Decision-making in real-world settings rarely follows a fixed script. Instead, it unfolds as a dynamic reasoning process in which the appropriate course of action evolves as new context and data become available. Traditional Business Process Management systems provide rigor, determinism, and auditability, yet they generally struggle to adapt their execution at runtime. Conversely, agentic systems based on Large Language Models (LLMs) bring flexibility to decision-making, but they are inherently opaque, often unreliable, and suffer from significant scalability constraints when operating over large datasets. To combine these complementary paradigms, we introduce VADAOrchestra, a neurosymbolic framework that models complex workflows as evolving reasoning processes. The framework adopts a hybrid approach: given a user query and a collection of data sources, an LLM-based orchestrator incrementally plans and adapts the workflow. This is encoded as a logic program in a fragment of Datalog+/- where predicates correspond to tool invocations and rules represent both predefined domain dependencies and logic constructs synthesized on demand to manipulate intermediate results. All logical inference tasks are then executed by a state-of-the-art Datalog+/- symbolic engine. This approach provides a verifiable reasoning trace, supporting the auditability and reproducibility of the entire process. Furthermore, by decoupling high-level orchestration from symbolic inference, it addresses scalability concerns, enabling complex reasoning over large datasets through targeted data querying. We evaluate VADAOrchestra on real-world financial use cases, demonstrating faithfulness, scalability, and explainability compared to standard agentic architectures.

During advanced surveillance missions, Unmanned Aerial Vehicles (UAVs) usually require the execution of edge-assisted computer vision (CV) tasks. In multi-hop UAV networks, the successful transmission of these tasks to the edge is severely challenged due to severe bandwidth constraints, and the possible node failures. To address these critical challenges, we propose a novel A 2 - UAV framework that optimizes the number of correctly executed tasks at the edge. In stark contrast with existing art, we take an application-aware approach and formulate a novel Application-Aware Task Planning Problem ( A 2 - TPP ) to optimize routing, data pre-processing and target assignment for each UAV. Our formulation explicitly takes into account (i) the relationship between CV task accuracy and image compression for the classes of interest based on the available dataset, (ii) the target positions , (iii) the current energy/position of the UAVs, and (iv) the possible node failures. We demonstrate A 2 - TPP is NP-Hard and propose a polynomial-time algorithm to solve it efficiently. We extensively evaluate A 2 - UAV through simulation and real-world experiments using a testbed composed by four DJI Mavic Air 2 UAVs. Results on image classification show that A 2 - UAV attains on average around 38% more accomplished tasks w.r.t. the state of the art, with a 400% improvement in tasks-intensive scenarios. Moreover, we show that our framework is able to reconfigure the network in case of nodes failure.

Flying ad-hoc networks (FANETs) are a powerful tool for inspecting safety-critical scenarios, including post-disaster areas or military fields, where they ensure prompt area monitoring and fast detection of events of interest. However, wide area deployment of FANETs requires fast and reliable communications among devices and their base station to ensure prompt intervention upon detection of anomalies. Existing long-range communication technologies are inadequate to meet the data rate requirements and delay constraints of safety-critical applications. Previous solutions to enable ad-hoc communications in mobile networks also fall short of exploiting the controllable mobility of FANETs. To face this challenge, we formulate the connected deployment problem, where we require the FANET to dynamically create connected coverage formations to ensure multi-hop low-latency communications while performing the monitoring task. We show that addressing the above problem under the joint requirement of maximizing event coverage is NP-hard. We propose a joint Task Management and Routing Algorithm called TaMaRA, a polynomial-time solution based on a two-phase approximation of the problem. By means of extensive simulations and real field experiments we show that our approach outperforms existing solutions in terms of monitoring accuracy and system responsiveness.

