Thesis (MScEng)-- Stellenbosch University, 2012. / ENGLISH ABSTRACT: This thesis presents the analysis, design, simulation and practical implementation of a control system to achieve an accurate autonomous landing of a fixed-wing unmanned aerial vehicle in the presence of wind gust atmospheric disturbances. Controllers which incorporate the concept of direct-lift control were designed based on a study of the longitudinal dynamics of the UAV constructed as a testbed. Direct-lift control offers the prospect of an improvement in the precision with which aircraft height and vertical velocity can be controlled by utilising actuators which generate lift directly, instead of the conventional method whereby the moment produced by an actuator results in lift being indirectly generated. Two normal specific acceleration controllers were designed. The first being a conventional moment-based controller, and the second a direct-lift-augmented controller. The moment-based controller makes use of the aircraft’s elevator while the direct-lift augmented controller in addition makes use of the flaps of the aircraft which serve as the direct-lift actuator. Controllers were also designed to regulate the airspeed, altitude, climb rate, and roll angle of the aircraft as well as damp the Dutch roll mode. A guidance controller was implemented to allow for the following of waypoints. A landing procedure and methodology was developed which includes the circuit and landing approach paths and the concept of a glide path offset to calibrate the touchdown point of a landing. All controllers and the landing procedure were tested in a hardware-in-the-loop simulation environment as well as practically in a series of flight tests. Five fully autonomous landings were performed, three of these using the conventional NSA controller, and the final two the direct-lift-augmented NSA controller. The results obtained during the landing flight tests show that the project goal of a landing within five meters along the runway and three meters across the runway was achieved in both normal wind conditions as well as in conditions where wind gusts prevailed. The flight tests also showed that the direct-lift-augmented NSA controller appears to achieve a more accurate landing than the conventional NSA controller, especially in the presence of greater wind disturbances. The direct-lift augmented NSA controller also exhibited less pitch angle rotation during landing. / AFRIKAANSE OPSOMMING: Hierdie tesis verteenwoordig die analise, ontwerp, simulasie en praktiese implementering van ’n beheerstelsel wat ten doel het om ’n akkurate en outonome landing van ’n onbemande vastevlerk vliegtuig in rukwind atmosferiese toestande te bewerkstellig. Gegrond op ’n studie van die longitudinale dinamika van die vliegtuig wat as proeftuig gebruik is, is beheerders ontwerp wat die beginsel van direkte-lig insluit. Direkte-lig beheer hou die potensiaal in om die vliegtuig se hoogte en vertikale snelheid akkuraat te beheer deur gebruik te maak van aktueerders wat lig direk genereer in teenstelling met die konvensionele metode waar die moment van die aktueerder indirek lig genereer. Twee normaal-versnellings beheerders is ontwerp. Die eerste is ’n konvensionele moment-gebaseerde beheerder wat gebruik maak van die hys-aktueerder van die vliegtuig, en die tweede is ’n direkte-lig-bygestaande beheerder wat addisioneel gebruik maak van die flappe van die vliegtuig wat as die direkte-lig aktueerder dien. Vedere beheerders is ontwerp wat die lugspoed, hoogte, klimkoers, en rolhoek van die vliegtuig reguleer asook die “Dutch roll” gedrag afklam. ’n Leiding-beheerder wat die volg van vliegbakens hanteer, is ingestel. Die landingsprosedure en -metodologie is ontwikkel wat die landingspad sowel as die sweef-pad bepaal en wat terselfdertyd ’n metode daarstel om die posisie van die landingspunt te kalibreer. Die beheerders en landingsprosedure is in ’n hardeware-in-die-lus omgewing gesimuleer en deur middel van ’n reeks proefvlugte getoets. Vyf ten volle outonome landings is uitgevoer waarvan drie van die konvensionele normaal-versnellings beheerder gebruik gemaak het, en die laaste twee die direkte-lig-bygestaande normaal-versnellings beheerder. Die vlugtoetsuitslae bevestig dat die navorsingsdoel om ’n landing binne vyf meter in lyn met en drie meter dwarsoor die landingstrook te bewerkstellig, behaal is. Hierdie akkuraatheid is verkry in beide goeie atmosferiese toestande sowel as toestande met rukwinde. Volgens die vlugtoetse blyk dit dat die direkte-lig-bygestaande normaalversnellings beheerder ’n meer akkurate landing kan bewerkstellig as die konvensionele normaal-versnellings beheerder, veral dan in toestande met rukwinde. Die direkte-ligbygestaande normaal-versnellings beheerder het ook ’n laer hei-hoek rotasie tydens die landing vertoon.

