Illegal waste disposal has a negative impact on the environment and people’s quality of life. Drone imagery enables law enforcement authorities to efficiently assess the environmental impact during on-site inspections of suspicious landfill sites. Automated tools based on Deep Learning techniques can then quickly analyze aerial images to recognize several waste materials and classify their hazard level. However, large high-quality datasets are required for training and testing waste recognition models. Currently, no such datasets are publicly available, which limits the development of accurate waste identification algorithms. This paper presents DroneWaste, a dataset of aerial images extracted from orthomosaics that are reconstructed from drone-collected imagery. The dataset is a collection of 4993 images of 17 solid waste dumps that contain 20 different types of materials. The DroneWaste dataset is publicly accessible from the Zenodo repository repository. Technical validation proves that the dataset can be used for building object detection models able to recognize several types of waste in aerial imagery.

This dataset provides UAV-based RGB and multispectral imagery for crop monitoring, weed mapping, and field-level analysis in Camelina sativa cultivation. Data were collected from three agricultural fields in Thessaloniki and Chalkidiki, Greece, during summer 2025 and winter 2025-2026, capturing variability across locations, seasons, crop growth stages, UAV platforms, flight altitudes, spatial resolutions, illumination conditions, and sensing modalities. The dataset includes 3023 manually annotated RGB UAV images with human expert-generated polygon annotations of weed instances. The annotation scheme includes both coarse weed categories, such as broadleaf, narrowleaf, and generic weed classes, and fine-grained species-level labels, supporting classification, object detection, semantic and instance segmentation, hierarchical learning, and weed distribution analysis. In addition, the dataset provides RGB and multispectral UAV imagery, the raw RGB and multispectral images used for orthomosaic reconstruction, and both RGB and multispectral orthomosaic products in GeoTIFF format. The data were acquired using DJI Phantom 4 Pro and DJI Mavic 3 M UAV platforms at different flight altitudes, resulting in multiple ground sampling distances and image resolutions. This dataset is intended to support the development, benchmarking, and validation of computer vision and precision agriculture methods under realistic field conditions. To the best of our knowledge, it is among the first publicly available UAV datasets specifically focused on weed monitoring and field analysis in Camelina sativa crops.

Efficient hay bale detection and counting are essential tasks within modern precision agriculture, significantly impacting yield estimation, logistics, and sustainable resource management. To address current limitations in dataset quality and environmental representation, we introduce BaleUAVision, a comprehensive dataset consisting of 2,599 high-resolution RGB images, each containing numerous human-annotated hay bales. Captured by Unmanned Aerial Vehicles (UAVs) across 16 diverse agricultural fields in Northern Greece, the dataset includes varying flight altitudes (50–100 meters), diverse speeds (3.7–5 m/s), and overlapping strategies to ensure robust data representation. BaleUAVision provides rich annotations through polygon-based semantic segmentation in multiple formats (COCO, CSV, JSON, YOLO, segmentation masks) and high-quality orthomosaics for precise spatial analysis. Technical validation demonstrated the dataset’s effectiveness in training robust hay bale detection models using YOLOv11, achieving high precision and recall under varying geographic and altitude conditions. Specifically, the dataset supported effective generalization across geographically distinct areas (Xanthi and Drama regions) and varying altitudes, highlighting its utility in real-world UAV operations. The dataset and supplementary tools, scripts, and analyses are publicly available on Zenodo and GitHub respectively, following FAIR principles to support wide-reaching applicability within the research community.

This article introduces RobotIQ, a framework that empowers mobile robots with human-level planning capabilities, enabling seamless communication via natural language instructions through any Large Language Model. The proposed framework is designed in the ROS architecture and aims to bridge the gap between humans and robots, enabling robots to comprehend and execute user-expressed text or voice commands. Our research encompasses a wide spectrum of robotic tasks, ranging from fundamental logical, mathematical, and learning reasoning for transferring knowledge in domains like navigation, manipulation, and object localization, enabling the application of learned behaviors from simulated environments to real-world operations. All encapsulated within a modular crafted robot library suite of API-wise control functions, RobotIQ offers a fully functional AI-ROS-based toolset that allows researchers to design and develop their own robotic actions tailored to specific applications and robot configurations. The effectiveness of the proposed system was tested and validated both in simulated and real-world experiments focusing on a home service scenario that included an assistive application designed for elderly people. RobotIQ with an open-source, easy-to-use, and adaptable robotic library suite for any robot can be found at https://github.com/emmarapt/RobotIQ .

