Rate-Aware Quantum-Inspired Trajectory Learning for Interference-Limited Multi-UAV Networks

arXiv:2606.25480 · cs.MA, cs.AI · Submitted 2026-06-24 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Computer Science. The day in computer science: software, systems and theory on arXiv.

Tom: We're looking at the paper: "Rate-Aware Quantum-Inspired Trajectory Learning for Interference-Limited Multi-UAV Networks".

Jane: Here's the gist of the paper. Unmanned aerial vehicle (UAV) can provide on-demand, high-capacity connectivity in disaster and normal situation <ref:2606.25480#pg2> The proposed Rate-Aware Quantum-Annealed Graph Condensation (RA-QAGC) scheme combines rate-aware graph abstraction with decentralized reinforcement learning to enable scalable, interference-aware UAV coordination <ref:2606.25480#pg2> Simulation results demonstrate the proposal outperformed over existing schemes by achieving 59.4 Mbps total throughput and 23.9 Mbps priority-user throughput <ref:2606.25480#pg2>.

Tom: Problem Context The major limiting factor in UAV deployment for trajectory optimization and network resource allocation is the high computational complexity <ref:2606.25480#pg2> As the network size increases, the number of possible network configurations grows combinatorially, significantly elevating the problem and leading to a severe curse of dimensionality <ref:2606.25480#pg2>.

Jane: Existing schemes in the literature addressed the energy efficiency maximization problem by decomposing multi-variable formulations into sequential convex subproblems evaluating trajectory, power allocation, and time-slot assignments <ref:2606.25480#pg2> Traditional tabular and standard deep multi-agent reinforcement learning (MARL) frameworks encounter a severe curse of dimensionality as the network expands, demanding decentralized architectures to maintain real-time decision-making capabilities <ref:2606.25480#pg2>.

Tom: How the training signal is built The framework decouples the problem into a sequential two-stage pipeline: (i) offline global state-space compression via quantum-inspired optimization, and (ii) online localized trajectory tracking handled by decentralized MARL <ref:2606.25480#pg2>.

Paper summary: Tom: So we're looking at this paper called Rate-Aware Quantum-Inspired Trajectory Learning for Interference-Limited Multi-UAV Networks. It looks like they’re tackling a huge problem in drone networks: how to make the UAVs move smartly while dealing with all that interference and keeping quality of service up.

Jane: Right, so the main idea here is using this Rate-Aware Quantum-Annealed Graph Condensation scheme to combine rate awareness with decentralized reinforcement learning for scalable coordination. It claims they can do this better than existing methods by achieving a total throughput of fifty-nine point four Mbps and a priority user throughput of twenty-three point nine Mbps, which is about fifteen percent and thirty-four percent better than the baseline schemes <ref:2606.25480#pg2>.

Lu: What’s interesting about their approach is that they decouple the problem into two stages: first, they do offline global state-space compression using quantum-inspired optimization, and then they handle the online trajectory tracking with decentralized reinforcement learning <ref:2606.25480#pg2>. That separation seems like a smart way to fight the complexity issue.

Meng: I wonder how much overhead that initial quantum-inspired compression adds to the actual deployment time. If we’re shipping this in production, we need something that runs fast and doesn't require massive precomputation before it can actually start moving <ref:2606.25480#pg2>.

Lalam: The paper suggests that by using a rate-aware cost objective function J, which maps physical channel constraints and interference into an energy minimization landscape, they create a compact representation of the search space <ref:2606.25480#pg2>. This process identifies optimal spatial coordinates corresponding to high rate and low interference corridors.

Tom: So the core innovation is this global, quantum-inspired probabilistic annealing mechanism guided by J that prunes and compresses the high-dimensional deployment space into a discrete waypoint candidate set C <ref:2606.25480#pg2>. It’s essentially finding the best spots before the agents even start learning how to fly optimally.

Jane: That condensation leads to this condensed graph, G equals (C, E), which then lets decentralized independent Q-learning agents operate with high efficiency <ref:2606.25480#pg2>. So they’re using that pre-filtered map for the agents to navigate and make decisions in real time.

