Reinforcement learning (RL) has seen an uptick in research interest in recent years, with many papers published in a plethora of different fields, topics and applications. A lot of that can be attributed to the recent advancements in machine learning (ML) and deep learning (DL) as a whole, the power of deep neural networks and the incorporation of them into reinforcement learning algorithms and techniques. This marriage of deep neural networks and traditional reinforcement learning methods has enabled reinforcement learning algorithms to achieve incredible results in use-cases that were previously deemed to be too complex. One of these highly complex scenarios is the realm of flight. In the last few years the number of papers published in this field has increased dramatically, but there is a fairly big divide between two major categories (quadcopters vs. fixed-wing aircraft). There is published work focusing on reinforcement learning using quadcopters, also known as drones, and there is published work focusing on fixed-wing aircraft, the more traditional aircraft, commonly referred to as airplanes. This paper focuses on the latter and will give an overview over the current state of reinforcement learning in the field of fixed-wing aviation. It will look into available toolkits usable for research as well as give an overview of different fields and tasks within reinforcement learning for fixed-wing aircraft control tasks. It contributes by providing detailed information about the methodologies used in each paper, while pointing out each paper’s contributions as well. A total of 48 papers on fixed-wing RL and airplane control are discussed. With the main focus lying on papers published post 2015, after the publication of Mnih et al.’s impactful deep RL paper.

Crops and their produce are vital to the livelihood of humans everywhere.World food security heavily relies on them, but still, even today, hundreds of millions of people world-wide are suffering from hunger. This is why it is essential to ensure that losses to the agricultural yield are kept at a minimum. Plant diseases, however, cause massive losses to the possible yield every year, rendering large amounts of the planted crops useless. And, if these diseases are not identified early enough, they will further infect more plants and therefore destroy even more yield. This is why plant diseases need to be recognized as fast as possible. Many diseases can be detected via the symptoms present on the plants leaves. As such, traditionally, staff samples and checks plants for their health in field manually. To speed up and increase accuracy, deep learning methods have been proposed to classify plant leaf images by their diseases. To train such models, however, one needs sufficiently large datasets. Such datasets are a gap in the field, with many datasets either being too small, not applicable (wrong plants or diseases), private, or taken under conditions that do not apply to the task at hand. One way to lower the need for more data and to overcome lesser data availability is transfer-learning, which utilizes unrelated rich, big, and available data to train feature extractors, which can then be re-trained to identify plant-leaves, based on the knowledge extracted from the prior task. In this work we will review and analyze a total of 84 CNN based transfer-learning papers from 2022 to 2025 out of 118 considered using PRISMA and discuss the insights gathered from them. This paper presents statistics on model, dataset, hyperparameter, pre-processing, augmentation, and metrics usage. Also, different methods for improving TL CNN models and training are listed. Additionally, performance comparisons of different prominent models for the field of plant leaf disease classification, as well as performance comparisons of different dataset types will be provided.

As denial of service attacks become more sophisticated, the source-side detection techniques are being studied to solve the limitations of target-side detection techniques such as delayed detection and difficulty in tracking attackers. Recently, some source-side detection techniques are being studied to use an adaptive attack detection threshold considering seasonal behavior of network traffic. However, because patterns of network traffic usage have become irregular with increased randomness and explosive traffic, the performance of the adaptive threshold technique has deteriorated. In addition, by limitations of the local view of a single site, distributed attacks from multiple sites may not be detected. In this paper, we propose a LSTM (Long Short Term Memory) based collaborative source-side DDoS (Distributed Denial of Service) attack detection framework which provides the attack detection result of a collaboration network in a global view. The proposed framework applies LSTM-based adaptive thresholds to each source-side network to mitigate performance degradation caused by irregular network traffic behavior. Also, in order to overcome the limitation of performance caused by the local view of single source-side network, the proposed framework constructs a collaborative network through multiple detection sites and aggregates feedback from each site, such as detection rates, local traffic patterns, and timestamp. The collaborative attack detection technique uses the aggregated feedback to determine whether the attack is finally detected and shares the finial detection results with multiple sites. Depending on this final detection result, the adaptive thresholds of each site are reset. Through extensive evaluation of actual network traffic data, the proposed collaborative source-side attack detection technique shows around 15% lower false positive rate than the single source-side attack detection technique while maintaining a high detection rate.

