Search Results (569)

The Q-Learning Relational Game (QLRG) framework provides a theoretically rigorous method for identifying minimal winning coalitions in online social networks (OSNs) under the restrictive assumption of global agent symmetry or uniform matroid structure. Real-world OSNs, however, exhibit significant asymmetry. This paper introduces the Cluster-Based Q-Learning Relational Game (C-QLRG), a practical extension that relaxes the global symmetry requirement by leveraging community structure. We partition the agent set into communities with bounded internal variation and represent the state solely by community membership counts of the seed set. Because the closure operator already captures all eventual influence spread, the problem reduces to a sequential seed selection task where the agent decides, at each step, from which community to add the next seed. We prove that the optimal Q-function of a suitably regularized reach-efficiency objective is Lipschitz continuous and derive a performance bound for the learned policy. The full algorithm is presented, and its complexity is analyzed. Empirical evaluations on a synthetic asymmetric network and Zachary’s Karate Club demonstrate that C-QLRG is highly sensitive to reward parameters, where default settings lead to premature stopping, but parameter tuning combined with a corrected minimality verification recovers high-efficiency coalitions by removing non-contributing agents. With tuned parameters, C-QLRG produces a near-winning coalition of size 11 and 99% reach on the synthetic network, surpassing the greedy baseline’s efficiency (size 12) despite a one-node coverage gap, while identifying the optimal winning coalition of size 1 on the Karate Club dataset, matching all baselines. The framework thus offers a principled trade-off between model fidelity and scalability, with the reward design choice being critical for practical deployment. Full article

Resource allocation for cognitive Vehicle-to-Everything (V2X) networks is challenging due to dynamic spectrum sharing, strong interference coupling, and stringent latency constraints for safety-critical Vehicle-to-Vehicle (V2V) traffic. Although recent Multi-Agent Reinforcement Learning (MARL) approaches report promising gains, many evaluations are conducted at limited and fixed network scales, which restricts insights into scalability under dense spectrum reuse. This paper investigates cooperative multi-agent learning for interference-aware and deadline-constrained V2X resource management. We propose a Q-value Transformation (QTRAN)-based value decomposition framework under centralized training with decentralized execution (CTDE) for joint resource-block and power allocation among V2V agents. The proposed approach is implemented in a realistic V2V/V2I simulator incorporating Manhattan grid mobility, fast fading, explicit cross-tier and co-channel interference, and per-link payload/deadline dynamics. Beyond communication-level performance, improved timely delivery of V2V safety messages can support cooperative maneuvering, collision avoidance, platooning, and infrastructure-assisted traffic management. Extensive simulations across varying numbers of V2V agents benchmark QTRAN against independent learning baselines including MARL and centralized single-agent learning (SARL). Results show that QTRAN improves performance compared with the selected learning baselines and enhances the throughput–reliability trade-off under interference-coupled spectrum reuse. For instance, at $N_{\mathrm{V}2\mathrm{V}}=20$, QTRAN achieves a V2V rate of $0.194\pm 0.004$ and a V2I rate of $9.117\pm 0.213$, while reaching a V2V success rate of $0.812\pm 0.017$ with a low Deadline Miss Ratio of $0.001\pm 0.000$. At higher density ($N_{\mathrm{V}2\mathrm{V}}=50$), QTRAN sustains strong reliability (V2V success rate of $0.719\pm 0.006$ and Completion Ratio of $0.716\pm 0.006$) while maintaining competitive infrastructure throughput (V2I rate of $9.251\pm 0.114$). These results indicate that QTRAN effectively captures non-linear interference interactions, enabling coordinated decentralized spectrum and power decisions under the adopted density-based evaluation setting, thereby enhancing V2V reliability and throughput in cognitive Internet of Vehicles. Full article