Lu: I see how this addresses the curse of dimensionality by reducing the continuous trajectory optimization problem into a discrete graph problem on that set C <ref:2606.25480#pg3>. But they’re relying on that initial compression being good enough to guide the learning process <ref:2606.25480#pg3>.

Meng: It sounds like a lot of work is happening upfront to find those candidate waypoints, which is fine theoretically, but what happens if the real-world channel conditions change drastically after that compression? We need robustness there <ref:2606.25480#pg2>.

Lalam: The paper states they use an exponential annealing cooling trajectory to explore the state space and isolate a refined matrix of discrete spatial centroids, which forms that condensed high-reward waypoint candidate set C <ref:2606.25480#pg5>. It keeps exploration going through a tunneling constant ptunnel equals zero point zero five.

Paper summary: Tom: That tunneling constant sounds like they’re trying to keep some search happening even when the system is supposed to be settling down, which is interesting for exploration <ref:2606.25480#pg5>. But the final result showed priority users getting a median data rate of about zero point eight five Mbps while normal users are around zero point five five Mbps <ref:2606.25480#pg6>.

Jane: That difference between the priority and normal user rates is quite noticeable, and they also noted that for edge users, the data rate is above zero, which means they managed to keep those users from being cut off while still maintaining quality of service <ref:2606.25480#pg6>.

Lu: The results show that the learned trajectories exhibit coordinated spacing between UAVs to minimize interference, which is a key physical behavior they achieved <ref:2606.25480#pg3>. That suggests the decentralized learning framework actually learned some useful spatial awareness on its own.

Meng: For deployment, I need to know how much computational power that entire quantum-inspired annealing process requires to run offline and how much memory that condensed graph G takes up at scale <ref:2606.25480#pg3>.

Lalam: The paper is clear about the benefits of this rate-aware abstraction, showing it enables UAVs to identify more favorable operating regions and utilize network resources more efficiently <ref:2606.25480#pg1>. It provides a scalable alternative to centralized optimization whose complexity increases rapidly with network size <ref:2606.25480#pg1>.

Tom: So, what’s the big picture here? This work shows that intelligent state-space reduction can significantly improve how we handle trajectory optimization in wireless networks where interference is a major constraint <ref:2606.25480#pg1>. It really shows how incorporating rate awareness into the abstraction process helps drones find better spots and use resources smarter.

Jane: What this means for the world is that for any drone network, especially in disaster relief or dense urban areas where interference is a big issue, we have a path toward scalable coordination that doesn't get instantly bogged down by massive combinatorial explosion <ref:2606.25480#pg1>.

Lu: Future work mentioned focuses on extending this framework to dynamic user mobility scenarios and continuous control models, and also to large-scale heterogeneous aerial networks <ref:2606.25480#pg1>. That suggests they see the next big challenge being making this adaptable to even faster, more unpredictable movements.

Meng: I'm still focused on deployment reality though. If we want to deploy this, we need a concrete plan for how we’d handle those dynamic mobility changes without losing the efficiency gains from that initial quantum-inspired step <ref:2606.25480#pg2>.

Lalam: The paper is quite sure about its findings, indicating that this rate-aware condensation strategy is effective at improving the practicality of multiUAV trajectory optimization in interference-limited wireless networks <ref:2606.25480#pg1>. It’s a solid piece of research that bridges quantum inspiration with practical reinforcement learning.

Conclusion: Tom: So we've seen how they used this whole Rate-Aware Quantum-Inspired Graph Condensation scheme to figure out how UAVs should move in crowded air, and now we get to the conclusion of this paper.

Jane: Right, so basically, they took a really complicated problem about drone paths and interference and broke it down into steps—first using quantum inspiration to shrink the search space, then using reinforcement learning for the actual flight decisions.

Lu: The authors are Khaoula Khaled, Muhammad Afaq, Ali Arshad Nasir, Zeeshan Kaleem. They're from King Fahd University of Petroleum and Minerals.

Meng: Okay, so what does this actually mean for deployment? Is this something we can plug into a real-world system right now without needing a supercomputer just to set up the initial map?

Lalam: The paper shows that state-space reduction is a big deal for making trajectory optimization practical in wireless networks with interference limitations.

Tom: Exactly, and what they proved with their simulation results is that this method achieves five ninety-four megabits per second total throughput and twenty three point nine megabits per second for the priority users.