Learning agent behaviors from observational data has shown to improve our understanding of their decision-making processes, advancing our ability to explain their interactions with the environment and other agents. While multiple learning techniques have been proposed in the literature, there is one particular setting that has not been explored yet: multi agent systems where agent identities remain anonymous. For instance, in financial markets labeled data that identifies market participant strategies is typically proprietary, and only the anonymous state-action pairs that result from the interaction of multiple market participants are publicly available. As a result, sequences of agent actions are not observable, restricting the applicability of existing work. In this paper, we propose a Policy Clustering algorithm, called K-SHAP, that learns to group anonymous state-action pairs according to the agent policies. We frame the problem as an Imitation Learning (IL) task, and we learn a world-policy able to mimic all the agent behaviors upon different environmental states. We leverage the world-policy to explain each anonymous observation through an additive feature attribution method called SHAP (SHapley Additive exPlanations). Finally, by clustering the explanations we show that we are able to identify different agent policies and group observations accordingly. We evaluate our approach on simulated synthetic market data and a real-world financial dataset. We show that our proposal significantly and consistently outperforms the existing methods, identifying different agent strategies.

To perform advanced surveillance, Unmanned Aerial Vehicles (UAVs) require the execution of edge-assisted computer vision (CV) tasks. In multi-hop UAV networks, the successful transmission of these tasks to the edge is severely challenged due to severe bandwidth constraints. For this reason, we propose a novel A2-UAV framework to optimize the number of correctly executed tasks at the edge. In stark contrast with existing art, we take an application-aware approach and formulate a novel Application-Aware Task Planning Problem (A2-TPP) that takes into account (i) the relationship between deep neural network (DNN) accuracy and image compression for the classes of interest based on the available dataset, (ii) the target positions, (iii) the current energy/position of the UAVs to optimize routing, data pre-processing and target assignment for each UAV. We demonstrate A2-TPP is NP-Hard and propose a polynomial-time algorithm to solve it efficiently. We extensively evaluate A2-UAV through real-world experiments with a testbed composed by four DJI Mavic Air 2 UAVs. We consider state-of-the-art image classification tasks with four different DNN models (i.e., DenseNet, ResNet152, ResNet50 and MobileNet-V2) and object detection tasks using YoloV4 trained on the ImageNet dataset. Results show that A2-UAV attains on average around 38% more accomplished tasks than the state of the art, with 400% more accomplished tasks when the number of targets increase significantly. To allow full reproducibility, we pledge to share datasets and code with the research community.

Synthetic time series are often used in practical applications to augment the historical time series dataset, amplify the occurrence of rare events and also create counterfactual scenarios. Distributional-similarity (which we refer to as realism) as well as the satisfaction of certain numerical constraints are common requirements for counterfactual time series generation. For instance, the US Federal Reserve publishes synthetic market stress scenarios given by the constrained time series for financial institutions to assess their performance in hypothetical recessions. Existing approaches for generating constrained time series usually penalize training loss to enforce constraints, and reject non-conforming samples. However, these approaches would require re-training if we change constraints, and rejection sampling can be computationally expensive, or impractical for complex constraints. In this paper, we propose a novel set of methods to tackle the constrained time series generation problem and provide efficient sampling while ensuring the realism of generated time series. In particular, we frame the problem using a constrained optimization framework and then we propose a set of generative methods including 'GuidedDiffTime', a guided diffusion model. We empirically evaluate our work on several datasets for financial and energy data, where incorporating constraints is critical. We show that our approaches outperform existing work both qualitatively and quantitatively, and that 'GuidedDiffTime' does not require re-training for new constraints, resulting in a significant carbon footprint reduction, up to 92% w.r.t. existing deep learning methods.

