Unmanned aerial vehicles (UAVs) is a rapidly expanding area of research due to their versatile usage, such as inspection of places inaccessible to humans and surveillance missions. This creates a demand for a reliable model that can accurately describe the dynamics of the system in order to improve the performance of the vehicle. System identification is a common tool used for the modelling of a system and is essential for developing an accurate and reliable model. The aim of this master's thesis is to develop an accurate non-linear grey-box model, with six degrees of freedom, of a fixed-wing UAV as well as a linearized version of the model. After a literature study a suitable model structure with sixstates and 28 parameters was chosen. The moment of inertia matrix is estimated separately using physical experiments,and the other parameters, related to the aerodynamic coefficients of the UAV, are estimated using flight experiments. Flight experiments are designed in order to capture all of the system dynamics and data was collected accordingly. The parameters are estimated using a prediction error method, which requires the solution of an optimal control problem. The derived models of the UAV are compared to each other and evaluated using model validation. In conclusion, the non-linear grey-box model shows great potential in becoming an accurate model, but further investigation and refining of the model is necessary.

The influx of unmanned aerial systems over the last decade has increased need for airspace awareness. Monitoring solutions such as drone detection, tracking, and classification become increasingly important to maintain compliance for regulatory and security purposes, as well as for recognizing aircraft that may not be so. Vision systems offer significant size, weight, power, and cost (SWaP-C) advantages, which motivates exploration of algorithms to further aid with monitoring performance. A method to classify aircraft using vision systems to measure their motion characteristics is explored. It builds on the assumption that at least continuous visual detection or at most visual tracking of an object of interest is already accomplished. Monocular vision is in part limited by range/scale ambiguity, where range and scale information of an object projected onto the image plane of a camera using a pin- hole model is generally lost. In an indirect effort to attempt to recover scale information via identity, classification of aircraft can aid in improvement of. These measured motion characteristics can then be used to classify the perceived object based on its unique motion profile over time, using signal classification techniques. The study is not limited to just unmanned aircraft, but includes full scale aircraft in the simulated dataset used to provide a representative set of aircraft scale and motion. / Doctor of Philosophy / The influx of small drones over the last decade has increased need for airspace awareness to ensure they do not become a nuisance when operated by unqualified or ill-intentioned personnel. Monitoring airspace around locations where drone usage would be unwanted or a security issue is increasingly necessary, especially for more range and endurance capable fixed wing (airplane) drones. This work presents a solution utilizing a single camera to address the classification part of fixed wing drone monitoring, as cameras are extremely common, generally cheap, information rich sensors. Once an aircraft of interest is detected, classifying it can provide additional information regarding its intentions. It can also help improve visual detection and tracking performance since classification can help change expectations of where and how the aircraft may continue to travel. Most existing visual classification works rely on features visible on the aircraft itself or its silhouette shape. This work discusses an approach to classification by characterizing visually perceived motion of an aircraft as it flies through the air. The study is not limited to just drones, but includes full scale aircraft in the simulated dataset used. Video of an airplane is used to extract motion from each frame. This motion is condensed to and expressed as a single time signal, that is then classified using a neural network trained to recognize audio samples using a time-frequency representation called a spectrogram. This transfer learning approach with Resnet based spectrogram classification is able to achieve 90.9% precision on the simulated test set used.