Jane: That means for a listener, it translates to a fifteen percent gain in total speed and thirty four percent gain for the users who need guaranteed service.

Lu: The core idea is that by making this global search mechanism rate-aware, you get a much better map of where the best routes are, which then helps the decentralized learning agents work much faster.

Meng: So it's about getting smarter about *where* to go before the drones even start trying to fly—that initial compression is key.

Lalam: The authors conclude that incorporating rate awareness into this abstraction process lets UAVs find better operating regions and use network resources way more efficiently.

Tom: It’s a scalable way to handle the complexity that usually crushes centralized optimization when you have a lot of aerial assets moving around.

Jane: But they did flag some stuff, you know, for the future. They're looking at how this framework can be extended to scenarios where users are constantly moving or where control is continuous instead of just discrete steps.

Lu: That’s right, they want to see if we can take this idea and make it work for faster user movements and more fluid control models on top of the current setup.

Meng: If they can handle dynamic mobility without losing that efficiency from the quantum-inspired step, that's where this paper becomes really useful for real deployment.

Khaoula Khaled, Muhammad Afaq, Ali Arshad Nasir, Zeeshan Kaleem

King Fahd University of Petroleum and Minerals

cs.MA, cs.AI

Submitted: 2026-06-24

Updated: 2026-11-09

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 81/100

The gist: Unmanned aerial vehicle (UAV) can provide on-demand, high-capacity connectivity in disaster and normal situation <ref:2606.25480#pg2> The proposed Rate-Aware Quantum-Annealed Graph Condensation…

Key concepts

Curse of Dimensionality
This problem arises because optimizing the trajectories and resource allocations for many UAVs in large networks creates an exponentially growing number of possible configurations. This makes real-time decision-making computationally impossible with traditional methods.
Rate-Aware Graph Condensation (RA-QAGC)
This scheme uses quantum annealing to prune a massive search space into a small set of promising spatial waypoints. It uses a cost function that balances high data rates against interference, effectively condensing the complex environment into a manageable graph for learning.
Decentralized Reinforcement Learning (MARL)
Instead of one central controller, this approach uses independent Q-Learning agents on the condensed graph. This allows each UAV to make local decisions efficiently based on its immediate surroundings and learned policies, avoiding the heavy computational load of centralized systems.

Terminology

Summary

Unmanned aerial vehicle (UAV) can provide on-demand, high-capacity connectivity in disaster and normal situation <ref:2606. p. 025480#pg2> The proposed Rate-Aware Quantum-Annealed Graph Condensation (RA-QAGC) scheme combines rate-aware graph abstraction with decentralized reinforcement learning to enable scalable, interference-aware UAV coordination <ref:2606. p. 025480#pg2> Simulation results demonstrate the proposal outperformed over existing schemes by achieving 59. p. 0 p. 04 Mbps total throughput and 23. p. 0 p. 0 p. 09 Mbps priority-user throughput <ref:2606. p. 0 p. 025480#pg2>.

Problem Context

The major limiting factor in UAV deployment for trajectory optimization and network resource allocation is the high computational complexity <ref:2606.25480#pg2> As the network size increases, the number of possible network configurations grows combinatorially, significantly elevating the problem and leading to a severe curse of dimensionality <ref:2606. p. 025480#pg2>. Existing schemes in the literature addressed the energy efficiency maximization problem by decomposing multi-variable formulations into sequential convex subproblems evaluating trajectory, power allocation, and time-slot assignments <ref:2606. p. 125480#pg2> Traditional tabular and standard deep multi-agent reinforcement learning (MARL) frameworks encounter a severe curse of dimensionality as the network expands, demanding decentralized architectures to maintain real-time decision-making capabilities <ref:2606. p. 125480#pg2>.

How the training signal is built

The framework decouples the problem into a sequential two-stage pipeline: (i) offline global state-space compression via quantum-inspired optimization, and (ii) online localized trajectory tracking handled by decentralized MARL <ref:2606.25480#pg2>. To evaluate the topological fitness of any multi-UAV spatial deployment layout, a communicationcentric, rate-aware cost objective function J is constructed that maps physical wireless channel constraints and co-channel interference characteristics into an energy minimization landscape <ref:2606. p. 325480#pg2>. This cost function is defined as J(qn) = − (wprRpri + wnrRnr) + λ · ¯IdB <ref:2606. p. 325480#pg2>.