Existing cross-chain bridges over-rely on static collateralization and post-event penalties, leaving them vulnerable to concealed on–off attacks and rational group collusion. To address these limitations, this paper proposes a Dynamic Trust Evaluation and Risk Control (DTERC) mechanism for heterogeneous cross-chain relay nodes. First, DTERC develops a multidimensional trust quantification model that combines temporal decay, robust multi-observer latency aggregation, verification accuracy, online stability, and an asymmetric one-strike penalty triggered only by cryptographic evidence. Second, DTERC constructs a threshold-aware N-player evolutionary game model to characterize the k-of-N signature structure of cross-chain relay consensus and introduces a dynamic staking function to reduce the economic incentive for collusion under bounded attack-value and parameter conditions. Third, DTERC designs a threshold-preserving FastPath mechanism to reduce redundant verification for low-risk transactions while retaining committee-level confirmation and challenge-based fallback. The empirical evaluation combines multi-agent simulation, smart-contract prototype testing, whitelist-compromise stress tests, malicious-oracle robustness analysis, network-jitter experiments, repeated trials, and parameter-sensitivity analysis. The results show that, under the tested settings, DTERC reduces the malicious transaction success rate to 0.15% under a 50% initial collusion scenario, lowers core contract Gas overhead by 35.7%, and reduces average end-to-end latency by approximately 10% in benign FastPath conditions. These findings indicate that DTERC improves the security–efficiency trade-off of heterogeneous cross-chain relay networks while making its assumptions and limitations explicit. Full article

Dengue virus infection is a febrile illness caused by the Orthoflavivirus Dengue, which is transmitted by the mosquitoes Aedes aegypti or Aedes albopictus. Despite the fact that Dengue virus (DENV) is present in tropical and subtropical areas, climate change with global warming has been associated with the spread of Aedes aegypti and Aedes albopictus mosquitoes in several other regions worldwide. Notably, as the presence of Aedes albopictus has been confirmed in Southern Europe, already locally transmitted cases of Dengue virus infection have been reported in Europe. Apart from Europe, Australia has reported DENV cases in the 21st century that have been associated with the transmission of Aedes aegypti in the neighboring islands. Climate change, namely increasing temperatures, higher humidity and rainfalls, together with the development of urban heat islands, uncontrollable deforestation and urbanization, travelling and trade, has contributed significantly to the spread of DENV infection. Modern diagnosis based upon the advent of “multi-omics” techniques and machinery learning programs will be of the utmost importance for the early and accurate diagnosis of DENV infection. Finally, preventive measures for controlling Dengue virus infection, such as the use of repellents, educational programs, and improvement in water storage and waste management at the community levels would be very useful. Regarding climate change, the One Health Approach by integrating collaboration of various sectors and raising public awareness seems to be of the utmost importance in this context. Further investigations regarding the development of antiviral agents and vaccines will be an important asset in our armamentarium against DENV infection. Full article

Precision Livestock Farming (PLF) requires real-time control systems that connect high-level Decision Support Systems with resource-constrained edge devices. This paper presents a hybrid Multi-Agent System (MAS) architecture for swine precision feeding designed to address the trade-off between semantic interoperability and real-time operational efficiency. The proposed Controlling Module uses a dual-layer communication strategy: a lightweight character-delimited TCP/IP protocol ensures deterministic performance for embedded controllers, while an XML-serialized format that maps to the FIPA Agent Communication Language preserves semantic interoperability. A custom serialization/deserialization algorithm was developed to process this XML structure within LabVIEW while avoiding the overhead typically associated with generic DOM/SAX parsers. The architecture was validated in a 120 h laboratory test that combined a Digital Twin simulation of 50 virtual feeders with Hardware-in-the-Loop testing of key sensing components. Under these test conditions, no communication failures were observed, all simulated network interruptions were recovered from, and the system operated with a modest resource footprint, including an average CPU use of 15% and a peak memory use of 350 MB. The platform also processed 2590 consumption events without reported data loss during the validation period. These results indicate that the proposed hybrid MAS architecture is a feasible solution for integrating interoperable decision support and deterministic edge control in PLF applications. Full article

When operating in complex and obstacle-dense environments, micro UAV swarms often face severe cooperative positioning failures due to transient non-line-of-sight (NLOS) interference and cascaded inertial sensor drift. To address this, this work proposes a fault-tolerant positioning framework integrating multi-agent deep reinforcement learning with cooperative extended Kalman filtering (MADRL-CEKF). The system incorporates a link-level dynamic soft isolation mechanism that dynamically adjusts observation covariance to effectively sever paths of cooperative error contagion. An adaptive Markov smoothing constraint is mathematically embedded to mitigate high-frequency control jitter typical of AI-driven policies. Crucially, the framework implements a resource-aware multi-objective reward architecture tailored for micro UAVs. Evaluated through high-fidelity simulations and offline physical datasets, the proposed framework achieves a 96.01% reduction in average tracking error (RMSE) under extreme multi-node cascaded failures, completely preventing system divergence. Furthermore, through autonomous multi-objective trade-offs, the system reduces processing delay by 44% (to 25.1 ms) and computational energy consumption by 41% with only a marginal accuracy compromise of 0.16 m, strictly keeping the execution time within the 50 ms real-time threshold. The MADRL-CEKF framework effectively bridges the gap between sophisticated AI decision-making and strict engineering constraints, providing a highly robust and resource-efficient navigation paradigm for swarm robotics. Full article