Unmanned Aerial Vehicles (UAVs) have revolutionized inspection tasks by offering a safer, more efficient, and flexible alternative to traditional methods. However, battery limitations often constrain their effectiveness, necessitating the development of optimized flight paths and data collection techniques. While existing approaches like coverage path planning (CPP) ensure comprehensive data collection, they can be inefficient, especially when inspecting multiple non-connected Regions of Interest (ROIs). This paper introduces the Fast Inspection of Scattered Regions (FISR) problem and proposes a novel solution, the multi-UAV Disjoint Areas Inspection (mUDAI) method. The introduced approach implements a two-fold optimization procedure, for calculating the best image capturing positions and the most efficient UAV trajectories, balancing data resolution and operational time, minimizing redundant data collection and resource consumption. The mUDAI method is designed to enable rapid, efficient inspections of scattered ROIs, making it ideal for applications such as security infrastructure assessments, agricultural inspections, and emergency site evaluations. A combination of simulated evaluations and real-world deployments is used to validate and quantify the method's ability to improve operational efficiency while preserving high-quality data capture, demonstrating its effectiveness in real-world operations. An open-source Python implementation of the mUDAI method can be found on GitHub1 and the collected and processed data from the real-world experiments are all hosted on Zenodo2. Finally, this on-line platform3 allows the interested readers to interact with the mUDAI method and generate their own multi-UAV FISR missions.

Multi-UAV Coverage Path Planning (mCPP) algorithms in popular commercial software typically treat a Region of Interest (RoI) only as a 2D plane, ignoring important 3D structure characteristics. This leads to incomplete 3D reconstructions, especially around occluded or vertical surfaces. In this paper, we propose a modular algorithm that can extend commercial two-dimensional path planners to facilitate terrain-aware planning by adjusting altitude and camera orientations. To demonstrate it, we extend the well-known DARP (Divide Areas for Optimal Multi-Robot Coverage Path Planning) algorithm and produce DARP-3D. We present simulation results in multiple 3D environments and a real-world flight test using DJI hardware. Compared to baseline, our approach consistently captures improved 3D reconstructions, particularly in areas with significant vertical features. An open-source implementation of the algorithm is available here: https://github.com/konskara/TerraPlan

In the original publication, the author has found few errors in the figures 2, 4, 11 and 12 which significantly impact both the paper's quality and the reputation of the journal.These issues undermine the clarity of the methodology and the novelty of the research.1.Each column represents a different case in the experiment, corresponding to the number of UAVs in both the obstaclefree and obstacle areas.However, in Fig. 2, the first row contains identical subfigures across all columns, while the subfigures in the remaining rows are identical to each other The incorrect Fig.

This dataset was collected to support Vehicle Routing Problem (VRP) optimization by providing structured time and distance matrices. A Third-Party Logistics (3PL) company granted access to its order management software, from which data on daily delivery problems involving pharmaceutical distribution were obtained. The dataset consists of carefully processed distance and time matrices, over a period of nine days. Each day’s problem involved 60-85 delivery stops that needed to be serviced. While the actual delivery routes covered only specific paths taken on the road, the generated matrices provide a complete view of travel distances and times between all locations, information essential for optimizing the routing process. To ensure confidentiality, only the structured matrices are provided, without the original address data. These matrices were generated using an API that computes travel durations based on historical traffic patterns, real-time data, and predictive models.From the API, we derived four distinct matrices: one for distances and three for travel times under different traffic scenarios: optimistic, pessimistic, and most likely. These matrices enable the modelling of realistic travel conditions accounting for the road congestion variability. Data retrieval was performed through automated API queries, ensuring consistency in structure and format. The collected matrices were processed and structured for direct use in VRP algorithms.The dataset offers substantial reuse potential by serving as a benchmark for evaluating VRP algorithms, enabling the comparison of optimization methods based on real-world logistics problems. It also supports statistical analysis and simulation, allowing researchers to assess travel time variability and model uncertainty in routing decisions through Monte Carlo simulations.Overall, this dataset offers valuable insights for optimizing delivery operations and addressing real-world logistics challenges. Its structured format, comprehensive traffic-based travel times, and applicability to VRP make it a valuable resource at the intersection of academia and industry.

