LEE- Savunma Teknolojileri-Yüksek Lisans
Bu koleksiyon için kalıcı URI
Gözat
Son Başvurular
1 - 5 / 31
-
ÖgeMissile evasion maneuver generation with model-free deep reinforcement learning(Graduate School, 2024-07-25)Unmanned Combat Aerial Vehicles (UCAVs) have fundamentally transformed modern military tactics and strategies, offering capabilities that were previously impossible or highly risky for manned aircraft. These vehicles can conduct a variety of critical missions, including intelligence gathering, reconnaissance operations, surveillance, reconnaissance, and target acquisition, which are essential for gaining and maintaining situational awareness in both peacetime and wartime scenarios. Additionally, UCAVs are equipped with sophisticated weaponry such as aircraft ordnance, missiles, bombs, and Anti-Tank Guided Missiles (ATGMs), making them versatile and effective in a wide range of combat situations. Missiles play a crucial role in modern warfare as long-range weapon systems. With their high speed and precision guidance capabilities, missiles can effectively strike land, sea, and air targets. The technological advancements in missile systems have transformed the nature of warfare, allowing for strategic targeting of enemy assets with great accuracy. However, the power of these weapons has also necessitated the development of advanced defense systems. Specifically, air defense systems have been designed with complex technologies to detect, track, and intercept enemy missiles, ensuring protection against these potent threats. Unmanned Combat Aerial Vehicles (UCAVs) equipped with artificial intelligence-based evasion systems offer significant advantages in terms of survivability and operational effectiveness on the modern battlefield. These systems can autonomously detect incoming missile threats and optimize the aircraft's control surfaces to perform evasive maneuvers effectively. The integration of AI technology allows UCAVs to respond quickly and efficiently to high-speed missile threats, executing precise avoidance actions. This AI-driven capability enhances the flexibility, resilience, and strategic value of UCAVs in combat situations, while also improving safety in unmanned operations. The strategic value of UCAVs lies not only in their operational capabilities but also in the distinct advantages they offer over traditional manned aircraft. One of the most significant benefits is the reduced risk to human pilots. By removing the need for an onboard human presence, UCAVs can be deployed in high-risk environments, including those contaminated by chemical, biological, radiological, or nuclear (CBRN) threats, or in areas under intense enemy fire. This capability greatly reduces the potential for human casualties and allows for more aggressive and daring operational tactics. Furthermore, UCAVs offer enhanced endurance and operational flexibility. Unlike manned aircraft, which are limited by human physiological constraints such as the need for rest and the impacts of fatigue, UCAVs can operate for extended periods, making them ideal for prolonged surveillance and reconnaissance missions. This endurance is particularly valuable in modern warfare, where persistent surveillance and intelligence gathering are critical for maintaining a tactical advantage over adversaries. However, the increasing deployment of UCAVs in military operations has led to the development of sophisticated countermeasures, including advanced air defense systems specifically designed to neutralize these unmanned threats. Among these countermeasures, high-speed missiles capable of reaching speeds up to 8 Mach have emerged as a significant threat to UCAVs, which typically have a maximum speed of around 2.5 Mach. This speed disparity presents a considerable challenge for UCAVs, as they must employ highly effective evasive maneuvers to survive in environments where they are targeted by such fast-moving threats. To address this challenge, researchers and engineers are developing innovative solutions that leverage advanced technologies such as artificial intelligence (AI) and machine learning. One particularly promising approach is the use of deep reinforcement learning (DRL) to generate online missile-evading maneuvers for combat aerial vehicles. DRL, a subset of machine learning, involves training algorithms through a process of trial and error, where the system learns to optimize its actions based on feedback from its environment. In the context of UCAVs, the DRL algorithm is designed to take direct control of the aircraft's aileron, rudder, and elevator, which are critical control surfaces for maneuvering. This setup allows the system to explore a wide range of potential escape maneuvers, optimizing them in real-time to evade incoming missile threats. The real-time nature of this engagement is crucial, as UCAVs must react swiftly to threats that can travel at hypersonic speeds, leaving little room for error or delay. Extensive simulations conducted using this DRL-based methodology have shown promising results, with a reported success rate of 88%. This high success rate indicates that the system is capable of effectively learning and applying evasive strategies that can withstand the demands of real-world combat scenarios. The ability to autonomously execute these maneuvers not only enhances the survivability of UCAVs but also extends their operational utility, allowing them to operate more safely in contested environments where the threat of missile attacks is high. The implications of integrating such advanced AI-driven maneuvering systems into UCAV operations