In the tracking of a moving ground target by small unmanned air vehicle (UAV) via camera vision, the target position and motion cannot be measured directly. Two different types of filters were assessed for their ability to estimate target motion, namely target velocity, directional heading on flat ground and distance from the UAV to target. The first filter is a nonlinear deterministic filter with stability guarantee. The second filter is based on nonlinear Kalman Filter technique. The application and performance of these two filters are presented, for simulated vision based target tracking.

Unmanned Aerial Vehicles (UAVs) are becoming vital warfare platforms because they significantly reduce the risk of human life while accomplishing important missions. A UAV can be used for example, as stand-in sensor for the detection of mobile, low-probability-of-intercept battlefield surveillance and fire control emitters. With many UAVs acting together as a swarm, the location and frequency characteristics of each emitter can be accurately determined to continuously provide complete battlefield awareness. The swarm should be able to act autonomously while searching for targets and relaying the information to all swarm members. In this thesis, two methods of autonomous control of a UAV swarm were investigated. The first method investigated was the Particle Swarm Optimization (PSO) algorithm. This technique uses a non-linear approach to minimize the error between the location of each particle and the target by accelerating particles through the search space until the target is found. When applied to a swarm of UAVs, the PSO algorithm did not produce the desired performance results. The second method used a linear algorithm to determine the correct heading and maneuver the swarm toward the target at a constant velocity. This thesis shows that the second approach is more practical to a UAV swarm. New results are shown to demonstrate the application of the algorithm to the swarm movement.

A new control law is being developed and implemented for the Vision Based Target Tracking (VBTT) system onboard a small unmanned aerial vehicle (SUAV). The new control law allows for coordinated SUAV guidance and vision-based target tracking of stationary and moving targets in the presence of atmospheric disturbances and measurements noise. The new control law is tested for its performance and stability in both the theoretical 6DOF simulation and the Hardware-in-the-Loop (HIL) simulation. Principal results show that realistic measures of performance of the control law are continuous and exhibit predictable degradation of performance with increase of target speed. The results are encouraging and comparable among theoretical predictions, actual hardware simulation results, and initial flight testing. The control law development, implementation, and trial processes and procedures are also examined and categorically documented in this thesis as future reference on the subject development, as well as for better knowledge retention, continuation and proliferation of the VBTT system.

This thesis was part of the ongoing research conducted at the Naval Postgraduate School to achieve greater collaboration between heterogeneous autonomous vehicles. The research addresses optimal control issues in the collaboration between an Unmanned Aerial Vehicle (UAV) and Autonomous Ground Vehicles (AGV). The scenario revolves around using the camera onboard the UAV to extend the effective field of view of the AGV. For military operations, this could be helpful in improving security for convoys and riverine patrols. There were three main problems addressed in this thesis. The first problem dealt with the design of a UAV control law that takes into consideration the relative speed differences between the UAV and the AGV. The UAV was assumed to have a greater speed compared to the AGV in this thesis. The second was the keystone field of view projection effect of the UAVâ s onboard camera onto the earth. The image captured by the camera was distorted due to the view angle of the camera from a high elevation. The third problem addressed was control of the location of the UAV to ensure the reliability of the communication network between the UAV and the AGV. The communication was assumed to be a linear function of the relative positions of the UAV and the AGV.