Agentic AI refers to autonomous systems that can perceive their environment, make decisions, and take actions to achieve goals with minimal or no human intervention. Recent advances in Large Language Models (LLMs) have opened new pathways to imbue robots with such “agentic” behaviors by leveraging the LLMs’ vast knowledge and reasoning capabilities for planning and control. This survey provides the first comprehensive exploration of LLM-based robotic systems integration into agentic behaviors that have been validated in real-world applications. We systematically categorized these systems across navigation, manipulation, multi-agent, and general-purpose multi-task robots, reflecting the range of applications explored. We introduce a novel, first-of-its-kind agenticness classification that evaluates existing LLM-driven robotic works based on their degree of autonomy, goal-directed behavior, adaptability, and decision-making. Additionally, central to our contribution is an evaluation framework explicitly addressing ethical, safety, and transparency principles—including bias mitigation, fairness, robustness, safety guardrails, human oversight, explainability, auditability, and regulatory compliance. By jointly mapping the landscape of agentic capabilities and ethical safeguards, we uncover key gaps, tensions, and design trade-offs in current approaches. We believe that this work serves as both a diagnostic and a call to action: as LLM-empowered robots grow more capable, ensuring they remain comprehensible, controllable, and aligned with societal norms is not optional—it is essential.

Affordable and modular ground robots are key in robotics research and education. However, many existing platforms rely on expensive hardware and rigid architectures, limiting their reproducibility and practical deployment in resource-constrained environments. This work presents the design and implementation of a low-cost, 3D-printed omnidirectional robot that combines mechanical flexibility with modern embedded software capabilities. The proposed system features a lightweight fused deposition modeling-printed chassis, mecanum wheels for holonomic motion, and a Raspberry Pi 5 as its central processing unit. The system supports full ROS 2 compatibility, integrating all core functionalities, including sensor acquisition, control logic, and real-time teleoperation. Extensive engineering validation confirms that the platform operates reliably under manual control, providing a reproducible and extensible base for future robotic research in constrained environments.

The increasing number of electric vehicles (EVs) necessitates the installation of more charging stations. The challenge of managing these grid-connected charging stations leads to a multi-objective optimal control problem where station profitability, user preferences, grid requirements and stability should be optimized. However, it is challenging to determine the optimal charging/discharging EV schedule, since the controller should exploit fluctuations in the electricity prices, available renewable resources and available stored energy of other vehicles and cope with the uncertainty of EV arrival/departure scheduling. In addition, the growing number of connected vehicles results in a complex state and action vectors, making it difficult for centralized and single-agent controllers to handle the problem. In this paper, we propose a novel Multi-Agent and distributed Reinforcement Learning (MARL) framework that tackles the challenges mentioned above, producing controllers that achieve high performance levels under diverse conditions. In the proposed distributed framework, each charging spot makes its own charging/discharging decisions toward a cumulative cost reduction without sharing any type of private information, such as the arrival/departure time of a vehicle and its state of charge, addressing the problem of cost minimization and user satisfaction. The framework significantly improves the scalability and sample efficiency of the underlying Deep Deterministic Policy Gradient (DDPG) algorithm. Extensive numerical studies and simulations demonstrate the efficacy of the proposed approach compared with Rule-Based Controllers (RBCs) and well-established, state-of-the-art centralized RL (Reinforcement Learning) algorithms, offering performance improvements of up to 25% and 20% in reducing the energy cost and increasing user satisfaction, respectively.