are profound. This technology represents a significant leap forward in the development of autonomous systems, merging the fields of artificial intelligence and aeronautical engineering to create more resilient and adaptive aerial platforms. These advancements not only improve the defensive capabilities of UCAVs but also open up new possibilities for their use in a broader range of military and non-military applications, including disaster response, border security, and environmental monitoring. Moreover, the technology developed for UCAVs can potentially be adapted for use in other types of vehicles, both aerial and ground-based, further expanding its impact. For instance, autonomous ground vehicles could use similar DRL-based systems to navigate complex environments or evade threats, enhancing their utility in military and civilian contexts alike. In conclusion, the development of deep reinforcement learning approaches for missile evasion marks a significant milestone in enhancing the capabilities of UCAVs. As air defense systems continue to evolve, the ability of UCAVs to autonomously and effectively evade threats will be crucial in maintaining their strategic advantage. Continued research, development, and testing of these systems are essential to ensuring their effectiveness in real-world scenarios, ultimately contributing to the broader goal of maintaining air superiority in increasingly complex and contested battle spaces. This ongoing innovation in autonomous technology promises not only to enhance military capabilities but also to lead to broader applications in various fields, showcasing the transformative potential of AI in shaping the future of aerial and ground-based operations
-
ÖgeAircraft trajectory optimization under wind effect by using optimal control : Environmental impact assessment(Graduate School, 2022)The aim of this thesis, Aircraft Trajectory Optimization Under Wind by Using Optimal Control and Environmental Impact of Aviation in terms of Aircraft Emissions, is to find the wind and weather optimized aircraft trajectories in the cruise phase by minimizing fuel consumption, time and air pollutants. Flight trajectories calculated by taking into account the wind factor are considered as a critical measure in terms of reducing fuel consumption. In addition, it is known that the models examined by including weather information give more realistic results than those that are not included. Trajectory planning calculations consist of various elements such as wind forecasts, operational constraints, amount of fuel, aircraft performance, atmospheric conditions. Temperature, pressure and air density parameters are considered standard atmospheric values. The performance model used is based on BADA. In order to achieve the aims of the thesis, first of all, the problem has been tried in 2 dimensions in terms of reducing complexity of operations. At this stage, the wind equation, which was created with a simple calculation, was added to the EoM in horizontal plane. The effect of the horizontal components of wind is clearly seen in the numerical simulation. Secondly, the problem was created according to BADA 3 and solved in a way that minimizes flight time and fuel in 3 dimensional space. Simultaneously, a wind model has been created with the wind tabular data obtained from the Global Forecast System (GFS). The GFS is a weather forecast model developed by the National Centers for Environmental Prediction (NCEP). In this study, wind factor was assumed to be stationary, wind uncertainty was not included in this study. Since there are erroneous measurements in the wind tabular data obtained from the Global Forecasting System (GFS), the data was improved by applying the interpolation method first and the error difference between the real data and the interpolated data was arranged to be the least. Then, with the smooth data, wind equations were obtained separately for seven barometric altitude levels. Thirdly, in addition to flight time and fuel consumption, an emission model was created based on the ICAO Engine Exhaust Data Bank [29] and Boeing Method 2 [30] to solve the multi-objective optimization problem. The developed new model was applied to the simulation environment created based on BADA 4. Finally, the wind equations in the horizontal plane obtained were included in the simulation environment developed on the basis of BADA 4, and the targeted model was created. Predetermined routes were filtered from the actual flight plan selected for the same day with the wind data to be examined in the case studies. The flight area was determined and a wind model was obtained for that region. All one-day flights for the selected route were examined. Wind equations are calculated by taking flight hours into account. The simulation results were obtained according to the flight information of the desired route obtained from the real flight plan. The optimized trajectories were calculated in the simulation environment by referring to the points where the aircraft started and ended the cruise phase. Thus, Turkish airspace, which has not been examined before, is presented as a case study specific to Istanbul-Ankara flights. As a second case study, European airspace is presented specific to Paris-Frankfurt flights. During these studies, it was clearly seen that cruising speed and cruising altitude are critical for fuel consumption under the wind effect. In addition, as a result of these studies, it has been shown that the proposed model gives more effective results as the flight distance increases. This study consists of five chapters