In current online economic dispatch problems of power systems, the cost function of each generation unit typically only includes time-varying generation costs, which does not account for the costs associated with current carbon emission market transactions and high-frequency communications. These factors need to be considered. Additionally, considering the lag in obtaining information about cost coefficients, feedback delay should be considered as a factor for the online optimization process of economic dispatch. This research addresses the challenge of online economic dispatch in the presence of delayed feedback and carbon emission expenses, proposing a decentralized, event-initiated algorithm for online optimization. To address these challenges, this paper proposes a decentralized, event-triggered online optimization algorithm tailored for Agentic AI in smart energy systems operating over an edge–cloud continuum. In each local optimization iteration, each node in the power system can access its local objective function with a delayed time sequence and updates its local decision-making behavior online based on information from multiple neighboring nodes, controlled by an event-triggering mechanism, to minimize the global cumulative generation cost and carbon emission cost. The research results show that, under the assumption that the time-varying balanced undirected communication topology remains connected, the designed online optimization algorithm ensures that the upper bound of static network regret grows sub-linearly, fundamentally related to feedback delays and event-triggering thresholds and scaled as O ( T ) . This work provides a foundational paradigm for Agentic AI in smart energy systems, bridging distributed optimization theory with practical constraints of modern edge-cloud infrastructures.
On-device deployment often relies on lightweight models that require fine-tuning when deployment environments differ from development conditions or when model performance changes over time. Although recent LLM-enabled agents are capable of reasoning about system states and forming adaptive strategies, they often incur substantial computational and budgetary overhead, making them impractical for resource-constrained edge settings. This paper presents Budgeted Agentic AI for Adaptive Lightweight Model Fine-Tuning (BA2), a framework for budget-constrained lightweight model adaptation that combines bounded optimization steps with auxiliary evaluation, rollback, and compression actions. BA2 enables agents to dynamically adjust adaptation strategies according to real-time budget conditions while maximizing system performance. BA2 explicitly accounts for constrained resources, including limited fine-tuning steps, tool invocation quotas, and token consumption. An Engineer-Manager architecture is introduced, where an LLM-based Engineer generates parameterized candidate operations from system logs, and a lightweight contextual-bandit Manager selects actions conditioned on both environmental states and remaining budgets. Experimental results demonstrate that BA2 achieves superior cost-performance trade-offs compared with static heuristics and inference-time search agents under tight budget constraints, while remaining competitive with strong baselines in higher-budget regimes.
Open Multi-Agent Systems (OMAS) are societies composed of autonomous agents that operate with diverse or aligned objectives. To enhance cooperation among these independently functioning agents, researchers have adapted the concept of social norms from human societies as regulatory mechanisms. However, social norms often exist as unwritten, unspoken rules that can only be acquired through observation and inference, a significant challenge for autonomous agents. This paper proposes a novel Dual-task Transformer architecture that enables agents to learn social norms by observing interactions in their environment. Our approach leverages the transformer's capability to capture long-range dependencies between observed actions, performing two simultaneous tasks: contextually classifying action sequences to predict punishments and rewards, and generating structured representations of the inferred social norms. The model is trained on a modified version of the Moral-Stories dataset, transformed to represent sequences of agent actions and their normative outcomes. We integrate this inference capability with rule-based systems to assist agents in their decision-making processes. Experiments conducted within the MESA agent-based modeling framework demonstrate that agents equipped with our dual-task inference model rapidly internalize learned social norms and effectively regulate their behavior, showing a significant reduction in norm-violating actions and an increase in the adoption of norm-compliant behavior. Our approach offers a cognitively inspired mechanism for norm learning in OMAS that is both extensible through fine-tuning and capable of capturing complex dependencies in observed social interactions.
Alzheimer's disease is a neurodegenerative disease that affects brain tissues, causing memory loss. The most common and primary symptoms of Alzheimer's are the complexity of remembering more information, behavioral changes, and a deep confusion about events, locations, and time. The detection of Alzheimer's disease is desirable for therapy and patient care. Still, it is difficult to handle heterogeneous and high-dimensional biomedical data for attaining a reliable prediction. To solve these complexities, this paper devises the Multi-Axis-Multi-Head Capsule Network (MaxMHCapsuleNet)-based detection and classification of Alzheimer's disease. Initially, image sharpening is performed using a Kriging-Weighted Laplacian kernel, and the Spatial-Channel Mamba-UNet (SCM-UNet) segments the affected region. Furthermore, the techniques, such as 90 degrees rotation and random masking, augment the images. In the feature extraction process, GoogLeNet and Visual Geometry Group Network (VGGNet) features are extracted. The Alzheimer's disease detection is done by MaxMHCapsuleNet, which detects the normal (non-demented) and abnormal classes. If the detected result is abnormal, multiclass classification is done by MaxMHCapsuleNet. In the classification process, the abnormal classes are classified into mild, moderate, and very mild demented stages. Moreover, the devised MaxMHCapsuleNet attains the accuracy, precision, and recall of 91.709%, 90.388%, and 92.036%.