Reinforcement learning agents for financial trading typically optimize reward functions that directly map profit and loss to learning signals, without accounting for the agent’s own decision certainty. This paper investigates whether modulating reward signals by a confidence estimate, without modifying network architecture, training procedures, or data pipelines, can meaningfully improve trading performance. We formalize five lightweight confidence estimation methods, each targeting a distinct uncertainty dimension: critic agreement (value estimation), temporal direction consistency (behavioral stability), state novelty (distributional familiarity), action magnitude stability (position sizing), and state-transition surprise (environmental predictability). Using a Twin Delayed Deep Deterministic Policy Gradient agent trained on hourly OHLCV data for Bitcoin, Litecoin, and Ethereum over five years encompassing diverse market regimes, we conduct a controlled experiment in which the confidence method is the sole variable across 18 experimental conditions. State novelty achieves the strongest improvement, raising mean test-period ROI from 5.7% to 24.9%, increasing Sharpe ratio (SR) from 0.34 to 1.57, and reducing maximum drawdown from 28.0% to 15.0% across the three cryptocurrencies. Four of the five methods reach statistical significance at $p<0.05$ on all assets; only state-transition surprise, the sole method requiring an auxiliary network, fails to distinguish itself from the baseline due to signal saturation. The proposed confidence-aware reward-shaping framework is plug-and-play, algorithm-agnostic, and directly applicable to other RL-based trading systems. Full article

The beef production chain plays a strategic role in Brazilian and global agri-food systems and faces growing demands for sustainability, transparency, and traceability. Building on official Animal Transit Guide (GTA) records from Mato Grosso do Sul, Brazil, this study examines whether a parsimonious agent-based model (ABM) can generate the main structural signatures of an observed cattle-trade network. The empirical benchmark is a directed and weighted network with 20,827 nodes and 258,120 weighted edges. The ABM represents producers and slaughterhouses as spatial agents connected by trade decisions based on three mechanisms: destination attractiveness, defined as the accumulated pull of a slaughterhouse based on previous simulated throughput; geographic distance, representing spatial friction; and relational memory, representing the tendency to repeat previous commercial ties. Producer choice is formalized through a local utility function that combines attractiveness, distance penalty, and relational memory under capacity, sourcing-radius, and saturation constraints. In the simulated scenarios, the top-five slaughterhouses accounted for 38.49 ± 2.56% of throughput at reduced scale and 14.40 ± 0.65% at intermediate scale, while weighted mean distances were 11.94 ± 0.56 and 9.07 ± 0.39 model units, respectively. The model reproduced, in structural and mechanistic terms, the emergence of dominant hubs, the concentration of flows, and the bounded increase in transaction distance with connectivity around the empirical threshold of kw ≈ 256. Sensitivity analyses indicated that attractiveness increases concentration, distance localizes transactions, and relational memory can stabilize repeated ties when recurrent activation is represented. Rather than reconstructing individual transactions, estimating policy impacts, or identifying a unique parameter vector, the model provides a generative explanation of how local trade rules can produce macro-level network patterns consistent with the observed cattle-trade regime. These findings support future prospective analyses of cattle governance, traceability, and sustainability within the broader context of Livestock 4.0. Full article

