This paper presents a hybrid optimization methodology integrating Biased Random-Key Genetic Algorithms (BRKGA), data mining, and reinforcement learning to address the Sensor Network Optimization Problem. Using statistical and clustering techniques, the method enhances BRKGA by incorporating patterns extracted from elite solutions. A reinforcement learning agent dynamically decides when and how to mine patterns, enhancing adaptivity and optimizing the process. The validated patterns are then integrated into a Quadratic Knapsack Problem (QKP) optimization model, guaranteeing compliance with budget constraints while efficiently exploring potential regions of the solution space. Experimental results demonstrate the approach's effectiveness in generating high-quality solutions for sensor allocation in Wireless Sensor Networks (WSNs), balancing exploration and exploitation, and achieving robust performance in complex optimization scenarios.

Neural rendering methods such as Neural Radiance Fields (NeRF) have achieved impressive novel-view synthesis results, but practical deployment demands a balance between accuracy and latency. This paper benchmarks two state-of-the-art neural rendering pipelines - Instant Neural Graphics Primitives (Instant-NGP) and 3D Gaussian Splatting - on a synthetic Blender scene (Lego) and a real LLFF scene (Fern). We evaluate rendering quality using PSNR, SSIM, and LPIPS, and measure rendering speed (milliseconds per frame and FPS) at 5122and 10242resolutions on an NVIDIA A100 GPU. Experiments cover single-view rendering and batched multi-view rendering. We further discuss the feasibility of deploying these pipelines in resource-constrained edge (TinyML) settings and for federated learning. Our results provide insights into the accuracy-latency trade-offs of hash-grid NeRF vs. point-based representations, guiding the choice of rendering pipeline for real-world applications.

The Automatic Identification System (AIS) has become essential for enhancing navigation safety and enabling innovative maritime applications. However, current research often relies on individual AIS data sources, limiting the comprehensiveness of maritime monitoring. This paper aims to develop a data fusion model that integrates AIS data from both fixed stations and satellite sources. Using a multi-sensor fusion approach, the model applies timestamp annotation, deduplication, and message decoding to AIS messages. Results demonstrate improved spatial and temporal coverage, supporting enriched situational awareness in maritime environments. This scalable model lays the foundation for advanced maritime monitoring and decision-making.

Optimizing vehicle routing is essential for logistics efficiency, but companies often operate independently due to competitive constraints, limiting potential gains from collaboration. In this work, we introduce a federated-inspired optimization approach for the Capacitated Vehicle Routing Problem (CVRP), where companies share optimized hyperparameters instead of raw data. Our method leverages Hybrid Genetic Search (HGS) for CVRP optimization while preserving data privacy. Unlike traditional federated learning, aggregation occurs outside the federated-inspired process, focusing on hyperparameter tuning rather than model updates. We evaluate different heuristic aggregation strategies, including FedAvg, Trimmed Mean, and Majority Vote. Experiments were conducted using CVRPLIB benchmark instances, varying the number of companies and problem instances to assess scalability and effectiveness. Our results indicate that federated-inspired hyperparameter aggregation can offer advantages over isolated optimization, particularly in reducing total routing costs. These findings suggest that controlled hyperparameter sharing may enhance logistics decisionmaking while maintaining operational autonomy.

Maritime Situational Awareness (MSA) is crucial for identifying suspicious vessel activities, such as dark-ship operations and prolonged loitering activities. However, the development of robust detection systems requires high-quality datasets that capture vessel encounters, particularly encounters that occur beyond 20 nautical miles (NM) from the coast. This paper presents the creation and analysis of a comprehensive data set that contains vessel trajectories associated with offshore encounters. The dataset, constructed using 12 months of data from the Marine Cadastre Automatic Identification System (AIS), leverages the H3 geohash system for spatial proximity detection and MovingPandas for trajectory extraction. The dataset analysis demonstrates that the dataset is a powerful tool for enhancing Maritime Domain Awareness (MDA), contributing to monitoring and security in the maritime environment. The analysis of encounter patterns highlights both the importance of reliable data and the need for a robust detection system to address uncertainties and information gaps.

Maritime vessel detection plays a critical role in navigation safety, surveillance, and environmental monitoring. While deep learning-based models, such as YOLOv8, offer high detection accuracy, they have the need for greater computational resources, making them less suited for embedded and real-time systems. This paper presents an efficient and lightweight vessel detection system by utilizing weightless neural networks for optimal object recognition and computational efficiency. Our approach was evaluated against YOLOv8 in terms of precision, recall, execution time, and energy consumption, using a publicly available ship dataset. The results demonstrate that our method achieves higher precision, comparable F1-score, and significantly lower computational overhead, reducing both inference time and power consumption. While YOLOv8 outperforms our approach in object localization via bounding box estimation, our model accurately localizes vessel positions with small error. Additionally, our framework reduces preprocessing and training time by up to$\mathbf{1 2}$times, making it highly effective for Edge AI and real-time maritime surveillance. By significantly lowering energy demands and computational latency, this research provides a scalable and sustainable alternative for vessel detection in autonomous maritime systems, smart surveillance networks, and embedded vision applications. Future work will explore hybrid models integrating weightless neural networks with deep learning to further enhance detection accuracy while maintaining energy efficiency.

