ISEP – CISTER – Artigos
Permanent URI for this collection
Browse
Recent Submissions
- Fusion flow-enhanced graph pooling residual networks for unmanned aerial vehicles surveillance in day and night dual visionsPublication . Noor, Alam; Li, Kai; Tovar, Eduardo; Zhang, Pei; Wei, BoRecognizing unauthorized Unmanned Aerial Vehicles (UAVs) within designated no-fly zones throughout the day and night is of paramount importance, where the unauthorized UAVs pose a substantial threat to both civil and military aviation safety. However, recognizing UAVs day and night with dual-vision cameras is nontrivial, since red-green-blue (RGB) images suffer from a low detection rate under an insufficient light condition, such as on cloudy or stormy days, while black-and-white infrared (IR) images struggle to capture UAVs that overlap with the background at night. In this paper, we propose a new optical flow-assisted graph-pooling residual network (OF-GPRN), which significantly enhances the UAV detection rate in day and night dual visions. The proposed OF-GPRN develops a new optical fusion to remove superfluous backgrounds, which improves RGB/IR imaging clarity. Furthermore, OF-GPRN extends optical fusion by incorporating a graph residual split attention network and a feature pyramid, which refines the perception of UAVs, leading to a higher success rate in UAV detection. A comprehensive performance evaluation is conducted using a benchmark UAV catch dataset. The results indicate that the proposed OF-GPRN elevates the UAV mean average precision (mAP) detection rate to 87.8%, marking a 17.9% advancement compared to the residual graph neural network (ResGCN)-based approach.
- Energy savings and emissions reduction of BEVs at an isolated complex intersectionPublication . Reddy, Radha; Almeida, Luis; Santos, Pedro Miguel; Kurunathan, Harrison; Tovar, EduardoImproving urban dwellers quality of life requires mitigating traffic congestion, minimizing waiting delays, and reducing fuel wastage and associated toxic air pollutants. Battery-electric vehicles (BEVs) are envisioned as the best option, thanks to zero exhaust emissions and regenerative braking. BEVs can be human-driven or autonomous and will co-exist with internal combustion engine vehicles (ICEVs) for years. BEVs can help at complex intersections where traffic is saturated. However, their benefits can be reduced by poor intersection management (IM) strategies that coordinate mixed traffic configurations inefficiently. This paper studies energy savings and emissions reduction using BEVs mixed with human-driven ICEVs under eight relevant IM approaches. It shows that adding BEVs has impacts on throughput, energy consumption, waiting delays, and tail-pipe emissions that depend on the specific IM approach used. Thus, this study provides the information needed to support an optimal choice of IM approaches considering the emerging trend towards electrical mobility.
- Reducing the gap between theory and practice in real-time systems with MARSPublication . Spilere Nandi, Giann; Pereira, David; Proença, José; Tovar, Eduardo; Nogueira, LuísA significant number of dependable systems rely on scheduling algorithms to achieve temporal correctness. Despite their relevance in real-world applications, only a narrow subset of the works in the literature of real-time systems are readily available to be reproduced in real-world hardware platforms. This lack of support not only hinders the reproducibility of research results, but also reduces the opportunity for new platform-specific research directions to emerge. In this work we discuss the use and development of an open-source tool named MARS capable of porting various scheduling tests and algorithms to hardware platforms used in distributed real-time dependable systems.
