: The approximation expresses rate as modified data rates plus compensation terms related to w, while w remains an intractable implicit function of P.The paper uses this approximation to recast the original optimization problem.
- Reformulation: The recast objective maximizes D_a(P,T,w,x) subject to the transformed constraints, including binary subchannel indicators, nonnegative powers, and nonnegative hovering times.The displayed constraints include x_n,u,g ∈ {0,1}, p_n,g,k ≥ 0, and T_n ≥ 0.
B. Problem Decomposition
The nonconvex efficiency-maximization problem is decomposed into subchannel allocation, coordinated power allocation, and hovering-time scheduling, then solved iteratively. Time-sharing relaxation and block coordinate descent yield tractable subproblems and locally optimal allocation updates.
- Problem Decomposition: The nonconvex problem is decomposed into three subproblems using block coordinate descent.The subproblems separately address subchannels, transmit power, and hovering times.
- Subchannel Allocation: Time-sharing relaxation converts the subchannel-allocation subproblem into a form solvable through iterative updates.The allocation variables can represent the fraction of time assigned to each UAV user and subchannel.
- Subchannel Allocation: The subchannel-allocation solution is obtained iteratively by updating allocation variables and Lagrangian multipliers.The method uses the Lagrangian dual function, KKT conditions, and subgradient updates.
- Subchannel Allocation: Algorithm 1 maximizes the approximated objective at every iteration and obtains at least a locally optimal solution to the subchannel subproblem.The selected subchannel is the one with the largest corresponding transmission-efficiency value for each user.
D. Coordinated Power Allocation
Coordinated power allocation addresses the coupled, nonconvex power problem through relaxation and iterative decomposition. The resulting algorithm converges to a saddle point and provides at least a locally optimal solution.
- Coordinated Power Allocation: The coupled power-allocation problem is relaxed and solved iteratively because its original coupling is difficult to handle directly.The method alternates between solving power variables and updating auxiliary variables.
- Coordinated Power Allocation: The reformulated power problem is decomposed into two subproblems whose equivalence to the original formulation is established.The decomposition separates the optimization of power variables from the associated auxiliary variables.
- Convergence: Algorithm 2 converges to a saddle point by following subgradient directions during the iterative solution process.Concavity in power variables and convexity in auxiliary variables support the saddle-point formulation.
- Convergence: The objective is non-decreasing across iterations, so Algorithm 2 derives at least a locally optimal solution.This monotonicity result is stated for the approximated objective under the coordinated power-allocation procedure.
E. Convergence Analysis
Algorithm 3 alternates among subchannel allocation, power allocation, and hovering-time optimization throughout its iterations. Its objective increases monotonically and is bounded, guaranteeing convergence to at least a locally optimal solution.
- Algorithm 3: Algorithm 3 initializes allocation, power, and hovering-time variables before repeatedly solving the three optimization subproblems.Each iteration updates x, P, and T in sequence until the stopping criterion is met.
- Algorithm 3: The subchannel update produces a locally optimal solution for the current power and hovering-time variables.The resulting objective value is no smaller than that of the previous subchannel allocation.
- Algorithm 3: The power and hovering-time updates are obtained by solving their respective subproblems after the subchannel update.The procedure therefore coordinates all three resource domains within each iteration.
- Convergence Analysis: The objective function increases at every iteration and is upper bounded by the available resources, guaranteeing convergence.Consequently, Algorithm 3 derives at least a locally optimal solution.
- Fairness: Favoring users with better channel conditions can improve transmission efficiency but does not guarantee fairness.Users remaining in poor channel environments may be served infrequently, motivating a separate fairness metric.
IV. PROCESS-ORIENTED MINIMUM DATA TRANSMISSION EFFICIENCY MAXIMIZATION
The paper introduces a minimum data transmission efficiency objective to improve fairness across UAV users. Its resulting max-min mixed-integer nonlinear problem is decomposed into three subproblems for iterative solution.
- Objective: User fairness is improved by using minimum data transmission efficiency as the objective function.The objective seeks to provide services to users as equally as possible.
- Problem Formulation: The resulting optimization problem is a max-min mixed-integer nonlinear program that is difficult to solve directly.The formulation retains the constraints from the earlier efficiency-maximization problem.
- Problem Decomposition: The problem is decomposed into subchannel, power, and hovering-time subproblems using block coordinate descent.The three subproblems are solved as components of the iterative resource-allocation procedure.
- Problem Decomposition: The hovering-time subproblem is a linear max-min optimization problem that can be solved directly with conventional tools.The subchannel and power subproblems remain integer-linear and non-convex, respectively.
B. Max-min Subchannel Allocation
The max-min subchannel allocation problem is transformed to prioritize the worst-served user while enabling a greedy iterative solution. The resulting algorithm improves minimum data transmission efficiency and converges to a locally optimal solution.
- Greedy allocation: The greedy algorithm prioritizes the user with the lowest data transmission efficiency and allocates subchannels to improve the minimum efficiency.This implements the max-min objective by serving the user in the worst condition as much as possible at every step.
