Topology-Aware Integrated Sensing, Communication, Charging in Massive Low-Altitude Wireless Network

I. INTRODUCTION

• The framework targets the coordination challenge created by combining UAV communication, target sensing and localization, and wireless power transfer, including sensing interference with communication.UAVs can also provide sensing, relay, and aerial base-station functions, while WPT is motivated by their limited battery capacity.Source: §I · p. 1

A. Related Work

• Prior ISAC methods use optimization procedures whose complexity grows with target and UE counts, while existing TA solutions lack joint multi-function coordination across terrestrial and non-terrestrial devices.The cited C-R-E analysis also assumes a single sensing target, communication UE, and energy-harvesting device.Source: A. Related Work · pp. 1–2

B. Contributions

• The TA framework models the LAWN as a bipartite graph whose nodes and edges represent heterogeneous functions through shared profit and cost terms.The graph is constructed using channel strength and selectively established AP-to-UE connections, making performance gains and structural costs part of a common representation.Source: B. Contributions · p. 2
• A unified resource-adjustment rule reconfigures topology by changing either resource providers or consumers, avoiding objective-specific heuristic tuning.The algorithm iteratively updates node activation states and edge relationships to adapt the network structure to functional requirements.Source: B. Contributions · p. 2
• Simulations under diverse network configurations and UE densities report that the TA method outperforms state-of-the-art approaches in system efficiency and multi-functional coordination.The paper also reports robust, scalable performance in large deployments with massive terrestrial and non-terrestrial UEs.Source: B. Contributions · p. 2

II. SYSTEM MODEL

• The system model includes terrestrial and UAV access points performing communication, sensing, and WPT, with communication, sensing, and charging UE sets.The APs use uniform planar arrays, and the model describes uplink channel estimation followed by downlink functional tasks.Source: §II · pp. 2–3

A. Uplink Phase

• Channel estimation uses UE pilots and models each AP's received signal with noise and the AP-to-UE channel, including large- and small-scale fading and array steering.The one-shot procedure uses binary pilots, so certain UEs are deactivated in a slot to mitigate interference and can be scheduled in later slots.Source: A. Uplink Phase · p. 3

B. Downlink Phase

• Downlink transmissions include communication, sensing, and charging precoders; each slot selects one sensing target and its serving AP, while the scheme can extend to additional UE categories.Sensing is framed as target detection rather than localization; a serving AP may determine target position using ToA and AoA measurements.Source: B. Downlink Phase · pp. 3–4
• Performance is measured by communication sum rate, sensing SINR, and received charging energy; sensing detection requires SINR at or above a prescribed threshold.For charging UEs, the entire received signal can be used for energy harvesting.Source: B. Downlink Phase · p. 4

• The framework combines graph-based reconstruction, objective representation, and TA-based algorithm design.Source: §III · p. 4
• This section focuses on constructing the LAWN bipartite graph and embedding multiple objectives in it.Source: §III · p. 4
• The workflow starts with graph construction and ends with resource and selection strategy outputs.Source: Figure 2 · p. 5

• The LAWN is represented as a bipartite graph G = (M, K, E), where AP–UE edges reflect feasible connectivity.The graph's edge weights are defined as wmk = amkβmk.Source: A. Graph Representation of LAWN · p. 5
• Links must meet the channel-strength threshold βτ, and each AP retains up to N strongest channels that also meet that threshold.The threshold can be controlled by comparing it with noise strength.Source: A. Graph Representation of LAWN · p. 5
• The representation aims to preserve dominant links while mitigating severe inter-UE interference from conflicting connections.Source: A. Graph Representation of LAWN · p. 5
• Charging-UE pilot contamination is expected to have limited impact, while downlink charging signals are treated as predictable and mitigable rather than stochastic interference.Source: A. Graph Representation of LAWN · p. 5
• The framework focuses on structural interference, treating estimation-induced interference from charging users as negligible in the considered scenario.Source: A. Graph Representation of LAWN · pp. 5–6

• The topological model approximates communication, sensing, and charging objectives using a simplified MRT-based precoding model and average power allocation.Source: B. Topological Objective Representation · p. 6
• Binary variables represent whether UEs are served, APs are active, sensing APs are selected, and charging precoders are assigned.Source: B. Topological Objective Representation · p. 6
• Sensing is always active and assigned to one AP, but charging UEs may receive no dedicated precoder when their channels cannot satisfy the minimum received-power requirement.Such UEs may still harvest negligible leakage energy.Source: B. Topological Objective Representation · p. 6

IV. LOW-COMPLEXITY ALGORITHM DESIGN

• The algorithm treats communication, sensing, and charging objectives as a shared structural problem governed by binary control variables.Source: §IV · p. 6

A. Profit and Cost Design

• The graph uses adjacency matrix A for dominant connections and weight matrix W for edge weights, while shared Profit and Cost concepts capture objective-specific gains and costs.Source: A. Profit and Cost Design · pp. 6–7
• Communication uses a sum-rate objective with SINR, and sensing uses sensing SINR within the same Profit-Cost framework.Source: A. Profit and Cost Design · p. 7
• Charging profit counts the intended precoder's gain as an approximation of received energy, although it does not exactly represent actual charging performance.Charging success requires received energy to exceed threshold τp,T.Source: A. Profit and Cost Design · p. 7