The exponential growth of video data across domains such as surveillance, transportation, and healthcare has raised critical challenges in scalability, real-time processing, and privacy preservation. While existing studies have addressed individual aspects of Video Big Data Analytics (VBDA), an integrated, up-to-date perspective remains limited. This paper presents a comprehensive survey of system architectures and enabling technologies in VBDA. It categorizes system architectures into four primary types as follows: centralized, cloud-based infrastructures, edge computing, and hybrid cloud–edge. It also analyzes key enabling technologies, including real-time streaming, scalable distributed processing, intelligent AI models, and advanced storage for managing large-scale multimodal video data. In addition, the study provides a functional taxonomy of core video processing tasks, including object detection, anomaly recognition, and semantic retrieval, and maps these tasks to real-world applications. Based on the survey findings, the paper proposes ViMindXAI, a hybrid AI-driven platform that combines edge and cloud orchestration, adaptive storage, and privacy-aware learning to support scalable and trustworthy video analytics. Our analysis in this survey highlights emerging trends such as the shift toward hybrid cloud–edge architectures, the growing importance of explainable AI and federated learning, and the urgent need for secure and efficient video data management. These findings highlight key directions for designing next-generation VBDA platforms that enhance real-time, data-driven decision-making in domains such as public safety, transportation, and healthcare. These platforms facilitate timely insights, rapid response, and regulatory alignment through scalable and explainable analytics. This work provides a robust conceptual foundation for future research on adaptive and efficient decision-support systems in video-intensive environments.

This paper addresses reliability of data dissemination applications when there are severe disruptions to the underlying physical infrastructure. Such massive simultaneous physical failures can happen during the geographical events such as natural disasters (earthquakes, floods, tornados) or sudden power outages - infrastructure failures in these cases are geographically correlated. In particular, we focus on overlay based data dissemination mechanisms and explore their ability to tolerate such large geographically correlated failures. Due to the tight correlation between multiple overlay links and a single physical link, a few physical failures may affect lots of overlay links. To enable reliable dissemination under such conditions, we propose overlay network construction methods that incorporate proximity-aware neighbor selection methods to improve the performance of the overlay data dissemination, to the extent possible, in terms of reliability and latency. In this approach, the overlay nodes select neighbors which are most likely distinct in presence of a geographical failure; we show how an overlay structure constructed using our proximity-aware neighbor selection techniques can disseminate data to over 80% of reachable end clients without any significant additional latency under various geographical failure conditions.

In large buildings, during the situation of fire or other hazards, a smart evacuation guidance system needs to fully consider manifold aspects of hazards to guide evacuees through exit gates as fast as possible by dynamic and safe routes. In this paper, we propose a smart evacuation guidance system with a dynamic evacuation routing approach by using the LCDT (Length-Capacity-Density-Trustiness) weighted graph model and partial view (PV) information which represents the hazard intensity and the crowd congestion information of a group of sections/floors in the building. The proposed system is designed as a distributed system with multiple layers of computing by using smart indicators. Given such a system, we develop an efficient distributed approach, so-called LCDT&PV, to find out effective evacuation routes dynamically. We then propose an estimating congestion strategy in order to improve the efficiency of dynamic evacuation routes. To validate the proposed system, we implement a simulator to compare the proposed evacuation routing approach with baseline approaches. Experimental results show that the proposed approach reduces up to 30% of the total evacuation time compared with others. Furthermore, through the results of initial smart indicator implementation, which can interact with the simulator, we show the viability of the proposed system.