- Algorithmic outcome: Algorithm 4 converges to a locally optimal solution because each allocation step improves the minimum data transmission efficiency.The method is designed for the max-min objective rather than the overall data transmission efficiency objective of the alternative algorithm.
C. Max-min Power Allocation
The max-min power allocation problem is handled by decoupling coupled power and auxiliary variables, then alternating convex subproblems. The resulting iterative method converges and yields a locally optimal solution under the stated formulation.
- Problem transformation: The original power-allocation problem is non-convex because transmit power and the auxiliary variable are coupled in constraints (55b) and (55f).Theorem 3 provides an equivalent reformulation of these constraints.
- Problem transformation: Theorem 3 decouples transmit power and the auxiliary variable, allowing successive convex optimization but motivating a lower-complexity iterative method.The paper notes that directly using successive convex optimization would incur excessive computational overhead.
- Iterative solution: The iterative algorithm alternates between solving a convex problem for power and slack variables and equations for the auxiliary variable.At each iteration, the auxiliary variable is held constant for the convex subproblem and then updated from the resulting solution.
- Convergence: The algorithm is guaranteed to converge and can obtain a locally optimal solution to the power-allocation problem.The saddle-point argument establishes convergence, while the iterative multi-domain procedure derives a locally optimal solution for the broader allocation problem.
- Joint allocation: The broader resource-allocation procedure jointly iterates over subchannels, transmit power, and hovering times because these domains are coupled in the optimization.Both optimization problems are solved iteratively to coordinate these three resource domains.
V. SIMULATION RESULTS AND DISCUSSIONS
Simulations evaluate convergence, algorithmic efficiency and fairness, adaptive coverage, and the effects of UAV-swarm size and subchannel count in CSUNs. The proposed algorithms converge quickly, improve efficiency and fairness, and produce adaptive cell-free coverage patterns under the tested settings.
- Simulation setup: The simulations use a CSUN with one satellite, K = 6 UAVs, 10 satellite users, 200 UAV users, 20 user groups, M = 6 antennas, 5.8 GHz operation, and G = 16 subchannels.UAV large-scale CSI is generated from real channel environments using simulation software.
- Convergence: Algorithm 3 converges in 2 iterations, while Algorithm 6 requires no more than 6 iterations across 10 user-location snapshots.The results indicate practical convergence potential for CSUN deployment.
- Algorithm comparison: Algorithm 6 improves user fairness compared with other schemes, while also providing some overall data transmission efficiency gains.At low interference temperature thresholds, Scheme 2 performs worse than Scheme 3 because it underestimates interference temperature.
- Coverage: Algorithm 6 produces irregular coverage areas whose shapes and ranges vary across subchannels, enabling adaptive cell-free CSUN coverage.The illustration uses a −92 dBm service threshold with four UAV users and five satellite users.
- UAV-swarm size: Increasing the number of UAVs improves both overall and minimum data transmission efficiency through greater diversity gain and more flexible coordination.The simulations also identify limited on-board energy as a dominant bottleneck for cell-free CSUNs.
- Subchannel count: Using more subchannels improves both algorithms' performance, but the number of available subchannels should be appropriately designed because Algorithm 6 derives a locally optimal solution.The performance gain from increased communication energy fluctuates with the number of subchannels.
VI. CONCLUSIONS
The paper proposes process-oriented multi-domain resource allocation for cell-free IoT-oriented CSUNs, optimizing UAV flight processes with slowly-varying large-scale CSI. Iterative allocation improves network efficiency and reveals cell-free coverage patterns for wide-area IoT access.
- VI. CONCLUSIONS: The framework jointly allocates subchannels, transmit power, and hovering times across the whole UAV flight process.It uses only slowly-varying large-scale CSI.
- VI. CONCLUSIONS: The optimization maximizes data transmission efficiency while guaranteeing user fairness under process-oriented resource allocation.Time-sharing and feasible-region relaxation methods are used before iterative allocation.
- VI. CONCLUSIONS: Simulation results show that the proposed methods improve network efficiency and produce a cell-free coverage pattern.The observed pattern indicates a promising approach for supporting massive access to wide-area IoT devices.
APPENDIX A PROOF OF THEOREM 2
The appendix proves Theorem 2 by showing that several constraints are redundant under the remaining conditions. It also establishes equivalence between two sets of conditions used for Theorem 3.
- APPENDIX A PROOF OF THEOREM 2: Constraints (51c)–(51e) are automatically satisfied whenever x satisfies (51b), (51f), and (51g).The proof handles both the all-zero vector and nonzero feasible vectors.
- APPENDIX A PROOF OF THEOREM 2: Because constraints (51c)–(51e) have no influence on (51), the conclusion of Theorem 2 follows.
- APPENDIX A PROOF OF THEOREM 2: Conditions (B.1) and (B.2) are equivalent to the more general conditions (B.4) and (B.5), yielding the conclusion of Theorem 3.