describing the stages of the thesis. The first chapter is a general introduction to the thesis topic. The thesis topic is explained and its aims are mentioned, the importance of the subject and why it is needed are presented. This section consists of three sub-titles. First of all, the scope and contributions of the thesis are mentioned. Afterwards, a wide literature review was made and studies in this field were presented. Finally, the structure of the thesis is mentioned. In the second chapter, the mathematical model required for this study is explained. Wind-optimized trajectories for an aircraft in the cruise phase are generated by solving a non-linear optimal control problem. For this reason, first of all, the general representation of the optimal control problem and its solution techniques are mentioned. The suitability of these solution techniques to the problem is discussed and the method to solve the problem is explained. It is known that multi-objective optimization problems give more realistic and ideal results than single-objective optimization problems. After this part, multi-objective optimization problem and constraints are defined. In the last part, the optimal control problem solution method is mentioned. GEKKO Python optimization module was used for numerical simulation in solving the aircraft trajectory optimization problem. This algorithm was developed to analyze the environmental impact of emitted aircraft emissions such as nitrogen oxides and carbon dioxide, using real air traffic data. In the third chapter, models used in trajectory generation optimized for wind and weather conditions are introduced. First, the assumptions are mentioned. Afterwards, the atmosphere model, aircraft performance model, wind model and emission model are explained in detail. In addition, the equations of motion of the aircraft in 2D and 3D are shown in this section. In the fourth chapter, two case studies on the subject and their results are presented. First of all, it is the main contribution to the literature to analyze the flights over Turkey, which has not been focused on before as a case study. First, the problem is defined in the case analysis. Then, the wind field over the Turkish airspace was examined and a wind model was created. The wind equations of the region, which was extracted to include Istanbul and Ankara, were obtained and added to the equations of motion of an aircraft, as explained in wind model section in third chapter. In the simulation environment created in this direction, the most optimized trajectories were calculated considering the Istanbul-Ankara flights. As the second case study, Paris-Frankfurt flights over European airspace were analyzed. As in the first application, after defining the problem respectively and examining the wind field covering the flight points, the multi-objective optimization problem was solved for this route. As a result of the case studies, the actual and calculated flight time, fuel consumption, NOx and CO2 emission findings for each flight are presented comparatively. In the fifth and also the last chapter, the results and other studies that can be done in this field are mentioned. The values obtained as a result of the case analyzes are emphasized again. Within the scope of the study, it has been shown that the adverse impacts of aviation on the climate are reduced by trajectory optimization, which is resolved by evaluating wind and environmental effects. The topics that can be studied on the basis of this study in the future are mentioned.
-
ÖgeFighter pilot behavior cloning and transferring to another aircraft(Graduate School, 2022)he "Pilot-in-the-loop" flight simulators are important tools in the development of fighter aircraft because they allow engineers and designers to test different scenarios and algorithms in a controlled environment without the risks associated with actual flight testing. However, these simulations often require high pilot skill levels and can be time-consuming and costly to arrange. It is important to create realistic models of human fighter pilots in order to reduce the reliance on skilled pilots to demonstrate agile/aerobatic maneuvers in flight simulators. Traditional controllers for aircraft require detailed knowledge about the aerodynamic model and physics of the aircraft in order to perform aerobatic maneuvers. Also, these control algorithms may not be able to match the performance of skilled human pilots, who are limited by a lack of bandwidth. This suggests that there may be potential for improving the performance of aircraft through the use of techniques that can take advantage of the superior speed and maneuverability of skilled pilots. In that case, imitation learning is a potential solution to eliminate the dependency need of skillful pilots in the flight simulator. Imitation learning also known as learning from demonstrations has benefited from computational progresses brought on by deep learning and increased availability of demonstration data. It is aimed to emulate desired behavior in a given task. An agent is trained to learn mapping between observations and actions by utilizing demonstrations. In this thesis aims the development of a pilot behavior model which is capable of autonomously performing agile maneuvers and is able to replace expert pilots' demonstrations over its full flight envelope in the flight simulator. Moreover, this model is transferable to other aircraft with limited data using transfer learning techniques. Besides all these features, the pilot behavior model can be able to run in real time in the flight simulator.