The proliferation of digital financial news offers unprecedented opportunities for automated analysis of market-moving events. This paper presents a framework for event-based sentiment analysis of financial news that leverages Large Language Models (LLMs). The approach brings together three complementary ideas: Retrieval-Augmented Generation (RAG) for contextual enhancement, Graph Neural Networks (GNNs) for modeling relationships between events, and a multi-agent ensemble for orchestrated reasoning. The methodology targets well-known difficulties in financial text processing, including domain-specific terminology, implicit event detection, and temporal reasoning, and it combines transformer-based event extraction with sentiment classification enhanced by external knowledge retrieval. We evaluate six model configurations on an aggregated corpus of 14,851 financial news samples. On the event-detection task, every configuration reaches a weighted F1-score of 100%; we show that this is a ceiling effect produced by a binary event/no-event formulation over a highly imbalanced dataset rather than evidence of a difficult problem being solved, and we discuss what it implies for how such systems should be evaluated. On three-way sentiment classification, the strongest configuration—the multi-agent ensemble—reaches 87.4% accuracy, narrowly ahead of a RoBERTa (Robustly Optimized BERT Pretraining Approach) baseline at 87.2%; however, because the gaps reported between models are small and we did not run significance testing, we report them as indicative rather than definitive. The GNN component is described as part of the proposed design, but it has not yet been validated experimentally, and we state this limitation explicitly. The framework produces interpretable, structured outputs suited to downstream use in algorithmic trading, risk assessment, and investment decision support, and the paper contributes a reusable financial NLP pipeline together with a candid account of where the current evidence is, and is not, convincing. Full article

This study investigates how structural properties of dependence networks and trust-based partner selection jointly shape cooperation in multi-agent societies. We develop a simulation framework grounded in a Block World environment in which agents pursue their goals under resource scarcity, heterogeneous competence, bounded network visibility, and explicit monetary exchange. Dependence emerges dynamically at the subgoal level, while trust thresholds regulate access to potential partners within structurally constrained influence zones. Results reveal a structural trade-off between effectiveness and efficiency: permissive trust strategies increase total utility, but at the cost of higher expenditures and lower utility-to-investment ratios. Conversely, more selective strategies reduce task completion but enhance economic sustainability. On the other hand, trustworthiness influences economic redistribution effects, representing a concrete advantage for reliable agents. These findings highlight the multidimensional nature of performance in dependence networks and provide theoretical and practical insights for scalable hybrid human–AI systems. Full article

To meet the goals of the Paris Agreement of avoiding dangerous climate change, decarbonisation of the global economy needs to proceed around three to five times faster over the coming decade than over the past two decades. This poses a great challenge for policy. Carbon pricing has often been put forward as the most efficient, or cost-effective, policy for achieving decarbonisation. This paper uses a stylised agent-based model to investigate whether implementing non-equilibrium dynamics and endogenous innovation results in more effective emission reductions for carbon tax compared with emission trading schemes. We find that the implementation of a carbon price is not policy-agnostic and that a carbon tax achieves faster emissions reduction, lower cumulative emissions, and lower cumulative (potentially wasted) investment in fossil fuel assets than a cap-and-trade policy with the same average carbon price. While a comparison between carbon pricing and alternative policies is outside the scope of this paper, we consider the broader policy implications that may be drawn from a new theoretical explanation for the difference in performance of the alternative carbon pricing approaches, and suggest that the traditional view that policy should aim to minimise the marginal emissions abatement cost is mistaken. Full article

The surging need for public security risk mitigation has placed stricter demands on the modernization of emergency response capacities. Unmanned aircraft systems (UASs) offer a promising solution for liquid anti-riot agent dispersion, yet the complex interaction between rotor-induced downwash and droplet trajectories makes nozzle layout optimization a significant challenge. To address the prohibitive computational costs of traditional Computational Fluid Dynamics (CFD) and the limitations of single-objective optimization, this study proposes an integrated “simulation–modeling–optimization–decision” framework. First, a linear nozzle layout was identified as superior to the traditional circular arrangement, achieving a 44.8% increase in deposition rate. Subsequently, Optimal Latin Hypercube Sampling (OLHS) and CFD simulations were combined to construct high-precision Kriging surrogate models for three key indicators: deposition rate, uniformity, and coverage rate. The NSGA-II algorithm was then employed to solve the multi-objective trade-off, followed by the entropy-weighted TOPSIS method to identify the optimal engineering solution. Results indicate that nozzle count is the dominant system-level variable under the constant per-nozzle flow-rate condition, showing strong positive correlations with all performance indicators. The identified optimal configuration (6 nozzles with a 1.88 m boom length) achieved a 66.1% increase in deposition rate and an 18.7% increase in coverage rate compared to the original circular layout. Furthermore, the surrogate-based framework improved optimization efficiency to 296% compared to full factorial methods. This study provides a scientific theoretical basis and a highly efficient technical pathway for the structural design of high-performance UAV spray systems. Full article