Approved for public release; distribution is unlimited / The NPS Tactical Horizon Extension Project objective is to define and demonstrate a concept by which task force-level commanders and below can obtain a persistent, over-the-horizon surveillance capability for the purpose of target development and other missions without tasking national or theater-level assets. Our goal is to increase the ISR capacity of units who normally would not rate the priority to task a Predator, Global Hawk, or U-2. There are two guiding tenets in developing this concept. First, the equipment and its control should be organic to the SOF unit or task force. Second, utilizing this capability should not require the soldier to carry any additional equipment into the field. Initial research led us to the idea of using networked unmanned aerial systems (UAS's) to generate an over-the-horizon surveillance capability for SOF. We demonstrated the concept by forming a network comprised of a forward ground team, an inexpensive, test-bed UAS equipped with an off-the-shelf video camera, a manned aircraft, and a tactical operations center (TOC). We attained connectivity through an ITT Mesh structure at 2.4 GHz, amplified to 1W. Researchers were from the Defense Analysis, Mechanical and Astronautical Engineering, and Information Sciences Departments. We conducted successful experiments through the USSOCOM-NPS Cooperative Field Experimentation Program. / Outstanding Thesis

無人直昇機是指在沒有搭載駕駛員下，能自動飛行的直昇機。這種飛行器除了不用令駕駛員冒上性命風險，亦由於其獨特飛行能力，例如空中旋停、急彎打轉等，令無人直昇機在執行危險任務上有著重要的優勢。但同時，由於沒有了駕駛員的關係，這種飛行器持續面對著一系列的技術難點。當中包括：（一）在飛行上，無人直昇機帶有一種未能確解的動態反應，偶合在主控軸上，令動態變化的預測變得困難；（二）絶大部份無人直昇機的導航倚賴全球定位儀，但這種定位儀卻極易受干擾，影響飛行安全；（三）由於每次任務的需要，機上搭載物的位置及多寡令飛行器上的重力分佈經常改變，令自動控制器需面密集的手動調度及系統識別。 / 對於無人直昇機的動態模型建構，我們提出了一種新辦法令過往未能確解的偶合動態反應，能被準確的量化及分析，因而使這種模型能更準確預測無人直昇機的動態變化。我們的辦法針對過往一種普遍使用但不能預測偶合動態反應，名為「遲相」的模型方法，並提出以主旋槳上的陀螺現像，去描述無人直昇機上的偶合動態反應。另外，我們的模型分析顯示出準確預測無人直昇機的動態反應需要識別一系列難以度測而又經常變化的系統參數，因此我們提出以一個相對隱性模型的辦法對解決往後有關導航及控制的問題更有利。 / 有關無人直昇機的導航，我們提出以慣性感測器去補償全球定位儀因干擾及訊號異常導致的感測數據中斷。通過一系列的噪音處理，我們的辦法能從慣性感測器的加速度讀數轉化為位移，從而令無人直昇機能在突如其來的全球定位儀中斷作出即時反應，避免因位置讀數中斷而墜毀。除此以外，我們亦提供全自動參數調節算法，令整個噪音處理能快速應用於不同慣性感測器的設置上。 / 有關無人直昇機的控制，由於一般需要廣泛的系統建模及識別，因此我們提出以「能力轉移」的概念，透過駕駛員的示範，提取控制直昇機的技巧從而令機器無需建構準確和全面的動態模型下，能準確控制無人直昇機。我們通過建立一個簡明的反饋模型，利用的駕駛員的示範而自動配置出最適合的參數去控制無人直昇機。 / 從更準確的動態模型分析，到抗干擾導航及新型控制器的設計，整個研究為無人直昇機的發展提出了多方面的觀點及解決辦法。與此同時，我們的最終目的是希望為無人直昇機的日常應用上掃除因為缺少了駕駛員而做成的障礙，令無人直昇機能真正以最少的人為干涉而達至全面的無人操作。 / Unmanned helicopters hold both tremendous potential and challenges. Without risking the lives of human pilots, these vehicles exhibit agile movement and the ability to hover and hence open up a wide range of applications in the hazardous situations. Sparing human lives, however, comes at a stiff price for technology. Some of the key difficulties that arise in these challenges are: (i) There are unexplained cross-coupled responses between the control axes on the hingeless helicopters that have puzzled researchers for years. (ii) Most, if not all, navigation on the unmanned helicopters relies on Global Navigation Satellite Systems (GNSSs), which are susceptible to jamming. (iii) It is often necessary to accommodate the re-configurations of the payload or the actuators on the helicopters by repeatedly tuning an autopilot, and that requires intensive human supervision and/or system identification. / For the dynamics modeling and analysis, we present a comprehensive review on the helicopter actuation and dynamics, and contributes toward a more complete understanding on the on-axis and off-axis dynamical responses on the