-
ÖgeDeveloping algorithm for automatic detection of caves using unmanned aerial vehicle data(Graduate School, 2023)Caves are underground, naturally occurring hollow spaces that are typically formed by the erosion of rock by water or other natural processes. They can range in size from small passages to vast underground networks, and are home to a wide variety of unique ecosystems and geological features. The exploration and mapping of caves is an important field of study for geologists, biologists, and other researchers, as it can provide valuable insights into the history of the earth and the processes that shape it. Furthermore, finding cave entrances is valueable for defence industry where these capabilities could greatly aid in counter-terrorism effort. Also, it is important to find cave like structures for archaeological exploration because the first settlemenents of humans are caves. In recent years, the use of unmanned aerial vehicles (UAVs) for cave exploration has become increasingly popular, due to their ability to access hard-to-reach areas and collect data from a safe distance. UAVs can be equipped with a range of sensors, such as cameras, Light Detection and Ranging (LIDAR), and thermal imaging, to collect data on the cave environment, including its topography, geology, and biological communities. In the past, the exploration and mapping of caves typically involved physical exploration, with researchers often crawling through small passages to gather data on the cave environment. However, with the advent of UAVs and their ability to collect data from a safe distance, cave exploration has become more efficient and less risky. Data collected by UAVs can include optical data, thermal data, LIDAR data, ground-penetrating radar data, and remote sensing data. Thermal data is particularly useful for cave detection, as cave entrances behave differently from the surrounding environment. For example, during the hottest hour of the day, cave entrances will appear cooler, while during the coldest hour of the day, they will appear warmer. Optical data can also be used to locate caves, by capturing images and labeling them where caves are located to create a dataset for training cave detection deep learning models. Moreover, creating a digital elevation model (DEM) with using LIDAR data is a solution method for exploring caves which could find the hidden caves. Additionally, cave detection systems based on thermal or optical imagery can detect caves that reach the ground surface, they cannot find caves with hidden entrances but ground-penetrating radar can detect underground cavities and sinkholes. Finally, remote sensing with different mutispectral bands and panchromatic bands can be used to identify cave-like structures where the caves are difficult to access. UAVs are becoming an increasingly popular tool for collecting data in the field of cave exploration due to their many advantages over traditional methods. UAVs are easy to manufacture and can be built at a lower cost than manned aircraft. They are also easy to operate and do not put human pilots at risk. Additionally, UAVs have a much lower cost of operation than traditional aircraft, making them a cost-effective choice for researchers and students. UAVs are highly maneuverable, which allows them to access hard-to-reach areas and collect data from a variety of angles and perspectives. This makes it easier to obtain detailed information about the cave environment. Moreover, the collected data has a lower ground sampling distance, which means it can be used to produce more detailed maps and other informative products. In developing an algorithm for automatic detection of caves using UAV data, several key concepts and techniques from the field of computer vision are employed. These include convolutional neural networks (CNNs), which are a type of deep learning algorithm that can learn to recognize patterns in images and other types of data. Optimization methods such as stochastic gradient descent are used to train the CNNs on large datasets of labeled images. Object detection is a common task in computer vision, and involves identifying and localizing objects within an image. Evaluation metrics such as precision and recall are used to measure the performance of the algorithm on test data. In addition, thermal imaging is an important component of the data collected by the UAVs, as it can provide valuable information about the temperature distribution within the cave environment. When used in combination with optical imaging, which captures visible light, thermal imaging can help to distinguish between cave entrances and other types of openings or anomalies. The use of UAV imaging allows for high-resolution data to be collected over large areas quickly and safely, making it an ideal tool for cave exploration and mapping. Several important decisions had to be made to carry out our project. Firstly, a comparison was made between the advantages and disadvantages of different types of UAVs, including rotary wing and fixed wing. After careful consideration, it was decided that a rotary wing UAV would be more appropriate for the project due to its ability to easily land and navigate in wild environments. The DJI Maverick 2 Enterprise was then selected as the UAV of choice, due to its high-resolution camera and thermal imaging capabilities. The location for UAV flight to collect the required data was then decided upon. Cappadocia, Gökova, Oymapinar, and Mersin were selected for the experiments as they are known to have unique geological formations and cave-like structures. Finally, the object detection model YOLOV7 was chosen to analyze the data collected by the UAV. YOLOV7 is a state-of-the-art model that can quickly and accurately identify objects in images. A dataset was generated for training the YOLOV7 model, as there were no existing open-source datasets available. The model was trained to identify cave entrances in the optical images collected by the UAV. An algorithm was then developed to analyze the optical images and identify cave-like structures. Another algorithm was generated for analyzing the thermal images and identifying potential cave entrances. Finally, a decision-making algorithm was developed to combine the results from the optical and thermal imaging analyses to detect the most likely cave entrances. These algorithms were tested and refined using the dataset and imagery collected during the UAV flights in the selected locations. In conclusion, it can be demonstrated that cave entrances can be located using data collected by UAVs as shown in our project. Several advantages are offered by the use of UAVs, such as the ease of use, lower operating costs, and their ability to access hard-to-reach areas. By generating a real-world dataset and developing algorithms for object detection in both optical and thermal imagery, cave-like structures were successfully identified in our experimental sites. Great results were generated even with a limited amount of data, thanks to Deep Learning techniques like YOLOV7. Thermal imaging was found to be an effective way to locate caves due to their characteristic temperature patterns. Astonishing results were achieved by combining optical and thermal imagery with a decision-making algorithm, providing new insights into the detection of underground caves and sinkholes. Overall, the potential of UAV technology and Deep Learning for the identification of cave-like structures and their applications in geological and environmental studies were showcased in our project.