Where the gains come from

The core innovation lies in the RA-QAGC scheme, which proposes a global, quantum-inspired probabilistic annealing mechanism guided by J to prune and compress the high-dimensional deployment space into high-reward discrete waypoint candidate set C <ref:2606.25480#pg2>. This process identifies optimal spatial coordinates corresponding to highrate, low-interference corridors <ref:2606. p. 125480#pg2>. The condensed graph G = (C, E) is then used to allow decentralized Independent Q-Learning (IQL) agents to operate with high efficiency <ref:2606. p. 325480#pg2>.

Performance Validation

The simulation results show that the RA-QAGC framework outperformed over existing schemes by achieving 59.4 Mbps total throughput and 23.9 Mbps priority-user throughput <ref:2606.25480#pg2>. The proposed scheme achieves the highest total throughput of 59. p. 64 Mbps, outperforming all benchmark methods <ref:2606. p. 625480#pg2>. Furthermore, the priority user achieved a median data rate of approximately 0. p. 585 Mbps, whereas normal users achieve around 0. p. 555 Mbps <ref:2606. p. 025480#pg2>. The learned trajectories exhibit several intelligent behaviors, such as UAVs maintaining coordinated spacing to minimize interference <ref:2606. p. 625480#pg2>.

Conclusion

The work demonstrates that intelligent state-space reduction can substantially improve the practicality of multiUAV trajectory optimization in interference-limited wireless networks <ref:2606.25480#pg2>. The results indicate that incorporating rate-awareness into the environment abstraction process enables UAVs to identify more favorable operating regions and utilize network resources more efficiently <ref:2606. p. 625480#pg2>. The decentralized learning strategy provides a scalable alternative to centralized optimization, whose complexity rapidly increases with network size <ref:2606. p. 625480#pg2>. Future research will focus on extending the framework to dynamic user mobility scenarios, continuous control models, and large-scale heterogeneous aerial networks <ref:2606. p. 625480#pg2>.

REFERENCES

arXiv:2606.25480v1 [cs. p. 0MA] 24 Jun 2026

--- Page 1 ---

Received XX Month, XXXX; revised XX Month, XXXX; accepted XX Month, XXXX; Date of publication XX Month, XXXX; date of current version April, 2026.

Digital Object Identifier 10. p. 01109/OJVT.2026.xxxx

Rate-Aware Quantum-Inspired Trajectory Learning for Interference-Limited Multi-UAV Networks

Khaoula Khaled∗, Muhammad Afaq †,Ali Arshad Nasir‡, Zeeshan Kaleem§, Senior Member, IEEE 1∗ Department of Computer Engineering, King Fahd University of Petroleum & Minerals (KFUPM), Dhahran 31261, Saudi Arabia 2† Computer Engineering Department and IRC for Intelligent Secure Systems, King Fahd University of Petroleum and Minerals (KFUPM), Dhahran 31261, Saudi Arabia 3‡ Interdisciplinary Research Center for Communication Systems and Sensing (IRC-CSS), Department of Electrical Engineering, King Fahd University of Petroleum and Minerals (KFUPM), Dhahran 31261, Saudi Arabia 4§ Computer Engineering Department and Interdisciplinary Research Center for Smart Mobility and Logistics, King Fahd University of Petroleum & Minerals (KFUPM), Dhahran 31261, Saudi Arabia Corresponding author: Zeeshan Kaleem (email: zeeshankaleem@gmail.com). The authors at King Fahd University of Petroleum & Minerals (KFUPM) would like to acknowledge the support provided by the Deanship of Research (DoR). p. 0 ABSTRACT Unmanned aerial vehicle (UAV) can provide on-demand, high-capacity connectivity in disaster and normal situation. p. 0 However, it faces a challenge of curse of dimensionality in trajectory optimization, where interference-limited environments and vast search spaces make real-time coordination computationally expensive. p. 0 To overcome this challenge, we propose the Rate-Aware Quantum-Annealed Graph Condensation (RA-QAGC) scheme, which combines rate-aware graph abstraction with decentralized reinforcement learning to enable scalable, interference-aware UAV coordination By identifying high throughput locations and guiding UAV trajectory adaptation toward throughput-optimal regions, RA-QAGC effectively balances network capacity by maintaining quality-of-service (QoS) requirements. p. 0 Simulation results demonstrate the proposal outperformed over existing schemes by achieving 59.4 Mbps total throughput and 23.9 Mbps priority-user throughput, representing gains of approximately 15% and 34%, respectively, over the baseline schemes. p. 0 INDEX TERMS Unmanned Aerial Vehicles (UAVs), quantum-inspired computing, trajectory optimization, independent Q-learning, reinforcement learning. p. 0