Swarms of Unmanned Aerial Vehicles (UAVs) are a key technology to support communication in many harsh environments where fixed infrastructures (e.g., 5G) are disrupted or not available. However, the fast mobility and highly dynamic network topology pose unique challenges and require the development of novel multi-hop routing protocols. Previous work in this direction extends geographical protocols or adapts approaches designed for Mobile Ad-hoc NETworks (MANETs), rarely taking full advantage of UAV capabilities. In this paper, we introduce a novel data offloading approach, namely Stop & Offload, that exploits the device controllable mobility to facilitate network routing. The swarm of UAVs performs data offloading synchronously and recurrently. At fixed intervals of time, the swarm interrupts the sensing mission (Stop) and moves, as little as possible, to build a connected formation to the base station and offload the data (Offload). We provide both centralized solutions — assuming a long-range control channel — and a distributed solution — working in the absence of a control channel. By means of extensive simulations we show that our proposals outperform state-of-the-art solutions, decreasing the time taken to build a connected formation of about 45% and increasing the time spent on sensing of 10%. Additionally, we compared our protocol with various routing strategies and observe remarkable improvements, including a 50% reduction in average packet delay.

Aerial drones are increasingly used to perform monitoring tasks in a large number of applications. Current solutions to trajectory planning rely on perfect knowledge of ongoing events requiring inspection. Nevertheless, in many scenarios the events’ time and position can only be estimated with someuncertainty. Unlike previous work, we consider critical scenarios where a squad of drones is required to autonomously inspect an area of interest underuncertaintyof time and location of target events. The main goal of the squad is to ensure maximum coverage of event monitoring with minimum average inspection delay. With no initial knowledge, the drones share their local observations of the environment and apply the Parzen-Rosenblatt approach to manage a dynamic probabilistic map of ongoing events. This map is integrated into a virtual force approach for a joint solution to distributed dynamic trajectory planning and collision avoidance. Through extensive simulations and real-field experiments, we compare our proposal againstAC-GAP, a state-of-art solution for UAVs, andSweep, a sweep-based algorithm for multiple robots. We show that our proposal discovers new events 30-40$\%$faster than the other algorithms, and outperforms them in terms of percentage of visited events and inspection delay, under a wide variety of scenarios.

In most of the developing countries, the economy is largely based on agriculture. The poor availability of skilled personnel and of appropriate supporting infrastructure, make crop fields vulnerable to the outbreak of plant diseases, possibly due to spreading viruses and fungi, or to adverse environmental conditions, such as drought. The mobile application PlantVillage Nuru provides an invaluable tool for early detection of plant diseases and sustainable food production. A mobile device endowed with Nuru is a powerful mobile sensor: it analyzes plant images and uses an AI engine to recognize health issues. In this article, we propose a crowdsensing framework, where Nuru is adopted at large scale in the farmer population. We tackle the device deployment problem, where device mobility is only partially controllable, mostly in an indirect manner, through incentives. We propose two problem formulations, and related algorithms, to minimize the number of required smartphones while providing sufficient geographical coverage. We study the proposed models in simulated as well as real scenarios, showing that they outperform the current solutions in terms of monitoring accuracy and completeness, with lower cost. Then, we describe the testbed implementation, confirming the applicability of the proposed crowdsensing framework in a real scenario in Kenya.

The fast and unconstrained mobility of Flying Ad-hoc NETworks (FANETs) brings about the need to develop solutions for packet routing in a highly dynamic topology scenario. Previous works in this direction aim at extending protocols designed for Mobile Ad-hoc NETworks (MANETs) to the more challenging domain of FANETs. Unlike previous approaches, we aim at exploiting the device controllable mobility to facilitate network routing. We propose MAD (Movement Assisted Delivery): a packet routing protocol specifically tailored for networks of aerial vehicles. MAD enables adaptive selection of the most suitable relay nodes for packet delivery, resorting to movement-assisted delivery upon need, which is supported by a reinforcement learning approach. By means of extensive simulations we show that MAD outperforms previous solutions in all the considered performance metrics including average packet delay, delivery ratio, and communication overhead, at the expense of a moderate loss in average device availability.