The railway infrastructure needs to be safe, reliable and efficient in order to meet the growing demand of sustainable transportation methods. One of the main problems the railway industry faces today is that a higher traffic load increases the need for maintenance, at the same time as it reduces the availability of gaps in the timetables to perform maintenance activities. Unmanned Aerial Vehicles, UAVs, have in recent years been adopted commercially due to their potential of increasing work efficiency and productivity. Different actors in the railway industry have recently started to explore and test the possibilities of implementing UAVs. The objective of this master thesis was to investigate and define use case scenarios where the use of UAVs would create value for railway infrastructure operations and maintenance activities. It is meant for both stakeholders in the railway industry to gain better understanding of capabilities and limitations of UAV technology but also provide recommendations to UAV manufacturers to understand the railway industry and potential UAV applications. Theoretical research and qualitative user studies with UAV professionals and relevant stakeholders within the railway industry were conducted in order to gain insight in the railway industry and to identify potential use case scenarios. The research showed that maintenance activities to a large extent are performed either manually by walking along the tracks which is inefficient, physically demanding and dangerous or by using test/measurement vehicles which require track occupancy. It was concluded that the use of UAVs would mainly create value by; enabling remote inspection and operation, accessing the infrastructure without track occupancy or the need of roads. At the same time, improve the working conditions, efficiency and quality of maintenance activities. The thesis resulted in 15 potential use case scenarios for UAVs in the railway industry and proposals for common UAV solutions based on functional requirements. / Järnvägssystemet måste vara säkert, pålitligt och effektivt för att möta den växande efterfrågan på hållbara transportmetoder. Ett av de största problemen som den svenska järnvägsindustrin står inför idag är att ökad trafikbelastning ökar behovet av underhåll, samtidigt som det minskar tillgängligheten för att utföra underhållsaktiviteter. Obemannade flygfordon, även kallade drönare, har under de senaste åren tillämpats mer frekvent i kommersiella syften för att bland annat uppnå ökad effektivitet och produktivitet. Aktörer inom järnvägsindustrin har nyligen börjat utforska och testa möjligheterna att använda drönare. Syftet med detta examensarbete var att undersöka och definiera potentiella tillämpningar av drönare med syfte att skapa värde för drift- och underhållsarbete inom järnvägen. Denna rapport är avsedd för intressenter inom järnvägsindustrin att få bättre förståelse för kapaciteten och begränsningar av drönarteknik samt ge rekommendationer till drönartillverkare för att bättre förstå järnvägsindustrin och potentiella användningsområden. Teoretisk undersökning och kvalitativa användarstudier med drönarexperter och relevanta intressenter inom järnvägsindustrin genomfördes för att få insikt i järnvägsindustrin samt för att identifiera problemområden. Studien visade att underhållsverksamheten i stor utsträckning utförs antingen manuellt genom att gå längs spåren vilket är ineffektivt, fysiskt krävande och farligt eller genom att använda test/mätfordon som kräver tillgång till spår. Arbetet resulterade i 15 potentiella tillämpningar av drönare i järnvägsindustrin samt förslag på gemensamma drönarlösningar baserade på funktionella krav. Slutsatsen drogs att tillämpningen av drönare i järnvägsindustrin främst kan skapa värde genom att; på distans utföra underhållsaktiviteter och inspektioner, få tillgång till infrastrukturen utan behov av spår eller vägar. Detta resulterar i förbättrade arbetsförhållanden samt ökad effektivitet och kvalitet på underhållsarbetet.

Fixed-wing unmanned aerial vehicles (UAVs) with the ability to hover combine the speed and endurance of traditional fixed-wing fight with the stable hovering and vertical takeoff and landing (VTOL) capabilities of helicopters and quadrotors. This combination of abilities can provide strategic advantages for UAV operators, especially when operating in urban environments where the airspace may be crowded with obstacles. Traditionally, fixed-wing UAVs with hovering capabilities had to be custom designed for specific payloads and missions, often requiring custom autopilots and unconventional airframe configurations. With recent government spending cuts, UAV operators like the military and law enforcement agencies have been urging UAV developers to make their aircraft cheaper, more versatile, and easier to repair. This thesis discusses the use of the commercially available ArduPilot open source autopilot, to autonomously transition a fixed-wing UAV to and from hover flight. Software modifications were made to the ArduPilot firmware to add hover flight modes using both Proportional, Integral, Derivative (PID) Control and Model Reference Adaptive Control (MRAC) with the goal of making the controllers robust enough so that anyone in the ArduPilot community could use their own ArduPilot board and their own fixed-wing airframe (as long as it has enough power to maintain stable hover) to achieve autonomous hover after some simple gain tuning. Three new hover flight modes were developed and tested first in simulation and then in flight using an E-Flight Carbon Z Yak 54 RC aircraft model, which was equipped with an ArduPilot 2.5 autopilot board. Results from both the simulations and flight test experiments where the airplane transitions both to and from autonomous hover flight are presented.