-
ÖgePanoramik görüntüler üzerinden su altı hedef tespiti: DBSCAN ve derin öğrenme ağları ile bütünleşik bir yaklaşım(Lisansüstü Eğitim Enstitüsü, 2024-07-10)Su altı akustik verileri, denizaltı savaşları, su altı hedef izleme, mayın tespiti, deniz güvenliği, su altı haritalama ve çeşitli su altı araştırmalarında kritik rol oynayan önemli bir çalışma konusudur. Bu alandaki sürekli gelişmeler, askeri amaçlı operasyonlardan bilimsel araştırmalara kadar geniş bir yelpazede uygulamalara olanak tanımaktadır. Su altı hedef tespiti, çeşitli sensörler, akustik sistemler ve görüntüleme teknikleri kullanılarak gerçekleştirilen önemli bir akustik faaliyet alanıdır. Bu sensörler su altındaki nesnelerin ses dalgalarını algılarlar. Böylece denizaltılar, torpidolar ve diğer su altı araçları tarafından kullanılarak düşman veya potansiyel tehditleri belirleme konusunda kritik bilgiler sağlarlar. Son yıllarda su altı ekipman sistemlerinin gelişmesi ile birlikte bu alanda yapılan çalışmalarda büyük bir artış meydana gelmiştir. Fakat, geniş su altı bölgelerindeki karmaşık topografyalar ve değişen çevresel koşulları nedeniyle su altı hedef tespiti kolay bir uygulama değildir ve çoğu zaman geleneksel tespit yöntemleri belirli bir alanda çalışmakta ve değişen çevre koşullarına uyum sağlamakta yeterli olmamaktadır. Bu nedenle bahsedilen zorluklarla baş edebilecek yeni yaklaşımların geliştirilmesine yönelik çalışmalar artmıştır. Bu tez çalışmasında, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) ve derin öğrenme ağlarının (yapay sinir ağlarının ve evrişimli sinir ağlarının) birleştirildiği yenilikçi bir yaklaşım sunulmuştur. Önerilen yaklaşımda simülasyon ortamında oluşturulmuş sonar verileri ve beamforming algoritması kullanılarak açı, frekans ve gürültü seviyesinden oluşan panoramik su altı resimleri oluşturulmuştur. Oluşturulan panorama resminde x ekseninde derece cinsinde açı değerleri, y ekseninde Hz cinsinden frekans verileri ve z ekseninde dB cinsinden gürültü seviyeleri bulunmaktadır. Oluşturulan veriler işlenerek kullanıma hazır hale getirilmiştir. Bu kapsamda ilk olarak açı ve frekans bandındaki ölçek farklılığını ortadan kaldırmak için frekans ve açı ekseni normalize edilmiş ve iki eksende 0 ile 1 değerleri arasına getirilmiştir. Gürültü seviyeleri ise z skor kullanılarak normalize edilmiştir. Veri işleme aşaması tamamlandıktan sonra model kısmına geçilmiştir. Sonar verilerinden oluşturulan panoramik deniz resimleri, ilk olarak DBSCAN algoritmasından geçirilmiştir. DBSCAN algoritması, su altı panoramik görüntülerdeki hedeflerin yoğunluk tabanlı bir şekilde kümelenmesini sağlar. Bu algoritma, geleneksel tespit yöntemlerinin ötesine geçerek, hedeflerin doğal olarak oluşan yoğunluk bölgelerinde daha etkili bir şekilde algılanmasını mümkün kılar. DBSCAN'in sunduğu bu avantajı kullanmak ve sistem performansını optimize etmek için, DBSCAN algoritması derin öğrenme ağlarıyla birleştirilmiştir. DBSCAN tarafından kümelenen veriler yapay sinir ağlarına (ANN) ve evrişimli sinir ağlarına (CNN) girdi olarak verilmiş ve modellerin hedef tespit performansları değerlendirilmiştir. Elde edilen sonuçlar sadece derin öğrenme ağları ile eğitilen modellerinin sonuçları ile karşılaştırılmıştır. Önerilen yaklaşım , sadece panoromik görüntülerle eğitilen modellerden, çok daha iyi performans elde etmiştir ve hedef tespiti konusunda daha başarılı olmuştur. Bu sonuçlar, derin öğrenme ağlarının, DBSCAN tarafından belirlenen yoğunluklu alanlarda daha spesifik ve hassas özellikler öğrenerek su altı hedeflerini daha doğru bir şekilde tanıma yeteneğinin olduğunu göstermiştir.