The Capacitated Quadratic Assignment Problem (CQAP) arises in logistics and network design, requiring the allocation of tasks to agents under quadratic interaction costs and capacity constraints. Classical exact solvers become computationally infeasible for large-scale instances, while heuristic methods such as Genetic Algorithms suffer from scalability limitations and sensitivity to local optima, leaving a gap for principled scalable approximations. In this paper, we address CQAP using the Gromov–Wasserstein (GW) framework, derived from Optimal Transport (OT) theory. In particular, we propose a multi-initialization GW strategy (GW_MultiInit) that mitigates the local optima problem inherent to non-convex GW optimization and scales efficiently to large problem sizes. Computational experiments on synthetic CQAP instances show that GW_MultiInit consistently achieves solutions close to the exact optimum for small- and medium-scale problems, and outperforms heuristic baselines such as the genetic algorithm at large scale in both runtime and solution quality across the benchmarks tested. To validate generalizability, we further evaluate GW_MultiInit On 17 QAPLIB benchmark instances adapted to the CQAP setting, GW_MultiInit achieves the best approximate result on 15 out of 17 instances with an average optimality gap of 0.34%, demonstrating strong generalizability beyond synthetic data. Additional comparisons with Entropic GW and Fused GW highlight practical trade-offs between accuracy, speed, and parameter sensitivity, offering guidelines for real-world deployment. Our results suggest that GW-based methods, and GW_MultiInit in particular, offer a promising and scalable approach for CQAP and related large-scale assignment problems within the problem scales examined. Full article

Internet of Things (IoT), sensor-network, and cyber-physical system (CPS) software increasingly relies on large language models (LLMs) and autonomous agents for code generation, maintenance, and vulnerability repair. However, LLM-generated edge services, telemetry APIs, configuration handlers, and data-aggregation routines can introduce SQL injection, path traversal, command injection, hard-coded credentials, and unsafe device-control logic, which may compromise sensing data integrity and system safety. Existing approaches largely rely on static post hoc analysis and lack a unified modeling of the generation process, making it difficult to achieve a principled trade-off between functionality and security. To address this challenge, we propose SafeCodeRL, a framework that integrates multi-agent collaboration with constrained reinforcement learning for trustworthy LLM-generated IoT/CPS software. SafeCodeRL models code generation as a security-aware sequential decision process, where Planner, Code, Security, Test, and Critic agents jointly optimize task decomposition, code synthesis, vulnerability auditing, and sandbox-based validation. We design a constraint-aware policy based on Proximal Policy Optimization, augmented with a Lagrangian mechanism and a shielding strategy to explicitly enforce security constraints. Experiments on real-world engineering and security benchmarks, including SWE-bench, SecurityEval, and CyberSecEval, show that SafeCodeRL reduces high-risk vulnerabilities by over 60% while maintaining high functional correctness. A scenario-level IoT/CPS case study further demonstrates that SafeCodeRL substantially improves secure pass rates for sensor telemetry, edge gateway, configuration-management, and data-aggregation tasks, providing a practical path toward trustworthy AI-assisted software development for sensor-driven systems. Full article

Prolonged downturns in the global economy have simultaneously increased banks’ credit risk exposures and intensified the need for effective liquidity management. This study develops a dynamic agent-based financial network comprising banks, depositors, firms, and the central bank to examine trade-offs in bank resolution under coupled liquidity and credit risks from the perspective of systemic stability. The simulation results show that, for liquidity risk management, when banks adopt the asset-sale strategy, both default probability and expected returns in the banking system exhibit a nonlinear pattern: they first decline and then rise as the asset depreciation ratio increases. Furthermore, at moderate levels of asset depreciation, the asset-sale strategy helps preserve heterogeneity within the banking system, thereby preventing excessive risk concentration, and performs better than the liability-expansion strategy. Regarding credit risk resolution, the debt-relief strategy significantly improves systemic stability, whereas the effectiveness of the debt-extension strategy depends critically on liquidity management conditions. Under liability-expansion scenarios, default risk initially declines but later rises as debt maturity is extended, whereas expected returns move in the opposite direction. Under asset-sale conditions, the debt-extension strategy enhances systemic stability only when the allowable number of debt extensions is sufficiently high. The analysis of strategic trade-offs indicates that combining the debt-relief strategy with the asset-sale strategy generates a positive synergistic effect and strengthens systemic resilience, whereas the interaction between the debt-extension and asset-sale strategies produces offsetting effects. These findings offer useful implications for banks and regulators in designing coordinated and adaptive frameworks for risk resolution and systemic stability. Full article