This paper focuses on Coverage Path Planning (CPP) methodologies, particularly in the context of multi-robot missions, to efficiently cover user-defined Regions of Interest (ROIs) using groups of UAVs, while emphasizing on the reduction of energy consumption and mission duration. Optimizing the efficiency of multi-robot CPP missions involves addressing critical factors such as path length, the number of turns, re-visitations, and launch positions. Achieving these goals, particularly in complex and concave ROIs with No-Go Zones, is a challenging task. This work introduces a novel approach to address these challenges, emphasizing the selection of launch points for UAVs. By optimizing launch points, the mission’s energy and time efficiency are significantly enhanced, leading to more efficient coverage of the selected ROIs. To further support our research and foster further exploration on this topic, we provide the open-source implementation of our algorithm and our evaluation mechanisms.

In this paper we present a Blender add-on named LFG that allows for easy, large and realistic, 3D model LandFill Generation. Large datasets of vast, diverse synthetic landfills are hard to come by, and greatly in need for the purposes of developing and evaluating a multitude of algorithms (e.g. waste classification, 3D-reconstruction, volume estimation algorithms) in the context of research against environment crime. Additionally, they can be used alongside UAV simulators for the development of path-planning algorithms. Although there are some 3D models of landfills available on online 3D-model marketplaces, these are often expensive, low-quality, low-variety and unalterable models. LFG offers customizable, expandable options and realistic features tailored for landfill generation and research.

This paper deals with the problem of informative path planning for a UAV deployed for precision agriculture applications. First, we observe that the “fear of missing out” data lead to uniform, conservative scanning policies over the whole agricultural field. Consequently, employing a non-uniform scanning approach can mitigate the expenditure of time in areas with minimal or negligible real value, while ensuring heightened precision in information-dense regions. Turning to the available informative path planning methodologies, we discern that certain methods entail intensive computational requirements, while others necessitate training on an ideal world simulator. To address the aforementioned issues, we propose an active sensing coverage path planning approach, named OverFOMO, that regulates the speed of the UAV in accordance with both the relative quantity of the identified classes, i.e. crops and weeds, and the confidence level of such detections. To identify these instances, a robust Deep Learning segmentation model is deployed. The computational needs of the proposed algorithm are independent of the size of the agricultural field, rendering its applicability on modern UAVs quite straightforward. The proposed algorithm was evaluated with a simu-realistic pipeline, combining data from real UAV missions and the high-fidelity dynamics of AirSim simulator, showcasing its performance improvements over the established state of affairs for this type of missions. An open-source implementation of the algorithm and the evaluation pipeline is also available: https://github.com/emmarapt/OverFOMO.

While reinforcement learning (RL) algorithms have generated impressive strategies for a wide range of tasks, the performance improvements in continuous-domain, real-world problems do not follow the same trend. Poor exploration and quick convergence to locally optimal solutions play a dominant role. Advanced RL algorithms attempt to mitigate this issue by introducing exploration signals during the training procedure. This successful integration has paved the way to introduce signals from the intrinsic exploration branch. ACRE algorithm is a framework that concretely describes the conditions for such an integration, avoiding transforming the Markov decision process into time varying, and as a result, making the whole optimization scheme brittle and susceptible to instability. The key distinction of ACRE lies in the way of handling and storing both extrinsic and intrinsic rewards. ACRE is an off-policy, actor-critic style RL algorithm that separately approximates the forward novelty return. ACRE is shipped with a Gaussian mixture model to calculate the instantaneous novelty; however, different options could also be integrated. Using such an effective early exploration, ACRE results in substantial improvements over alternative RL methods, in a range of continuous control RL environments, such as learning from policy-misleading reward signals. Open-source implementation is available here: https://github.com/athakapo/ACRE .