In this research work, a novel multimodal data-based recommendation system is introduced to ultimately enhance student's talents and broaden knowledge without any interference. Initially, the required multimodal data, such as video, images, and text data are gathered from standard sources and it is given to the feature extraction phase. Here, the Bidirectional Encoder Representations from Transformers (BERT), Visual Geometry Group 16-based Autoencoder (VGG16-AE) and 3D VGG16-AE mechanisms are utilized for accurately recognizing and differentiating the significant feature sets such as, first set of features (f1), second set of features (f2) and third set of features (f3) from the input data. Then, the extracted three set of features are fed into the learning recommendation phase here, the Multi-scale Feature Fusion-based Residual Bidirectional Gated recurrent unit with Sparse Attention (MFF-RBG-SA) mechanism is designed to efficiently analyze and generate better solution in an earlier stage. Moreover, the overall performance of the designed approach is estimated with various performance measures and compared with conventional methods to guarantee its reliable effectiveness in the educational sector. In the validation phase, the designed approach has attained higher accuracy value of 95.35%, 95.28%, 95.38% and 95.33% in terms of Linear, Tanh, softmax and sigmoid activation functions, which demonstrate its superior learning recommendation performance than other conventional methods.
Recent advances in deep learning have enabled the development of highly powerful models, but these models are also extremely resource-intensive. This results in high financial costs, prolonged training times, and a significant environmental impact. Consequently, optimizing the training process is essential to mitigate these costs while maintaining or enhancing performance. In this context, software-level optimization plays a crucial role by improving training pipelines, memory management, and overall code efficiency. Fine-tuning hyperparameters, as part of this code optimization process, is the focus of this review, which examines how to effectively combine and adjust these parameters to boost overall model performance while reducing both costs and environmental impact.
This review provides a comprehensive analysis of cooperative Multi-Agent Reinforcement Learning (MARL) approaches for robotic systems, with particular emphasis on methodological foundations, practical implementations, and emerging challenges. We first examine the evolution of distributed intelligence in robotics, tracing its development from early architectures to modern learning-based frameworks. Our analysis focuses on two complementary paradigms: cooperative methods utilizing Centralized Training with Decentralized Execution (CTDE), and hierarchical approaches that address complexity through temporal and task decomposition. We systematically compare actor-critic methods, value-based approaches, and hierarchical frameworks across theoretical foundations, implementation characteristics, and application domains spanning aerial, ground, and maritime robotics. Our comparative analysis reveals important trade-offs between expressiveness, computational efficiency, and implementation complexity, highlighting that method selection must align with specific application requirements. Furthermore, we identify critical challenges, including the sim-to-real gap, scalability constraints, communication limitations, safety verification, and coordination in heterogeneous teams, mapping promising research directions to address these barriers to widespread deployment. This survey bridges theoretical understanding with practical implementation, providing a structured framework for researchers and practitioners working on multi-agent learning for advanced robotic systems.
Human Activity Recognition (HAR) is a challenging task that involves accurately classifying diverse daily movements from data captured by sensors, videos, or images. In this study, we propose a robust HAR framework that integrates CatBoost with a stacked ensemble learning (SEL) strategy, combining multiple base classifiers to enhance accuracy and generalization beyond conventional machine learning approaches. The framework was first evaluated on the benchmark WISDM, RealWorld and PAMAP2 datasets, comprising raw triaxial accelerometer signals segmented with a sliding window approach, demonstrating its effectiveness. The CatBoost model within the SEL framework achieved strong performance in identifying activities such as walking and jogging, while also delivering nearly perfect recognition for stair-related activities, with average scores of 87.06% accuracy, 89.25% recall, 79.93% precision, 84.26% F1-score, and 85.43% ROC-AUC across all WISDM activities. To assess generalization, the framework was further tested on the RealWorld HAR and PAMAP2 datasets. On RealWorld HAR, it achieved 99.2% accuracy, 99.06% recall, 99.23% precision, 99.13% F1-score, and 99.1% ROC-AUC, whereas on PAMAP2, it attained 99.43% accuracy, 99.33% recall, 99.53% precision, 99.43% F1-score, and 99.36% ROC-AUC. These results highlight the capability of ensemble learning combined with boosting methods to advance sensor-based HAR across multiple benchmark datasets, offering high reliability and generalization in real-world scenarios.