Many safety critical scenarios, including post-disaster areas, or military fields, require prompt area monitoring and fast detection of events of interest. Flying Ad-hoc Networks (FANETs) provide a powerful tool to search the area, and locate anomalies. Nevertheless, wide-area deployment of FANETs poses a number of challenges. Existing long range communication technologies are inadequate to meet the data rate and delay requirements of a safety critical application. To face this challenge, we formulate the connected deployment problem, where we require the FANET to create connected formations to ensure multi-hop low-latency communications while performing the monitoring task. We show that addressing the above problem with the aim of maximizing event coverage is NP-hard. We propose a polynomial time solution, called Greedy Connected Deployment (GCD), based on a two phase approximation of the problem. By means of extensive simulations and real field experiments, we show that our approach outperforms existing solutions to related problems, both in terms of monitoring accuracy and system responsiveness.

Recent years have witnessed the design and development of several smart devices that are wireless and battery-less. These devices exploit RFID backscattering-based computation and transmissions. Although singular devices can operate efficiently, their coexistence needs to be controlled, as they have widely varying communication requirements, depending on their interaction with the environment. The design of efficient communication protocols able to dynamically adapt to current device operation is quite a new problem that the existing work cannot solve well. In this article, we propose a new communication protocol, called ReLEDF, that dynamically discovers devices in smart buildings and their active and nonactive status and when active their current communication behavior (through a learning-based mechanism) and schedules transmission slots (through an Earliest Deadline First-- (EDF) based mechanism) adapt to different data transmission requirements. Combining learning and scheduling introduces a tag starvation problem, so we also propose a new mode-change scheduling approach. Extensive simulations clearly show the benefits of using ReLEDF, which successfully delivers over 95% of new data samples in a typical smart home scenario with up to 150 heterogeneous smart devices, outperforming related solutions. Real experiments are also conducted to demonstrate the applicability of ReLEDF and to validate the simulations.

Fleets of cooperative drones are a powerful tool in monitoring critical scenarios requiring early anomaly discovery and intervention. Due to limited energy availability and application requirements, drones may visit target points in consecutive trips, with recharging and data offloading in between. To capture timeliness of intervention and prioritize early coverage, we propose the new notion of Weighted Progressive Coverage, which is based on the definition of time dependent weights. Weighted progressive coverage generalizes classic notions of coverage, as well as a new notion of accumulative coverage specifically designed to address trip scheduling. We show that weighted progressive coverage maximization is NP-hard and propose an efficient polynomial algorithm, called Greedy and Prune (GaP), with guaranteed approximation. By means of simulations we show that GaP performs close to the optimal solution and outperforms a previous approach in all the considered performance metrics, including coverage, average inspection delay, energy consumption, and computation time, in a wide range of application scenarios. Through prototype experiments we also confirm the theoretical and simulation analysis, and demonstrate the applicability of our algorithm in real scenarios.

Has now become more important than ever to guarantee an always present connectivity to users, especially in emergency scenarios. However, in case of a disaster, network infrastructures are often damaged, with consequent connectivity disruption, isolating users when are more in need for information and help. Drones may supply with a recovery network, thanks to their capabilities to provide network connectivity on the fly. However, users typically need special devices or applications to reach these networks, reducing their applicability and adoption.

We consider the problem of assigning tasks and related trajectories to a fleet of drones, in critical scenarios requiring early anomaly discovery and intervention. Drones visit target points in consecutive trips, with recharging and data offloading in between. We propose a novel metric, called weighted coverage, which generalizes classic notions of coverage, as well as a new notion of accumulative coverage which prioritizes early inspection of target points. We formulate an ILP problem for weighted coverage maximization and show its NP-hardness. We propose an efficient polynomial algorithm with guaranteed approximation. By means of simulations we show that our algorithm performs close to the optimal solution and outperforms a previous approach in terms of several performance metrics, including coverage, average inspection delay, energy consumption, and computation time, under a wide range of application scenarios.