- Assessing short-range Shore-to-Shore (S2S) and Shore-to-Vessel (S2V) wifi communicationsPublication . D'Orey, Pedro; Gutiérrez Gaitán, Miguel; Santos, Pedro Miguel; Ribeiro, Manuel; Sousa, J. Borges de; Almeida, LuísWireless communications increasingly enable ubiquitous connectivity for a large number of nodes, applications and scenarios. One of the less explored scenarios are aquatic ecosystems, specially when enabled by near-shore and short-range communications. Overwater communications are impaired by a number of distinguishing dynamic factors, such as tides, waves or node mobility, that lead to a widely fluctuating and unpredictable channel. In this work, we empirically characterize near-shore, overwater channels at 2.4 GHz under realistic conditions, including tidal variations, and relatively short TX-RX separations. To this end, we conducted experiments in a coastal estuarine region and on a harbor to characterize Shore-to-Shore (S2S) and Shore-to-Vessel (S2V) communication channels, respectively, and to identify major factors impairing communication in such scenarios. The empirical results show that constructive/destructive interference patterns, varying reflecting surface, and node mobility (i.e. travel direction and particular maneuvers) have a relevant and noticeable impact on the received signal strength. Thus, a set of parameters should be simultaneously considered for improving the performance of communication systems supporting S2S and S2V links, namely tidal variations, reflection surface changes, antenna height, TX-RX alignment and TX-RX separation. The results useful provide insights into realistic S2S and S2V link design and operation.
- MARS: Safely instrumenting runtime monitors in real-time resource-constrained distributed systemsPublication . Nandi, Giann; Pereira, David; Proenca, José; Tovar, EduardoAdvancements in the energy efficiency and computational power of embedded devices allow developers to equip resource-constrained systems with a greater number of features and more complex behavior. As complexity of a system grows, so does the difficulty in demonstrating its overall correctness. Formal methods have been successfully applied in a variety of verification and validation scenarios, but their wide adoption in the industry and academia is still lackluster. Among the explanations listed in the literature for the low adoption of these techniques are the perceived difficulty of getting into formal practices and how formal tools are not usually aimed at practical use cases. Striving to address these issues, we present MARS, an open-source domain-specific language for the safe instrumentation of runtime verification monitors into real-time resource-constrained distributed systems. Our main objective with MARS is to ease the integration of runtime verification monitors in distributed applications while also providing developers with evidence of their correct instrumentation in the context of systems where dependability and temporal requirements need to be respected even under extreme resource constraints. We present the language syntax, the set of tools embedded into its compiler, its functionalities, and a use case to exemplify its use in a practical distributed application.
- Leverage variational graph representation for model poisoning on federated learningPublication . Li, Kai; Yuan, Xin; Zheng, Jingjing; Ni, Wei; Dressler, Falko; Jamalipour, AbbasThis article puts forth a new training data-untethered model poisoning (MP) attack on federated learning (FL). The new MP attack extends an adversarial variational graph autoencoder (VGAE) to create malicious local models based solely on the benign local models overheard without any access to the training data of FL. Such an advancement leads to the VGAE-MP attack that is not only efficacious but also remains elusive to detection. VGAE-MP attack extracts graph structural correlations among the benign local models and the training data features, adversarially regenerates the graph structure, and generates malicious local models using the adversarial graph structure and benign models’ features. Moreover, a new attacking algorithm is presented to train the malicious local models using VGAE and sub-gradient descent, while enabling an optimal selection of the benign local models for training the VGAE. Experiments demonstrate a gradual drop in FL accuracy under the proposed VGAE-MP attack and the ineffectiveness of existing defense mechanisms in detecting the attack, posing a severe threat to FL.
- On the path loss performance of underwater visible light communication schemes evaluated in several water environmentsPublication . Almonacid, Lucas; Játiva, Pablo Palacios; Meza, Cesar A. Azurdia; Dujovne, Diego; Soto, Ismael; Firoozabadi, Ali Dehghan; Gutiérrez Gaitán, MiguelThis paper presents an in-depth study into the necessity of efficient communication systems in underwater environments, with a primary focus on Underwater Visible Light Communication (UVLC). A novel path loss model that adapts to different water types is proposed to improve existing UVLC channel models. Validation against various scenarios, including different water types and receiver aperture diameters, is carried out using Monte Carlo simulations. The results demonstrate the efficiency and accuracy of the model by carefully fitting the actual performance of the UVLC systems. The results show a considerable improvement over previous models that only considered Lambert’s path loss and geometric path loss. Despite some variations observed at larger distances between the transmitter and receiver, the proposed model exhibits significant promise in the understanding and application of UVLC in different underwater environments. This study serves as a preliminary step toward developing more sophisticated and efficient models for UVLC systems.