--- Page 2 ---

Khoula et al.: Quantum-Aware Trajectory Optimization geometric clustering approaches, such as K-means, have been adopted to compress the environmental state space Nevertheless, their ability to meet QoS requirements is limited, as they neglect the dynamics of real-time wireless channel conditions. p. 1 Moreover, frameworks that jointly optimize trajectory design, power allocation, and user association lead to highly non-convex optimization problems, posing substantial computational challenges. p. 1 Existing schemes in the literature addressed the energy efficiency maximization problem by decomposing multi-variable formulations into sequential convex subproblems evaluating trajectory, power allocation, and time-slot assignments. p. 1 Similarly, Successive convex approximation and block coordinate descent techniques have also been extensively deployed in wirelesspowered communication networks [8], however, these approaches require intensive, centralized iterative computations that scale poorly with network size. p. 1 In literature, clustering based approaches has also been adopted to compress large user spaces into discrete UAV service zones [9]. p. 1 These schemes still have a challenge even after simplifying the spatial problem, as they remain channel-blind. p. 1 They calculate spatial centroids based strictly on Euclidean distances, completely ignoring vital wireless channel parameters and interferences. p. 1 Despite of their success, traditional tabular and standard deep multi-agent reinforcement learning (MARL) frameworks encounter a severe curse of dimensionality as the network expands, demanding decentralized architectures to maintain real-time decision-making capabilities. p. 1 They also have limitation as they struggle with memory constraints and fail to generalize across unseen states when facing continuous high-dimensional actions. p. 1 To break the scalability barriers of classical optimization and deep MARL search spaces, quantum computing paradigms, specifically Quantum Annealing (QA) and Quantum-Inspired Optimization (QIO) have emerged as highly efficient alternatives for continuous and combinatorial problems. p. 1 For instance, QA has been successfully applied to satellite communication systems to resolve beam placement and frequency assignments by formulating hybrid quantum-classical Ising pipelines that outperform standard commercial optimization solvers. p. 1 In aerial networks, quantum annealing (QA)-based approaches have been investigated for sum-rate maximization through the joint optimization of user clustering, subchannel assignment, and power allocation. p. 1 Similarly, [12] addressed joint UAV trajectory design and radio resource allocation by formulating the problem as a Markov decision process and employing a non-iterative cooperative optimization strategy to obtain high-quality solutions. p. 1 Although these studies apply QIO or integrate quantum concepts into reinforcement learning frameworks, their focus remains on trajectory optimization and resource allocation. p. 1 The use of QIO for reducing the state space prior to learning has received limited attention. p. 1 In particular, existing works do not exploit quantum-inspired probabilistic annealing together with a rate-aware condensation objective to construct a compact representation of the search space, which can improve the scalability of multi-agent reinforcement learning for trajectory optimization. p. 1

