Carlos Viegas

Swarm robotics (SR) and swarm intelligence (SI) share decentralization and collective behavior but address different problems. SR concerns physical multi-robot systems whose collective behavior must hold under sensing, communication, energy, and safety constraints, while SI includes computational models and optimization methods, many of which involve no robots. Conflating them obscures which SI results transfer to robot swarms and what machine learning (ML) contributes. This review examines SR coordination, SI optimization, and ML-augmented hybrid methods from a deployment perspective. It proposes a tiered framework for deciding whether a system is a robotic swarm, a taxonomy of five hybrid SR–SI–ML architecture types, and feasibility conditions for combining methods on a given platform. Across 17 coded studies, SI optimizers such as Ant Colony Optimization and Particle Swarm Optimization are evaluated mainly on benchmarks or in simulation and reach hardware only with supporting infrastructure; hardware experiments, with one exception, use twenty robots or fewer, and learned controllers transfer to hardware in some studies but show a large simulation-to-reality gap in others. Layered hybrid architectures, in which SR provides coordination, SI provides optimization and ML provides adaptation, remain a design hypothesis. The review proposes a reporting checklist for future studies.

Swarm robotics enables groups of autonomous agents to coordinate and perform tasks beyond the capabilities of individual robots. This state-of-the-art review focuses on the defining features, principles, and algorithms of swarm robotic systems, with an emphasis on recent advances. It examined classical and modern bio-inspired coordination strategies, decentralized control algorithms and hybrid approaches, highlighting their strengths, limitations and applicability to real-world deployments. Key challenges, such as scalability, robustness, adaptability and the gap between simulation and hardware implementation are analyzed.

Localization and mapping remain critical challenges for Unmanned Ground Vehicles (UGVs) operating in unstructured natural environments, such as forests and agricultural fields. While Visual SLAM (VSLAM) and Visual–Inertial SLAM (VI-SLAM) have matured significantly in structured and urban scenarios, their extension to outdoor natural domains introduces severe challenges, including dynamic vegetation, illumination variations, a lack of distinctive features, and degraded GNSS availability. Recent advances in Deep Learning have brought promising developments to VSLAM- and VI-SLAM-based pipelines, ranging from learned feature extraction and matching to self-supervised monocular depth prediction and differentiable end-to-end SLAM frameworks. Furthermore, emerging methods for adaptive sensor fusion, leveraging attention mechanisms and reinforcement learning, open new opportunities to improve robustness by dynamically weighting the contributions of camera and IMU measurements. This review provides a comprehensive overview of Visual and Visual–Inertial SLAM for UGVs in unstructured environments, highlighting the challenges posed by natural contexts and the limitations of current pipelines. Classic VI-SLAM frameworks and recent Deep-Learning-based approaches were systematically reviewed. Special attention is given to field robotics applications in agriculture and forestry, where low-cost sensors and robustness against environmental variability are essential. Finally, open research directions are discussed, including self-supervised representation learning, adaptive sensor confidence models, and scalable low-cost alternatives. By identifying key gaps and opportunities, this work aims to guide future research toward resilient, adaptive, and economically viable VSLAM and VI-SLAM pipelines, tailored for UGV navigation in unstructured natural environments.

We present an algorithm for wildfire suppression based on graph theory that builds on realistic simulations of the propagation, is efficient, produces results in a short time, and has the potential to help achieve a faster suppression of a real fire. The graph-based suppression algorithm, which looks for the selection of the best locations to attack the suppression of wildfires, is formally correct, and its complexity is computed. The performance of the algorithm is compared to random and brute-force approaches. The simulations demonstrate that the graph-based algorithm outperforms the other two, being one order of magnitude faster than the brute-force approach. Furthermore, based on extensive computer simulations using real-life wildfire data, we demonstrate the effectiveness of the proposed algorithm. We also present a set of real-life wildfire data for wildfires that occurred in Portugal. The algorithm has potential for operational applications, supporting fire-suppression strategies and aiding civil protection agencies, insurance companies, and forest management services. It is expected to assist firefighters in determining optimal fire propagation strategies.

Trees are vital to both environmental health and human well-being. They purify the air we breathe, support biodiversity by providing habitats for wildlife, prevent soil erosion to maintain fertile land, and supply wood for construction, fuel, and a multitude of essential products such as fruits, to name a few. Therefore, it is important to monitor and preserve them to protect the natural environment for future generations and ensure the sustainability of our planet. Remote sensing is the rapidly advancing and powerful tool that enables us to monitor and manage trees and forests efficiently and at large scale. Statistical methods, machine learning, and more recently deep learning are essential for analyzing the vast amounts of data collected, making data the fundamental component of these methodologies. The advancement of these methods goes hand in hand with the availability of sample data; therefore, a review study on available high-resolution aerial datasets of trees can help pave the way for further development of analytical methods in this field. This study aims to shed light on publicly available datasets by conducting a systematic search and filter and an in-depth analysis of them, including their alignment with the FAIR—findable, accessible, interoperable, and reusable—principles and the latest trends concerning applications for such datasets.

Ground segmentation in LiDAR point clouds is a foundational capability for autonomous systems, enabling safe navigation in applications ranging from urban self-driving vehicles to planetary exploration rovers. Reliably distinguishing traversable surfaces in geometrically irregular or sensor-sparse environments remains a critical challenge. This paper introduces a hybrid framework that synergizes multi-resolution polar discretization with sparse convolutional neural networks (SCNNs) to address these challenges. The method hierarchically partitions point clouds into adaptive sectors, leveraging PCA-derived geometric features and dynamic variance thresholds for robust terrain modeling, while a SCNN resolves ambiguities in data-sparse regions. Evaluated in structured (SemanticKITTI) and unstructured (Rellis-3D) environments, two different versions of the proposed method are studied, including a purely geometric method and a hybrid approach that exploits deep learning techniques. A comparison of the proposed method with its purely geometric version is made for the purpose of highlighting the strengths of each approach. The hybrid approach achieves state-of-the-art performance, attaining an F1-score of 95.4% in urban environments, surpassing the purely geometric (91.4%) and learning-based baselines. Conversely, in unstructured terrains, the geometric variant demonstrates superior metric balance (80.8% F1) compared to the hybrid method (75.8% F1), highlighting context-dependent trade-offs between precision and recall. The framework’s generalization is further validated on custom datasets (UMH-Gardens, Coimbra-Liv), showcasing robustness to sensor variations and environmental complexity. The code and datasets are openly available to facilitate reproducibility.

The autonomous navigation of unmanned ground vehicles (UGVs) in unstructured environments, such as agricultural or forestry settings, has been the subject of extensive research by various investigators. The navigation capability of a UGV in unstructured environments requires considering numerous factors, including the quality of data reception that allows reliable interpretation of what the UGV perceives in a given environment, as well as the use these data to control the UGV’s navigation. This article aims to study different PID control algorithms to enable autonomous navigation on a robotic platform. The robotic platform consists of a forestry tractor, used for forest cleaning tasks, which was converted into a UGV through the integration of sensors. Using sensor data, the UGV’s position and orientation are obtained and utilized for navigation by inputting these data into a PID control algorithm. The correct choice of PID control algorithm involved the study, analysis, and implementation of different controllers, leading to the conclusion that the Vector Field control algorithm demonstrated better performance compared to the others studied and implemented in this paper.