helicopter. We focus on a commonly used modeling technique, namely the phase-lag treatment, and employ a first-principles modeling method to justify that (i) why that phase-lag technique is inaccurate, (ii) how we can analyze the helicopter actuation and dynamics more accurately. Moreover, these dynamics modeling and analysis reveal the hard-to-measure but crucial parameters on a helicopter model that require the constant identifications, and hence convey the reasoning of seeking a model-implicit method to solve the state estimation and control problems on the unmanned helicopters. / For the state estimation, we present a robust localization method for the unmanned helicopter against the GNSS outage. This method infers position from the acceleration measurement from an inertial measurement unit (IMU). In the core of our method are techniques of the sensor error modeling and the filtering method for the sensor noise compensation. Moreover, we provide a fully automatic algorithm to tune our method. Finally, we evaluate our method on an instrumented gasoline helicopter. Experiments show that the technique enables the robust positioning of flying helicopters when no GNSS measurement is available. / The design of an autopilot for an unmanned helicopter is made difficult by its nonlinear, coupled and non-minimum phase dynamics. Here, we consider a reinforcement learning approach to transfer motion skills from human to machine, and hence to achieve autonomous flight control. By making efficient use of a series of state-and-action pairs given by a human pilot, our algorithm bootstraps a parameterized control policy and learns to hover and follow trajectories after one manual flight. One key observation our algorithm is based on is that, although it is often difficult to retrieve the human pilots’ hidden desiderata that formulate their state-feedback mechanisms in controlling the helicopters, it is possible to intercept the states of a helicopter and the actions by a human pilot and then to fit both into a model. We demonstrate the performance of our learning controller in experiments. / The results described in this dissertation shed new and important light on the technology necessary to advance the current state of the unmanned helicopters. From a comprehensive dynamics modeling that addresses perplexing cross-couplings on the unmanned helicopters, to a robust state estimation against GNSS outage and a learn-from-scarcesample control for an unmanned helicopter, we provide a starting point for the cultivation of the next-generation unmanned helicopters that can operate with the least possible human intervention. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Lau, Tak Kit. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2012. / Includes bibliographical references (leaves 182-198). / Abstract also in Chinese. / Abstract --- p.i / Acknowledgement --- p.v / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- The State-of-the-Art --- p.5 / Chapter 1.2 --- Problemand Approach --- p.8 / Chapter 1.3 --- Contributions and Organization of Chapters --- p.13 / Chapter 2 --- Actuation and Dynamics of Helicopter --- p.16 / Chapter 2.1 --- Introduction --- p.17 / Chapter 2.2 --- Input-Output Relation of Hingeless Helicopters --- p.20 / Chapter 2.3 --- Helicopter Dynamics --- p.30 / Chapter 2.3.1 --- Aerodynamics --- p.34 / Chapter 2.3.2 --- Discrepancy and Conventional Treatment --- p.36 / Chapter 2.3.3 --- Gyroscopically InducedMoment --- p.38 / Chapter 2.3.4 --- Example of Pitching Forward --- p.42 / Chapter 2.4 --- Empirical Evaluations --- p.44 / Chapter 2.4.1 --- Numerical Simulation --- p.45 / Chapter 2.4.2 --- Experiments --- p.47 / Chapter 2.5 --- Summary --- p.48 / Chapter 3 --- State Estimation against GNSS Outage --- p.53 / Chapter 3.1 --- Introduction --- p.55 / Chapter 3.2 --- Sensor Noises --- p.62 / Chapter 3.2.1 --- Sensitivity Analysis of Estimation Accuracy --- p.63 / Chapter 3.2.2 --- Sensor Error Characteristics --- p.68 / Chapter 3.2.3 --- Random Walk Acceleration Noise --- p.71 / Chapter 3.3 --- State Estimation Algorithm --- p.76 / Chapter 3.3.1 --- Process and Measurement Models --- p.79 / Chapter 3.3.2 --- Selection of OperationModes --- p.83 / Chapter 3.3.3 --- Prioritized Propagation of States --- p.85 / Chapter 3.4 --- Automatic Tuning of Filter Parameters --- p.87 / Chapter 3.4.1 --- Performance Index --- p.88 / Chapter 3.4.2 --- Tuning Algorithm --- p.90 / Chapter 3.5 --- Experimental Results --- p.93 / Chapter 3.6 --- Summary --- p.97 / Chapter 4 --- Model-based Control for Unmanned Helicopters --- p.106 / Chapter 4.1 --- Introduction --- p.107 / Chapter 4.2 --- Effect of Center of Gravity --- p.108 / Chapter 4.3 --- Design of Attitude Controller --- p.110 / Chapter 4.4 --- Augmentation in Controller and Experimental Results --- p.113 / Chapter 4.5 --- Summary --- p.117 / Chapter 5 --- Learning Control for Unmanned Helicopters --- p.120 / Chapter 5.1 --- Introduction --- p.121 / Chapter 5.2 --- RelatedWork --- p.124 / Chapter 5.3 --- Preliminary --- p.129 / Chapter 5.3.1 --- Derivation of Policy Gradient --- p.133 / Chapter 5.4 --- Controller Design --- p.133 / Chapter 5.4.1 --- Learning Dynamics --- p.135 / Chapter 5.4.2 --- Learning Control --- p.137 / Chapter 5.5 --- Performance Analysis --- p.141 / Chapter 5.5.1 --- Baseline --- p.141 / Chapter 5.5.2 --- Numerical Analysis --- p.145 / Chapter 5.6 --- Convergence Analysis --- p.152 / Chapter 5.7 --- Evaluation --- p.153 / Chapter 5.7.1 --- Simulation - Acrobot --- p.153 / Chapter 5.7.2 --- Simulation - Fixed-wing Aircraft --- p.154 / Chapter 5.7.3 --- Experiment - RC Car --- p.158 / Chapter 5.7.4 --- Experiment - Helicopter --- p.161 / Chapter 5.8 --- Summary --- p.169 / Chapter 6 --- Conclusion --- p.171 / Chapter A --- List of Publications & Awards --- p.177 / Chapter A.0.1 --- Publications --- p.177 / Chapter A.0.2 --- Awards --- p.181 / Bibliography --- p.182

相對於在地面上行走的機器人, 自動無人飛機擁有高自由度及靈活操控性。在對人類有危險的環境中，無人飛機更特別可以全自動或供操作員安全地遙距控制來執行危險工作。但無線連結不能確保安全及穩定，所以基於安全性考慮自動無人機上必需擁有自主運算能力以用於自動定位，返航及避障。 / 可是制作全自動無人機有很多的限制，如小型無人飛機上只能使用較小型馬達，載重十分少，故能被作為主要的計算機也不能使用強大的圖像處理器(GPU)，所以只能使用運算能力較低的一般處理器(CPU)。此外，傳統的圖像處理定位方法也只能在強大硬件下實時運行，這種種的限制加起來使得小型無人飛機的自動定位成為難題。 / 一般現在的無人機，例如作為航空拍攝的，都是使用簡單而小型的全球定位系統感測器，所需要的運算器硬件要求也不高。但是這些無人飛機卻不能被用於室內或靠近建築物的地方，因為衛星定位信號會被多重反射甚至是阻隔開來。此外，全球定位系統普不能給予無人機有關機體附近環境的訊息，例如視覺特徵及三維結構等。所以無人機並沒有能力避開環境中的障礙物或甚至危害人身安全。 / 我們設計了飛行控制器，內含恣態感測組件，光流傳感器及高度計等元件作控制及感測。我們也提出了一個全自動的校準方法用於低成本的恣態感測器，當中並不需要任何人為的操作或昂貴和精準的的校準儀器。 / 除了恣態感測組件用於機體動作測量外，環境資訊都是對自動飛行有很大用的。我們所建構的無人飛機平台上使用了多種不同的運算組件及視覺定位算法，目的在於提拱一個有效,快速且能於飛機上裝置的電腦實時運算的定位系統。 / 從本論文所展示的結果中可得悉我們在無人機的及展，包括如何把低成本的恣態感測器校正的方法以增加準確性及從環境視覺特徵及三維結構幫助計算無人機自身位置的定位方法。我們希望本研究論文能對下一世代的自動無人機發展有幫助。 / Unmanned Aerial Vehicles (UAVs) process great maneuverability over ground vehicles, legged and tracked robots. UAVs can do more dangerous tasks where human can safely tele-operate from far away. However stable communication channels may not be guaranteed and therefore autonomous onboard processing for safety-concerned functions such as localization, homing or obstacles avoidance are essential. / UAVs have more limitations than others, typically the payload weight is significantly reduced compared to ground vehicles with powerful motors. Other limitations such as limited computing power for light-weight computer modules being put on-board lend the difficulty to the localization and control of the agile UAVs using traditional computationally exhausting Visual SLAM approaches. / Although the traditional simple GPS-Inertial approaches are beautiful that the hardware requirement is not so demanding, they cannot be applied in indoor or near indoor environments due to signal jamming and multiple-reflection. Additionally, GPS cannot provide any information about the environment arround the UAV so it raises risk for the UAV to danger such as obstacles collision which may cause serious injury to human. / To implement an autonomous aerial vehicle we built our flight-controller with a low-cost inertial measurement unit (IMU) and optical ow sensor and altimeter for motor control and sensing. The low-cost IMU suffers from bad measurement bias and drifts. We proposed a fully autonomous calibration method where no human intervention nor precise calibration equipment is needed. / Besides the IMU which can only provide body motion sensing, the perception about the environment is required for autonomous flying and avoid obstacles. We implemented a visual localization algorithm which runs efficiently in realtime on our light-weight and low-cost computer module. We have built a set of UAV system with an RGB-D Kinect sensor, which fuse different sensor measurements together for the localization. / The results presented in this thesis show a new possiblity to advance current development on Unmanned Aerial Vehicles. From the autonomous calibration method for low-cost IMU to improve accuracy orientation estimation to 3D localization using the perception of the environment features and structures from RGB-D sensor. At the same time we hope that our experience on the hardware building and real-time localization algorithm implementations can help boosting the advancement of the next generation of autonomous aerial vehicles. / Cheuk, Chi Ming. / Thesis M.Phil. Chinese University of Hong Kong 2015. / Includes bibliographical references (leaves 88-96). / Abstracts also in Chinese. / Title from PDF title page (viewed on 05, October, 2016). / Cheuk, Chi Ming. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only.

MANETs are known to be useful in situations where mobile nodes need to communicate and coordinate in dynamic environments with no access to fixed network infrastructure. However, connectivity problems can occur when sub-groups within a MANET move out of communication range from one another. The increasingly prolific use of UAVs in military and civilian contexts suggests that UAVs may be very useful for facilitating connectivity between otherwise disconnected mobile nodes. Network Centric Warfare (NCW) theory makes heavy use of MANETs, and UAVs also fit well into the NCW theory; this paper describes the work involved in integrating network enabled UAVs into a previously-developed simulation of ground troop mobility called UMOMM. Specifically, we created a simple decision model for constrained, constant-radius UAV movements, and developed a target-based method by which UAVs can distribute themselves in order to improve the connectivity of the ground members of the MANET. / Graduation date: 2012

Thesis (Ph. D.)--Oregon State University, 2007. / Printout. Includes bibliographical references. Also available on the World Wide Web.