- Two-ray model analysis for overwater communication at 28 GHz with different heightsPublication . Celades-Martínez, Jorge; Rodríguez, Mauricio; Gutiérrez Gaitán, Miguel; Almeida, LuisThis research aims to assess the signal propagation behavior of millimeter waves (mmWaves) over maritime environments. It focuses on the path loss performance of shore-to-vessel and vessel-to-vessel overwater communication at 28 GHz when considering line-of-sight conditions. The study is conducted by means of synthetic simulations at four different receiver antenna heights with respect to the water surface, representing emerging maritime Internet-of-Things application scenarios. Simulation results are shown concerning the path loss and the excess path loss – additional path loss relative to that in free-space – for each particular antenna height, over different TX-RX separations. We also show the cumulative distribution function of the excess path loss. The outcomes reveal variations of up to 10 dB in path loss performance depending on the height-distance setup. The results also reveal an initial distance range for all antenna heights in which the excess path loss is below 3 dB with 90% probability.
- Poisoning federated learning with graph neural networks in Internet of DronesPublication . Li, Kai; NOOR, ALAM; Ni, Wei; Tovar, Eduardo; Fu, Xiaoming; Akan, Ozgur B.Internet of Drones (IoD) is an innovative technology that integrates mobile computing capabilities with drones, enabling them to process data at or near the location where it is collected. Federated learning can significantly enhance the efficiency and effectiveness of data processing and decision-making in IoD. Since federated learning relies on aggregating updates from multiple drones, a malicious drone can generate poisoning local model updates that involves erroneous information, leading to incorrect decisions or even dangerous situations. In this paper, a new data-independent model poisoning attack is developed to manipulate federated learning accuracy, which does not rely on training data at drones. The proposed attack leverages an adversarial graph neural network (A-GNN) to generate poisoning local model updates based on the benign local models overheard. Particularly, the A-GNN discerns the graph structural correlations between the benign local models and the features of the training data that underpin these models. The graph structural correlations are reconstructively manipulated at the malicious drone to crafts poisoning local model updates, where the training loss of the federated learning is maximized.
- MobiWise: Eco-routing decision support leveraging the Internet of thingsPublication . Aguiar, Ana; Fernandes, Paulo; Guerreiro, Andreia; Tomás, Ricardo; Agnelo, João; Santos, José Luís; Araújo, Filipe; Coelho, Margarida C.; Fonseca, Carlos M.; D'Orey, Pedro; Luís, Manuel; Sargento, SusanaEco-routing distributes traffic in cities to improve mobility sustainability. The implementation of eco-routing in real-life requires a diverse set of information, including different kinds of sensors. These sensors are often already integrated in city infrastructure, some are technologically outdated, and are often operated by multiple entities. In this work, we provide a use case-oriented system design for an eco-routing service leveraging Internet-of-Things (IoT) technologies. The methodology involves six phases: 1) defining an eco-routing use case for a vehicle fleet; 2) formulating a routing problem as a multi-objective optimisation to divert traffic at a relevant hub facility; 3) identifying data sources and processing required information; 4) proposing a microservice-based architecture leveraging IoT technologies adequate to a multi-stakeholder scenario; 5) applying a microscopic traffic simulator as a digital twin to deal with data sparsity; and 6) visually illustrating eco-routing trade-offs to support decision making. We built a proof-of-concept for a mid-sized European city. Using real data and a calibrated digital twin, we would achieve hourly total emissions reductions up to 2.1%, when applied in a car fleet composed of 5% of eco-routing vehicles. This traffic diversion would allow annual carbon dioxide and nitrogen oxides savings of 400 tons and 1.2 tons, respectively.