In recent years, the interest in flying multiple Unmanned Aerial Vehicles (UAVs) in formation has increased. One challenging aspect of achieving this is the relative positioning within the swarm. This thesis evaluates two different methods for estimating the relative position and orientation between two fixed wing UAVs by fusing range measurements from Ultra-wideband (UWB) sensors and orientation estimates from Inertial Measurement Units (IMUs). To investigate the problem of estimating the relative position and orientation using range measurements, the performance of the UWB nodes regarding the accuracy of the measurements is evaluated. The resulting information is then used to develop a simulation environment where two fixed wing UAVs fly in formation. In this environment, the two estimation solutions are developed. The first solution to the estimation problem is based on the Extended Kalman Filter (EKF) and the second solution is based on Factor Graph Optimization (FGO). In addition to evaluating these methods, two additional areas of interest are investigated: the impact of varying the placement and number of UWB sensors, and if using additional sensors can lead to an increased accuracy of the estimates. To evaluate the EKF and the FGO solutions, multiple scenarios are simulated at different distances, with different amounts of changes in the relative position, and with different accuracies of the range measurements. The results from the simulations show that both solutions successfully estimate the relative position and orientation. The FGO-based solution performs better at estimating the relative position, while both algorithms perform similarly when estimating the relative orientation. However, both algorithms perform worse when exposed to more realistic range measurements. The thesis concludes that both solutions work well in simulation, where the Root Mean Square Error (RMSE) of the position estimates are 0.428 m and 0.275 m for the EKF and FGO solutions, respectively, and the RMSE of the orientation estimates are 0.016 radians and 0.013 radians respectively. However, to perform well on hardware, the accuracy of the UWB measurements must be increased. It is also concluded that by adding more sensors and by placing multiple UWB sensors on each UAV, the accuracy of the estimates can be improved. In simulation, the lowest RMSE is achieved by fusing barometer data from both UAVs in the FGO algorithm, resulting in an RMSE of 0.229 m for the estimated relative position.

In aviation, safety is a big concern. Knowing the position of an aircraft at all times is of high importance. Today most aircraft utilize Global Navigation Satellite Systems (GNSS) for localization and precision navigation because of the small position error which do not increase over time. However, recent research show that GNSS can easily be jammed or spoofed. An alternative navigation method is Visual Odometry (VO). VO is navigation through visual input and is a key-part in development of fully autonomous vehicles. This thesis investigates the Deep Learning-based Visual Odometry (DL-VO) for aircraft at altitudes over 100 m. DL-VO deployed at high altitude is almost none existing. Therefore, this thesis investigates the deployments of ground developed DL-VO in the aerial domain. DeepVO is a Frame-To-Frame optical flow estimation method which is trained supervised and end-to-end. The domain change, from ground to high altitude aerial, brought bigger issues and had larger impact on the performance than first though. The use of full 6 Degrees of Freedom (DoF) pose estimation increases the complexity and was much harder than 2D estimation (x, y, yaw). A good angle representation was of higher importance during training and testing in the aerial domain. Since in the aerial domain the full 3D rotation is not unique in all representations of the orientation and issues with Gimbal lock can occur. Results on simulated data show that the propose method fails to estimate 6 DoF poses. However, the reduced 2D estimations shows that a trajectory can be maintained even is drift is present. The result on real world dataset shows that it easier to recover scale at lower speeds and with a less down angled camera. The difference between simulated and non-simulated data has not been investigated to the extent that a fair assessment on how the method’s performance is affected by the data character. / Flygsäkerhet är av stor vikt inom flygindustin. Att som pilot alltid veta var planet befinner sig är av stor vikt. Global Navigation Satellite Systems (GNSS) är idag den mest använda metoden för lokalisering och precisionsnavigering då GNSS har liten felmarginal som inte förvärras över tid. Nyligen har forskare visat att GNSS kan lätt störas och alternativa lokaliseringsmetoder behövs. En av dem är Visual Odometry (VO). VO metoder försöker navigera sig i olika miljöer genom att estimera kamerors rörelser i sekvens av bilder. Det pågår mycket forsking på området då det är ett nyckelkoncept för autonoma fordon. Detta arbete undersöker användadet av Deep Learning-based Visual Odometry (DL-VO) för flygfarkoster på höjder över 100 m. Det är väldigt få som har testat DL-VO på annat än små drönare vilket skiljer från flygplan på högre höjd som stöter på andra problem där alla obejekt är väldigt små. Då forskingen på DL-VO för flygplan på högre höjd är minimal undersöker arbetet ett domän byte genom att ta en metod utveklad för markfordon och flytta den till flygdomänen. För att undersöka bytet av domän avändes en anpassad version av DeepVO nätverket. DeepVO använder sig av realtiv Frame-To-Frame optiskt flödes estimeringar och är tränad end-to-end enligt supervised learning metoden. Domän bytet, från mark till luft, medförde större problem än först trott och det ökade komplexiteten på problemet. Estimeringar med 6 frihetgrader är mer komplexa och en bra vinkel representation är av mycket större vikt. Minimering av vinklar under träningen skapade andra problem i flygdomänen än vad det gjorde på ursrungliga datasetet. Resultaten på simiulerad data visar att den framtagna metoden inte klarar estimeringar med 6 frihetgrader. Men om problemet reduceras så kan metoden estimaera 2D banor på en fixerad höjd i luften även om viss drift över tid existerar. Kameravinkeln och hastighet påverkar metodens förmåga att hålla en korrekt skala. Resultat på verklig data visar att det är lättare att uppnå korrekt skala vid lägre hastighet och mindre nervinklad kamera. Skillnaderna mellan simulerad och verklig data har inte undersökts i den utsträktning som behövs för att göra en korrekt slutsats om dess efftekter på resultatet.