B. Low-Complexity Algorithm I

• A sensing target whose links from all APs fall below the graph-construction threshold becomes isolated and cannot be supported in the current topology.Those negligible links are discarded during graph construction.Source: B. Low-Complexity Algorithm I · p. 7
• For communication UEs, individual channel strengths may exceed βτ even when aggregate received signal strength remains insufficient.The passage begins to describe the scheduling consequence but is truncated.Source: B. Low-Complexity Algorithm I · p. 7
• Algorithm 1 takes matrix A, W, and thresholds as inputs and returns UE and AP status plus a sensing-selection vector.The listed thresholds are τc,T and τs,T.Source: B. Low-Complexity Algorithm I · p. 8

Step 2: Sensing Objective

• Communication UEs not selected in a slot remain eligible for reconsideration in later slots under the same topology.Source: §8 · p. 8
• The sensing objective is served by one AP, making sensing SINR sensitive to interference from active APs and communication UEs.The method selects the AP with the strongest channel connectivity to the target.Source: §8 · p. 8
• If the selected configuration misses the sensing SINR requirement, the algorithm identifies the APs or communication UEs contributing most to interference and adjusts their activation.It reevaluates qs under alternative adjustments and selects the configuration with better performance, repeating until the requirement is satisfied.Source: §8 · p. 8

Step 4: Charging Objective

• The charging stage checks whether charging is possible and checks AP connectivity.Source: §14 · p. 8
• For a conflicting charging UE, the algorithm may close one AP connection or all of that UE’s connectivity.Source: §14 · p. 8
• After updating connectivity, profit, and cost, the algorithm repeats the charging checks and returns UE status, AP status, sensing selection, and charging-AP selection.Source: §14 · p. 8

C. Low-Complexity Algorithm II

• Thresholds govern channel strength, sensing connections, AP connections, and the maximum users per AP; the stated settings include τs,T = 1 and generally NAP,in = 1.The four sequential steps make decisions from both UE and AP perspectives, and the framework is presented as a broader structural paradigm for multi-objective network optimization.Source: C. Low-Complexity Algorithm II · pp. 8–9
• When a threshold is unmet, the framework compares adjusting the resource provider with altering the resource consumer and repeats the higher-profit action until thresholds are met.Source: C. Low-Complexity Algorithm II · p. 9
• Communication and sensing adjustments deactivate interfering UEs or APs, while charging adjustments deactivate a precoder or its full precoding vector.The framework represents interference mitigation and power reallocation as a common resource-reconfiguration problem.Source: C. Low-Complexity Algorithm II · p. 9
• Overall complexity is O(MK(M + KC)), upper bounded by O(MK(M + K)).Steps 2 and 3 each require O(MK(M + KC)); Step 4 requires O(M^2K).Source: C. Low-Complexity Algorithm II · p. 9

A. Simulation Setups

• The simulations use 64 terrestrial and 64 UAV APs across a 2 km × 2 km region, with 80% of UEs airborne and 100 deployment and fading realizations.No Selection activates all communication UEs; UE-Centric Selection chooses UEs by dominant channel relationships, and both benchmarks prune each UE to its three strongest AP links.Source: A. Simulation Setups · pp. 9–10
• The proposed method also filters weak users and can generally serve more UEs than UE-Centric Selection by jointly selecting UEs and APs and deactivating interfering APs.With 200 users and 128 APs, approximately 50 users meet the channel-strength threshold over 2 km × 2 km, versus approximately 120 over 1 km × 1 km.Source: A. Simulation Setups · p. 10

B. Communication Performance

• As UE numbers increase, the TA-based method consistently outperforms the benchmark schemes in sum rate, with its advantage growing in overloaded, interference-limited conditions.Joint UE- and AP-side decisions let the method deactivate interfering APs and enlarge the feasible UE activation set.Source: B. Communication Performance · p. 10, Figure 4 · p. 11

C. Sensing Performance

• The TA-based method consistently outperforms UE-centric selection in target sensing detection probability by identifying and suppressing links that cause significant interference.The comparison uses the probability that the target UE's sensing SINR exceeds its required threshold.Source: C. Sensing Performance · pp. 10–11, Figure 5 · p. 11

D. Power Transfer Performance

• The TA-based method achieves a 100% charging success rate, significantly outperforming user-centric and No Selection schemes; total received energy’s gap versus benchmarks is within approximately 2 to 3 dBm in most percentile ranges.Source: D. Power Transfer Performance · p. 11

E. Threshold Sensitivity Analysis

• When βτ is overly large, the proposed method can underperform UE-centric selection; below approximately σ2/40 for MMSE, its performance is less sensitive to further reductions.βτ filters UEs with excessively weak channels; smaller values retain more potentially useful UEs and reduce sensitivity across a broad range.Source: E. Threshold Sensitivity Analysis · pp. 11–12, Figure 8 · p. 12
• When τc,T is relatively large, fewer communication UEs are retained and sum rate falls below the benchmarks; below approximately −70 dBm, sum rate remains nearly unchanged.Users filtered by τc,T may be deactivated through joint UE/AP selection and scheduled in later time slots.Source: E. Threshold Sensitivity Analysis · p. 12, Figure 9 · p. 12
• The best sum-rate performance occurs at NAP,in = 1; increasing the threshold reduces sum rate because added interference outweighs the gain from serving more UEs.NAP,in limits communication and sensing connections per AP, thereby controlling channel overlap among selected entities.Source: E. Threshold Sensitivity Analysis · p. 12

VI. CONCLUSION

• Simulations show that the TA-based framework outperforms typical benchmarks in sum rate and successful sensing and charging probabilities, including as mobile-user numbers increase.The conclusion also reports that the framework remains advantageous as participation scales.Source: §VI · p. 12