In precision viticulture, deep learning and computer vision have been increasingly employed to automate grape cluster counting, a crucial task for yield estimation and farm management. However, existing methods fail to address the unique challenges posed by the pergola vine training system - a widely used but understudied method, particularly in Brazil's Northeast, where viticulture plays a key role in economic and social development. To address these limitations, we propose a novel, lightweight approach that combines object detection (YOLOv8) and tracking (ByteTrack) to count grape clusters in handheld smartphone videos captured under diverse environmental conditions. Unlike previous works, our technique is applied to the unique layout of the pergola system, where occlusions, perspective variations, and irregular cluster distributions complicate automated counting. We evaluate our method on real-world vineyard footage, achieving$75.1 \% \text{mAP} {@} 50$and a 69.9 % F1-score for cluster detection, with a final cluster count only 27 % above the actual value-establishing the first baseline for this challenging task. By enabling cost-effective, accessible yield estimation, our work contributes to improving efficiency and equity in viticulture. We publicly release our implementation to encourage further research and practical adoption (https://github.com/artsasse/pergola-grape-count/blob/main/grape_cluster_count.ipynb).

This study introduces a novel approach to the Sensor Allocation Problem (SAP) in Wireless Sensor Networks (WSNs) by integrating reinforcement learning with a multi-centrality heuristic. The goal of SAP is to optimize the network topology to minimize energy consumption while maintaining connectivity. We propose a reinforcement learning agent that interacts with a multi-centrality heuristic to dynamically select sensor placement. Extensive computational experiments were performed on both regular and semi-regular grid topologies of varying sizes. Comparisons with a Mixed-Integer Linear Programming (MILP) model reveal that our approach not only yields high-quality solutions in small to medium instances but also remains effective in larger networks, where the MILP solver often fails to produce feasible solutions. Notably, the RL module consistently improves upon the baseline heuristic allocation, demonstrating significant cost reductions while preserving full network coverage. By uniting a centrality-driven heuristic with an adaptive learning mechanism, the proposed hybrid framework addresses challenges in SAP more effectively than traditional exact methods. Consequently, it provides a promising avenue for large-scale WSN deployments, maintaining robust connectivity and efficiently managing energy consumption across various network configurations.

This study addresses the Wireless Sensor Network Planning Problem with Multiple Sources/Destinations (WSNPMSD), an optimization challenge focused on reducing the sensor count within a network topology for a specified area, considering numerous sources and destinations. We introduce a hybrid strategy for tackling WSNP-MSD, particularly effective for large-scale scenarios, combining a Biased Random-key Genetic Algorithm with a Local Branching Technique. This methodology is justified by the limitations exact methods may encounter when the number of variables increases. Through computational experiments, we demonstrate the superiority of our proposed method over conventional exact methods in managing large instances of the WSNP-MSD.

Detecting fishing trajectories in maritime surveillance is of the utmost importance for identifying illegal fishing activity. In the event of illegal fishing activity, the maritime authority can mobilize resources to engage the vessel; hence, a false flag can be costly. This study investigates the efficacy of ensemble learning techniques for boosting individual model performance and decreasing uncertainty. Employing a range of machine learning models, including logistic regression, decision trees, random forests, neural networks, gradient boosting, and recurrent neural networks, the research evaluates the combination of these using ensemble methods like ensemble mean, weighted ensemble, and stacking approaches to enhance precision and decrease uncertainty. The primary dataset comprises a combination of fishing vessel and cargo vessel trajectories to train and test the models. Methodologically, the paper details the process of data analysis and the application of ensemble learning. A comparative assessment of individual models versus ensemble techniques forms the crux of this study. Results indicate a marked improvement in accuracy and consistency when employing ensemble methods, with weighted and stacking ensembles showing particular promise. These findings suggest that ensemble models outperform their individual counterparts in the context of maritime surveillance. This research makes a notable contribution to the maritime surveillance domain, demonstrating the potential of ensemble learning in enhancing detection capabilities for illegal fishing activities. The implications of these advancements are critical for maritime authorities as they strive to effectively monitor and protect marine ecosystems.

The UFRJ Nautilus is a student-driven engineering project team at Federal University of Rio de Janeiro, focused on building and designing AUVs to compete in the AUSVI RoboSub Competition. There are several challenges on developing an AUV: location, computer vision, filters, collect and evaluate data from several sensors. The priority of the team was deliver a robot capable of localizing it self on a pool, with more reliability from all hardware and mechanical systems. We have developed a echo-localization algorithm based on the traditional beamforming that considers both time and frequency in order to have a faster and less power intensive procedure, Simulation showed that our algorithm achieved those objectives.