<p> Within recent years, there has been a renewed interest in developing data-driven learning based algorithms for solving longstanding challenging control problems. This interest is primarily motivated by the availability of ubiquitous data and an increase in computational resources of modern machines.  However, there is a prevailing concern on the lack of provable performance guarantees on data-driven/model-free learning based control algorithms. This dissertation focuses the following key aspects: i) with what facility can state-of-the-art learning-based control methods eke out successful performance for challenging flight control applications such as aerobatic maneuvering?; and ii) can we leverage well-established tools and techniques in control theory to provide some provable guarantees for different types of learning-based algorithms?  </p> <p>To these ends, a deep RL-based controller is implemented, via high-fidelity simulations, for Fixed-Wing aerobatic maneuvering. which shows the facility with which learning-control methods can eke out successful performances and further encourages the development of learning-based control algorithms with an eye towards providing provable guarantees.<br> </p> <p>Two learning-based algorithms are also developed: i) a model-free algorithm which learns a stabilizing optimal control policy for the bilinear biquadratic regulator (BBR) which solves the regulator problem with a biquadratic performance index given an unknown bilinear system; and ii) a model-free inverse reinforcement learning algorithm, called the Model-Free Stochastic inverse LQR (iLQR) algorithm, which solves a well-posed semidefinite programming optimization problem to obtain unique solutions on the linear control gain and the parameters of the quadratic performance index given zero-mean noisy optimal trajectories generated by a linear time-invariant dynamical system. Theoretical analysis and numerical results are provided to validate the effectiveness of all proposed algorithms.</p>

The goal of this bachelor thesis has been to evaluate and test the available open source software and commercial hardwarefor potential later use as the electrical system in the ALPHAUAV. ALPHA is a student project, with the goal of building an autonomous drone capable of high altitude, long-endurance missions to gather data from electromagnetic phenomena in the atmosphere. Data later to be used in research at the facility ofSpace and Plasma physics at KTH. The evaluation has been done by constructing of an MVP, to prove that the open source softwareand commercial hardware can be used to build an autonomousUAV. / Målet med denna kandidatuppsats har varit att evaluera och testa öppen källkod tillsammans med kommersiell hårdvara för att potentiellt kunna nyttjas som elektriskt system i ALPHA UAV. ALPHA UAV är ett studentprojekt, med målet att bygga en autonom drönare kapabel att genomföra höghöjdsflygningar med lång uthållighet för att kunna samla in data från elektromagnetiska fenomen i atmosfären. Data som senare kan nyttjas i forskningssyfte på institutionen för rymd-och plasmafysik på KTH. Evalueringen har gjorts genom att konstruera en MVP, för att bevsia att öppen källkod och kommersiell hårdvara kan nyttjas för att bygga en autonom UAV. / Kandidatexjobb i elektroteknik 2022, KTH, Stockholm