This work focuses on the efficiency improvement of grid-based Coverage Path Planning (CPP) methodologies in real-world applications with UAVs. While several sophisticated approaches are met in literature, grid-based methods are not commonly used in real-life operations. This happens mostly due to the error that is introduced during the region’s representation on the grid, a step mandatory for such methods, that can have a great negative impact on their overall coverage efficiency. A previous work on UAVs’ coverage operations for remote sensing, has introduced a novel optimization procedure for finding the optimal relative placement between the region of interest and the grid, improving the coverage and resource utilization efficiency of the generated trajectories, but still, incorporating flaws that can affect certain aspects of the method’s effectiveness. This work goes one step forward and introduces a CPP method, that provides three different ad-hoc coverage modes: the Geo-fenced Coverage Mode, the Better Coverage Mode and the Complete Coverage Mode, each incorporating features suitable for specific types of vehicles and real-world applications. For the design of the coverage trajectories, user-defined percentages of overlap (sidelap and frontlap) are taken into consideration, so that the collected data will be appropriate for applications like orthomosaicing and 3D mapping. The newly introduced modes are evaluated through simulations, using 20 publicly available benchmark regions as testbed, demonstrating their stenghts and weaknesses in terms of coverage and efficiency. The proposed method with its ad-hoc modes can handle even the most complex-shaped, concave regions with obstacles, ensuring complete coverage, no-sharp-turns, non-overlapping trajectories and strict geo-fencing. The achieved results demonstrate that the common issues encountered in grid-based methods can be overcome by considering the appropriate parameters, so that such methods can provide robust solutions in the CPP domain.

This paper presents a modular and holistic Precision Agriculture platform, named CoFly, incorporating custom-developed AI and ICT technologies with pioneering functionalities in a UAV-agnostic system. Cognitional operations of micro Flying vehicles are utilized for data acquisition incorporating advanced coverage path planning and obstacle avoidance functionalities. Photogrammetric outcomes are extracted by processing UAV data into 2D fields and crop health maps, enabling the extraction of high-level semantic information about seed yields and quality. Based on vegetation health, CoFly incorporates a pixel-wise processing pipeline to detect and classify crop health deterioration sources. On top of that, a novel UAV mission planning scheme is employed to enable site-specific treatment by providing an automated solution for a targeted, on-the-spot, inspection. Upon the acquired inspection footage, a weed detection module is deployed, utilizing deep-learning methods, enabling weed classification. All of these capabilities are integrated inside a cost-effective and user-friendly end-to-end platform functioning on mobile devices. CoFly was tested and validated with extensive experimentation in agricultural fields with lucerne and wheat crops in Chalkidiki, Greece showcasing its performance.

The current paper proposes a hierarchical reinforcement learning (HRL) method to decompose a complex task into simpler sub-tasks and leverage those to improve the training of an autonomous agent in a simulated environment. For practical reasons (i.e., illustrating purposes, easy implementation, user-friendly interface, and useful functionalities), we employ two Python frameworks called TextWorld and MiniGrid. MiniGrid functions as a 2D simulated representation of the real environment, while TextWorld functions as a high-level abstraction of this simulated environment. Training on this abstraction disentangles manipulation from navigation actions and allows us to design a dense reward function instead of a sparse reward function for the lower-level environment, which, as we show, improves the performance of training. Formal methods are utilized throughout the paper to establish that our algorithm is not prevented from deriving solutions.

In this paper, we are trying to examine the role of robots in the Industry 4.0 Era. After a brief chronology and an updated literature review on the field, we are trying to examine one of the biggest problems in today's industrial automation, namely in Robotic bin-picking. Its objective is to manage to control a robot with multiple sensory motors attached and be able to collect identified objects with random poses out of a bin, containing a collection of those objects, using any kind of robot-end effector and place it in a predetermined location in the working area. As we observe, the Robotic bin-picking task is divided into three main subcategories: i) perception of the environment, ii) the tooling and iii) the processing architecture. These subcategories are observed thoroughly in the paper and the state-of-the-art techniques used are presented.