--- Page 3 ---

We consider the uplink of a multi-UAV-assisted wireless network spanning a continuous terrestrial region A ⊂ R squared. The network comprises a set N = N 2 of N 2 UAVs acting as Aerial Base Stations (ABSs) to service a set K = K stationary ground users (GUs). p. 2 Each ABS n ∈ N flies at a constant altitude hn. p. 2 At any discrete time slot t ∈ T, where T denotes the finite mission horizon, the 3D coordinate vector of ABS n is defined as un[t] = [xn[t], yn[t], hn]⊤, where qn[t] = [xn[t], yn[t]]⊤ ∈ A denotes its time-varying horizontal position. p. 2 The stationary GUs possess fixed coordinates wk = [xk, yk, 0]⊤, ∀k ∈ K. p. 2 To implement QoS, K is partitioned into two mutually exclusive subsets: priority users (Kpr) requiring strict data rate guarantees, and normal users (Knr = Ksetminus Kpr) processing standard best-effort traffic. p. 2 To mitigate the curse of dimensionality inherent in continuous trajectory optimization, the spatial search domain is mapped to a finite set of M discrete candidate waypoints (centroids), precomputed via the RA-QAGC framework as C = c m ∈ R 2M m=1. p. 2 Consequently, the horizontal positioning of each agent must satisfy qn[t] ∈ C, ∀n ∈ N, ∀t. p. 2 Spatial state transitions between successive time slots qn[t] → qn[t + 1] are governed by a global connectivity graph G = (C, E), where a directional transition edge (c m, c m′) ∈ E exists if and only if:∥c m − c m′∥2 ≤ vmax∆t, where vmax is the maximum horizontal velocity of the UAV, and ∆t is the discrete slot duration. p. 2 The time-varying Euclidean distance dk,n[t] and the corresponding elevation angle θk,n[t] between terrestrial GU k and flying ABS n are expressed respectively as dk n[t] =∥un[t] − wk∥2, θk n[t] = arcsin hn/dk n[t]. p. 2 Let θ◦ k,n[t] = 180/π θk,n[t] represent the elevation angle in degrees. p. 2 The probability of establishing a Line-of-Sight (LoS) link follows a standard sigmoidal logistic distribution as PLoS,k,n[t] = 1/(1+a exp(−b[θ◦ k,n[t]-a]), where a and b are environmental constant parameters dependent on the urban topology, and the corresponding Non-LoS (NLoS) probability is PNLoS,k,n[t] = 1 − PLoS k,n[t]. p. 2 The aggregate large-scale effective path loss (in dB) is represented as Leff,k,n[t] = K0 + 10α log10(dk n[t])+χLoSPLoS k,n[t] + χNLoSPNLoS k,n[t], (1) where K0 = 20 log10(4πfc/c) represents the freespace path loss at a one meter reference distance for carrier frequency fc. p. 2 Here, α denotes the environmentspecific path loss exponent, and χLoS, χNLoS are statedependent excess path loss variables assigned to LoS and NLoS conditions. p. 2 Compounding large-scale attenuation with exponential small-scale Rayleigh fading variance fk,n[t]2 ∼ exp(1) yields the total linear channel gain: gk n[t] = 10−Leff,k,n[t]/10 · fk n[t]2. p. 2 To maintain energy efficiency, GUs implement a fractional open-loop power control mechanism. p. 2 The uplink transmission power of ground user k is dynamically bounded in accordance with: Ptx k[t] = min Pmax, P0 + αOLLeff,k,nˆk[t][t] + 10 log10(NRB). p. 2 Terrestrial users associate with ABS that provides the maximum average received signal power computed as nˆk[t] = arg max n∈N (ptx k[t] · gk nˆk[t]), where ptx k[t] is the linear equivalent wattage of the logarithmic value Ptx k[t]. p. 2 Under a full frequency reuse, the cumulative co-channel inter-cell interference power In,k[t] = P j∈K,j! p. 2=k ptx j [t] · gj n[t], and the resulting received SINR γk[t] for GU k are γk[t] = ptx k[t]·gk nˆk[t][t]/σ2+Inˆk[t], k[t]. p. 2 we assume an equal-sharing intra-cell bandwidth allocation policy among associated users, the achievable uplink data rate for GU k associated with its target ABS is defined as Rk[t] = BKnˆk[t][t] · log2(1 + γk[t]), where Kn[t] = P j∈K ⊮nˆj [t] = n represents the instantaneous user association cardinality of ABS n, evaluated using the indicator function ⊮[. p. 2 To enforce priority, the aggregate network utility T [t] is formulated as a weighted sum-rate as T [t] = wpr P k∈Kpr Rk[t] + wnr P k∈Knr Rk[t], where wpr and wnr denote the priority and normal tier weights respectively, satisfying wpr ≫ wnr. p. 2 The primary network objective is to maximize this utility over the global mission horizon. p. 2