With the proliferation of unmanned aircraft designed for national security and commercial purposes, opportunities exist to create high-fidelity aerodynamic models with flight test techniques developed specifically for remotely piloted aircraft. Then, highly maneuverable unmanned aircraft can be employed to their greatest potential in a safe manner using advanced control laws. In this dissertation, novel techniques are used to identify nonlinear, coupled, aerodynamic models for fixed-wing, unmanned aircraft from flight test data alone. Included are quasi-steady and unsteady nominal flight models, aero-propulsive models, and spinning flight models. A novel flight test technique for unmanned aircraft, excitation with remote uncorrelated pilot inputs, is developed for use in nominal and nonlinear flight regimes. Orthogonal phase-optimized multisine excitation signals are also used as inputs while collecting gliding, aero-propulsive, and spinning flight data. A novel vector decomposition of explanatory variables leads to an elegant model structure for stall spin flight data analysis and spin aerodynamic modeling. Results for each model developed show good agreement between model predictions and validation flight data. Two novel applications of aerodynamic modeling are discussed including energy-based nonlinear directional control and a spin flight path control law for use as a flight termination system. Experimental and simulation results from these applications demonstrate the utility of high-fidelity models developed from flight data. / Doctor of Philosophy / This dissertation presents flight test experiments conducted using a small remotely controlled airplane to determine mathematical equations and parameter values, called models, to describe the airplane's motion. Then, the models are applied to control the path of the airplane. The process to develop the models and predict an airplane's motion using flight data is described. New techniques are presented for data collection and analysis for unusual flight conditions, including a spinning descent. Results show the techniques can predict the airplane's motion very well. Two experiments are presented demonstrating new applications and the usefulness of the mathematical models.

Vision-based conventional landing of a fixed wing UAV is addressed in this thesis. The work includes mathematical modeling, interface to a software for rendering the outside scenery, image processing techniques, control law development and outdoor experimentation. This research focuses on detecting the lines or the edges that flank the landing site, use them as visual cues to extract the geometrical parameters such as the line co-ordinates and the line slopes, that are mapped to the control law, to align and conventionally land the fixed wing UAV. Pre-processing and image processing techniques such as Canny Edge detection and Hough Transforms have been used to detect the runway lines or the edges of a landing strip. A Vision-in-the-Loop Simulation (VILS) set up on a personal computer or laptop, has been developed, without any external camera/equipment or networking cables that enables visual serving toper form vision-based studies and simulation. UAV mass, inertia, engine and aero data from literature has been used along withUAV6DOF equations to represent the UAV mathematical model. The UAV model is interfaced to a software using UDP data packets via ports, for rendering the outside scenery in accordance with the UAV’s translation and orientation. The snapshots of the outside scenery, that is passed through an internet URL by including the ‘http’ protocol, is image processed to detect the lines and the line parameters for the control. VILS set has been used to simulate UAV alignment to the runway and landing. Vision-based alignment is achieved by rolling the UAV such that the landing strip that is off center is brought to the center of the image plane. A two stage proportional aileron control input using the line co-ordinates, bringing the midpoints of the top ends of the runway lines to the center of the image, followed by bringing the mid points of the bottom ends of the runway lines to the center of the image has been demonstrated through simulation. A vision-based control for landing has been developed, that consists of an elevator command that is commiserate with the acceptable range of glide slope followed by a flare command till touch down, which is a function of the flare height and estimated height from the 3rd order polynomial of the runway slope obtained by characterization. The feasibility of using the algorithms for a semi-prepared or unprepared landing strip with no visible runway lines have also been demonstrated. Landing on an empty tract of land and in poor visibility condition, by synthetically drawing the runway lines based on a single 3rd order slope. vs height polynomial solution are also presented. A fixed area, and a dynamic area search for the Hough peaks in the Hough accumulator array for the correct detection of lines are addressed. A novel technique for crosswind landing, quite different from conventional techniques, has been introduced, using only the aileron control input for correcting the drift. Three different strategies using the line co-ordinates and the line slopes, with varying levels of accuracy have been presented and compared. About 125 landing data of a manned instrumented prototype aircraft have been analysed to corroborate the findings of this research. Outdoor experiments are also conducted to verify the feasibility of using the line detection algorithm in a realistic scenario and to generate experimental evidence for the findings of this research. Computation time estimates are presented to establish the feasibility of using vision for the problem of conventional landing. The thesis concludes with the findings and direction for future work.