Video Surveillance is generally utilized in highways, residential zones, schools, and other public areas to monitor events happening in those areas, where detecting abnormal events in video surveillance effectively contributes to guaranteeing the safety of public areas. Although various methods have been created in this field, many unsolved issues remain, such as higher computational complexity, irrelevant features, and low learning capability, are exist in the existing methods, which limit them from obtaining an accurate abnormal event detection. Hence, a Supervised Incremental Learning based Multihead Attention Convolutional Network (SIL-MACoN) model is proposed in this research to detect the abnormal events accurately by eliminating the existing drawbacks. The unification of the Multihead Attention (MA) mechanism helps to increase the ability of the SIL-MACoN model to understand complex features by capturing the variances among the features by multiple heads. Moreover, the utilization of incremental and supervised contrastive learning mechanisms improves the MACoN model's learning capability and performance through updating its knowledge without forgetting the previously learned features and producing similar and dissimilar set features for training, respectively. The SIL-MACoN model attains 97.34% accuracy, 97.36% specificity, and 97.33% sensitivity with 90% of training data using the ShanghaiTech Campus Dataset, respectively.
Options markets are the platforms where traders buy/sell contracts, are inherently complicated and volatile, presenting difficulties for algorithmic trading systems. Conventional approaches, which rely on human evaluation of financial documents and charts, are frequently ineffective and prone to mistakes. To overcome this constraint, this study suggests an automated options trading system that uses tailored Proximal Policy Optimization (PPO). The system trains an RL agent incorporating time series data, technical indicators (Momentum Indicators, Level Indicators), and a risk-adjusted reward function defined as cumulative returns penalized for exceeding a 5% daily loss. According to our findings, the PPO-based trading strategy outperforms a number of current trading strategies, such as Moving Averages (MA), Relative Strength Index (RSI), Momentum Trading, and provides a success rate of 85%, indicating that total trades resulted in profit. This provides a robust environment for reviewing stock data, conducting trading simulations, and evaluating performance metrics. The core functionality leverages a Proximal Policy Optimization (PPO) model to forecast trading moves based on historical price movements. This highlights the significance of integrating technical indicators with systematic strategy assessment for reliable automated trading systems. Empirical results demonstrate a predictive accuracy of 85% for profitable trades using the PPO model, superior to state-of-the-art. This facilitates data-driven decision-making, enabling users to identify and implement optimal, risk-managed trading strategies.
The increasing reliance on smart grids to manage power distribution efficiently has introduced significant cybersecurity vulnerabilities due to their interconnected nature. Traditional security approaches often fall short in real-time protection, particularly against advanced threats such as data manipulation, unauthorized access, and Distributed Denial-of-Service (DDoS) attacks. This paper proposes a novel Smart Grid Secure Protocol (SGSP), integrating Attribute-Based Zero-Knowledge Proofs (AB-ZKP), Redundant Consensus Mechanisms combining Proof of Stake (PoS) and Practical Byzantine Fault Tolerance (PBFT) for scalable and fault-tolerant consensus, and Grid Safe Smart Contracts (GSSC) to enhance data confidentiality, automate security enforcement, and resist cyber threats. The AB-ZKP mechanism ensures selective attribute verification while preserving privacy and keeping sensitive data off-chain. The hybrid consensus mechanism merges energy-efficient PoS with fault-tolerant PBFT, securing the Blockchain layer against DDoS and Sybil attacks. Meanwhile, GSSCs automate transaction validation and policy enforcement, reducing human intervention and enabling real-time anomaly detection. Experimental results in a simulated environment demonstrate high resilience, improved data privacy (98.7% compliance), fast consensus (1.2 s), low energy consumption (0.09 kWh/transaction), and strong DDoS resistance (92.5/100). The proposed approach significantly outperforms traditional methods, paving the way for secure, scalable, and privacy-preserving smart grid ecosystems.
The increasing complexity and critical nature of air traffic control systems demand robust and scalable security mechanisms to ensure the safe and efficient management of airspace. This research proposes a novel approach that uses multi-agent systems with adaptive risk-aware access control for air traffic control systems. The framework incorporates real-time risk assessment, past behavior analysis, and current air traffic context to compute an agent's risk profile before granting access to the sensitive air traffic control data. This ensures that agents with higher risk are not granted restricted access to critical resources. Additionally, security features such as non-persistent connections and two factor authentication are incorporated to further enhance the system's resilience against the cyber threats. The proposed framework is evaluated through a case study, demonstrating that it significantly enhances the security of air traffic control systems compared to the traditional approaches. By ensuring that the access rights are dynamically adjusted based on an agent's behavior and current air traffic conditions, the system mitigates potential risks while maintaining operational efficiency.