--- Page 4 ---

In this section, we formulated the weighted sum-rate maximization problem by jointly optimizing the UAV trajectory and resource allocation over a finite operational mission horizon T, represented as max

X T t=1 (wpr X i∈Kpr R i [t] + wnr X j∈Knr R j [t]) (2) s.t. Xmin ≤ xn[t] ≤ Xmax, Ymin ≤ yn[t] ≤ Ymax, (3)∥qn[t + 1] − qn[t]∥2 ≤ vmax∆t, ∀n ∈ N, t < T, (4)qn[t] ∈ C, ∀n ∈ N, ∀t, (5) (qn[t], qn[t + 1]) ∈ E, ∀n ∈ N, t < T, (6) pmin ≤ ptx k[t] ≤ pmax, ∀k ∈ K, ∀t, (7) X i∈Kn[t] Bi [t] ≤ B, ∀n ∈ N, ∀t. p. 2

Constraint (3) specifies that the ABS horizontal trajectories must remain strictly bounded within the designated terrestrial deployment region A. p. 2 Kinematic step bounds are enforced by (4). p. 2 Constraints (5) and (6) enforce discrete waypoint restrictions and graph topology adherence during the trajectory tracking phase. p. 2 Here, the ABS positions are restricted to the pre-screened condensed waypoint matrix C, and their spatial state transitions must map across connected geometric edges defined within the valid physical edge domain E of the graph GC. p. 3

--- Page 5 ---

To minimize this non-convex cost function without becoming trapped in sub-optimal configurations, a configuration state s = [s1, s2, s3]⊤ tracks coordinate indices pointing to a pre-screened candidate grid Ncand = 200. A candidate state s′ is proposed by selecting nswap = 2 random agents and modifying their positions:

s'n = (randi(Ncand), n ∈ Rswap sn, otherwise. This state transition exploration is governed by an exponential annealing cooling trajectory: Tκ+1 = ρ · Tκ, T0 = 1. p. 30 (12) where the cooling index ρ = 0. p. 39985 sustains the exploration behavior over a terminal iteration boundary of Imax = 3000, reaching a terminal temperature floor of TImax ≈ 0. p. 30119. The state transition couples classical Metropolis metrics with a quantum-inspired probabilistic floor: Paccept = max exp −∆J T ptunnel · T T0 (13) where ∆J = J(s′) − J(s). p. 3…

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Quantum-Inspired State Space Reduction for Trajectory Optimization: The RA-QAGC scheme uses a global, quantum-inspired probabilistic annealing mechanism guided by J to prune and compress the high-dimensional deployment space into high-reward discrete waypoint candidate set C. This allows the system to reduce the search space from continuous coordinates to a manageable set of 100 discrete waypoints before trajectory planning begins.

  2. Rate-Aware Cost Function for Interference Management: The cost function J maps physical wireless channel constraints and co-channel interference characteristics into an energy minimization landscape, where the term IdB = 1/K P i=1 10 log10(In i,i + epsilon) quantifies the time-averaged geometric system interference expressed in decibels. This ensures that the optimization actively suppresses severe co-channel self-interference loops by penalizing high interference levels.

  3. Decentralized Independent Q-Learning for Real-Time Adaptation: The framework employs a decentralized IQL framework where each agent n in N maintains an independent action-value function Qn(sn, an) over localized features. This structure allows agents to perform online continuous trajectory tracking efficiently while mitigating the curse of dimensionality by operating with a computational profile of O(M 2) per agent instead of O(MN).

  4. Joint Throughput and QoS Maximization: The system maximizes the utility T[t] defined as "wpr P k in Kpr Rk[t] + wnr P k in Knr Rk[t], satisfying wpr >> wnr." This capability enables the UAVs to dynamically adjust their trajectories and service regions to maintain strict data rate guarantees for priority users while maximizing overall network throughput.

  5. Environment-Aware Spatial Clustering: The method is environment-aware spatial clustering schemes unlike the conventional K-means heuristics by mapping complex, physics-driven wireless channel behaviors directly into the clustering routine. This bypasses channel-blind geometric limitations by using a cost function that balances throughput and interference instead of just minimizing Euclidean centers.

Sources

Related papers