Interaction protocols are commonly used in agent-based systems. They ensure good coordination between agents by proposing a specific message exchange pattern. However, these interaction protocols are not perfect; they need more extensions to offer, among others, better performance and scalability, mainly when tight deadlines are involved. In this case, participants often fail to answer some requests before their deadlines due to overload, bottlenecks, slow network, or being busy or blocked. Designing agents without considering this issue may decrease their sociability, which wastes valuable chances to obtain the best goals. The proposed approach uses the participant's experience to train supervised learning models to predict if the replies will reach initiators before deadlines or not, thereby enabling a prioritization mechanism for handling interaction requests more effectively. The proposed approach has been evaluated using multiple Contract Net interaction scenarios of two case studies under the JADE platform. The promising results show a significant increase in agents' sociability measured by a new metric that we have proposed called Sociability Degree via Interaction Protocols (SDIP) where it was maintained even when systems scale up in term of number of agents and initiated interactions.
In a cloud computing environment, the allocation of virtual machines for executing the user-submitted task is a challenging process. Specifically for large task sizes in the cloud environment, finding an optimal task scheduling solution is regarded as an NP-hard problem. Optimizing the tasks in a virtual machine's data center while reducing important, influential, and cost-effective parameters such as energy usage, makespan, and cost is the best course of action. Hence to minimize these parameters, an effective heuristic algorithm such as the dynamic predator-prey optimization technique is proposed to execute the task scheduling process in the cloud server that assists the virtual machine manager (VMM) to allocate the task in an optimal way that enhances and secures the performance. While the physical machine consumes more data and space the proposed method utilizes a Virtual Machine (VM) for task scheduling where the VMM holds the responsibility for scheduling the task for VMs. The dynamic predator-prey optimization is a behavioral combination of grey wolf optimization and sparrow search optimization where the foraging behavior assists in finding the solution in a large search space and the hunting behavior of grey wolf optimization is integrated. The overall combined behavior helps the VMM in scheduling the task to the correct VM considering the capacity and capability to perform the task. The dynamic predator-prey optimization method attains a makespan of 794.97 s, a throughput of 1 bps, a degree of imbalance is 0.180 and an Average Resource Utilization Ratio is 0.95. The result findings state that the dynamic predator-prey optimization obtains the best results in allocating the tasks for the VM.
The brain is an incredible example of biological engineering, managing human cognition, behavior, and perception. Autism Spectrum Disorder (ASD) emerges within this complex organ. It brings a variety of challenges and distinct traits that influence social interactions, communication, and sensory experiences. ASD is a developmental condition that appears in early childhood, impacting how individuals perceive and engage with their surroundings. Detecting ASD involves observing behavioral patterns, assessing communication skills, and recognizing specific behavioral traits. In this work, the ASD detection is done using the ResNeXt Convolutional Forward Harmonic Network (ResCFHN). At first, the input image is acquired from the specified dataset and is pre-processed using the Region of Interest (RoI) and Kuwahara filter techniques. Then, the extraction and identification of pivotal region is done using the Box Neighborhood Search Algorithm based on functional connectivity. Thereafter, the ResCFHN is used for performing ASD detection. The proposed ResCFHN is obtained by integrating ResNeXt with Convolutional Neural Network (CNN). The ResCFHN has achieved 91.094% of accuracy, with 90.774% of sensitivity and 92.879% of specificity.
Mobile agents offer numerous benefits in healthcare but encounter critical security challenges, including masquerading, unauthorized access, and data tampering, primarily due to reduced control and visibility during migration. To address these issues, this paper proposes a blockchain-based model to enhance the security of mobile agents in healthcare. The model integrates blockchain technology to secure mobile agents, and eliminate malicious agents, utilizing a smart contract to regulate communications between agents and the blockchain. The design, implementation, and evaluation of the model are detailed, with security tested against various attack scenarios. Experimental results demonstrate the model’s ability to monitor patients accurately with minimal latency and achieve a 100% detection rate for malicious agents. Detection times vary based on attack type, with a maximum average of 7400 ms for integrity breach detection.
This paper provides an extensive overview of recent advancements in Multi-Agent Systems, focusing on three principal areas: cooperative observation, cooperative control, and collaborative decision-making. First, the cooperative observation is highlighted for its utilization of distributed sensing and state estimation techniques, enabling agents to collectively perceive their surroundings with improved accuracy and resilience. Then, the cooperative control is emphasized for the significance of distributed algorithms and consensus protocols, which facilitate synchronized behaviours and system stability under dynamic and uncertain conditions. Next, the collaborative decision-making is discussed as an integration of cognitive models and adaptive strategies, empowering agents to make informed, real-time decisions as a group. By scrutinizing these aspects, the cognitive models like the Observe-Orient-Decide-Act (OODA) loop are used for a dynamic and adaptive decision-making in MAS. Two key aspects are discussed: 1) how the amalgamation of these techniques amplifies the performance, adaptability, and robustness of MAS in diverse applications such as robotics, cyber-physical systems, and intelligent transportation systems has been discussed, and 2) how the centrality of agent coordination and cooperation facilitates the evolution of MAS. Overall, the document emphasizes the importance of integrating cooperative observation, control, and decision-making for optimized MAS operation and identifies future research avenues that include refining these techniques and exploring their potential in emerging application domains.
Vehicle platooning is an active area of research in which vehicles travel closely together to improve fuel efficiency and safety, and relies heavily on the trustworthiness and cooperation of the participating vehicles. Traditional methods of managing trust in such systems face significant challenges, including vulnerability to malicious behavior and lack of transparency. Any such attack on the platoon from outside or inside will be disastrous as human lives are involved. Such attacks can be dealt with if we only allow trust worthy vehicles. In this research we propose a blockchain-based reputation model for vehicle platooning, designed to address these challenges and enhance the overall reliability and efficiency of platooning systems. Vehicles will only be allowed to join the platoon after calculating their reputation scores and these scores will be saved in a smart contract. This will ensure that every agent can reliably compute the score and no entity can compromise the working of the platoon.
Malaria remains a major global health issue, with over 229 million cases and 409,000 deaths reported annually, particularly in sub-Saharan Africa. Current diagnostic methods, such as microscopic examination of blood smears, are time-consuming and often lack accuracy due to human error and variability in slide quality. This study introduces Malaria-Net, a novel framework integrating advanced data preprocessing techniques with a Parasite Specific Attention Convolutional Neural Network (PSA-CNN) for enhanced feature extraction and Probabilistic Extremely Randomized Trees (PERT) for classification. The proposed approach begins with preprocessing steps, including image normalization, augmentation, and noise reduction to improve image quality and consistency. The PSA-CNN focuses on relevant features specific to malaria parasites, enhancing the network's ability to distinguish between different stages of infection. The PERT is then utilized for classification, leveraging its ability to handle high-dimensional data and provide probabilistic outputs. This method aims to improve diagnostic accuracy and reduce the reliance on manual interpretation, offering a more reliable and efficient solution for malaria detection. The proposed Malaria-Net achieves an accuracy of 99.937%, demonstrating its strong overall classification performance. It shows high precision (99.669%) and recall (99.337%), indicating that the model correctly identifies positive cases and minimizes false negatives. The F1-score of 99.539% reflects a balanced performance, combining precision and recall into a single metric, confirming its robustness in malaria detection.
Nowadays, scientific developments in the vehicle and infrastructure automation sectors are developing steadily to benefit end users in ways like reducing traffic congestion, enhancing safety, facilitating stress-free travel, reducing fuel costs, and enabling smart parking. Autonomous vehicles sometimes known as self-driving cars, can entirely replace conventional automobiles to accomplish safe and reliable intelligent transportation systems (ITS). Autonomous vehicles have improved sensing capabilities and can drive on roads that would be difficult for humans. Software, technology, and people, together with their interactions, are the building blocks of Autonomous vehicles. Even though Autonomous vehicles have been studied extensively, many issues have yet to be fixed. One of the biggest obstacles to developing Autonomous vehicles is improving their ability to communicate with other vehicles and pedestrians. Second, it's unclear whether society will benefit or suffer from people's tendency to place an excessive amount of faith in Autonomous vehicles and assume that they will never malfunction. Capabilities analysis is essential to making sure autonomous vehicles are safe and widely accepted. The proposed review outlines the key benefits, drawbacks, and difficulties of autonomous vehicle technology and provides a detailed analysis of the different autonomous vehicle mechanisms. Further, the developed review intends to formulate future prospects by examining the advantages and disadvantages of autonomous driving technology to encourage the development of innovative autonomous vehicle technologies to prevent human error and increase road safety and security.
