Motion planning is the field of computer science that aims at developing algorithmic techniques allowing the automatic computation of trajecto- ries for a mechanical system. The nature of such a system vary according to the fields of application. In computer animation it could be a humanoid avatar. In molecular biology it could be a protein. The field of application of this work being aerial robotics, the system is here a four-rotor UAV (Unmanned Aerial Vehicle) called quadrotor. The motion planning problem consists in computing a series of motions that brings the system from a given initial configuration to a desired final configuration without generating collisions with its environment, most of the time known in advance. Usual methods explore the system’s configuration space regardless of its dynamics. By construction the thrust force that allows a quadrotor to fly is tangential to its attitude which implies that not every motion can be performed. Furthermore, the magnitude of this thrust force and hence the linear acceleration of the center of mass are limited by the physical capabilities of the robot. For all these reasons, not only position and orientation must be planned, higher derivatives must be planned also if the motion is to be executed. When this is the case we talk of kinodynamic motion planning. A distinction is made between the local planner and the global planner. The former is in charge of producing a valid trajectory between two states of the system without necessarily taking collisions into account. The later is the overall algorithmic process that is in charge of solving the motion planning problem by exploring the state space of the system. It relies on multiple calls to the local planner. We present a local planner that interpolates two states consisting of an arbitrary number of degrees of freedom (dof) and their first and second derivatives. Given a set of bounds on the dof derivatives up to the fourth order (snap), it quickly produces a near-optimal minimum time trajectory that respects those bounds. In most of modern global motion planning algorithms, the exploration is guided by a distance function (or metric). The best choice is the cost-to-go, i.e. the cost associated to the local method. In the context of kinodynamic motion planning, it is the duration of the minimal-time trajectory. The problem in this case is that computing the cost-to-go is as hard (and thus as costly) as computing the optimal trajectory itself. We present a metric that is a good approximation of the cost-to-go but which computation is far less time consuming. The dominant paradigm nowadays is sampling-based motion planning. This class of algorithms relies on random sampling of the state space in order to quickly explore it. A common strategy is uniform sampling. It however appears that, in our context, it is a rather poor choice. Indeed, a great majority of uniformly sampled states cannot be interpolated. We present an incremental sampling strategy that significantly decreases the probability of this happening.

Visual inspection in autonomous robotics is a task in which autonomous agents are required to gather visual information of objects of interest, in a manner that ensures safety, efficiency and comprehensive coverage. It is, therefore, crucial for identifying key landmarks, detecting cracks or defects, or reconstructing the observed object for detailed analysis. This thesis delves into the  challenges encountered by autonomous agents in executing such tasks and presents frameworks for scenarios ranging from operations by multiple spacecrafts in close proximity to celestial bodies in Deep Space to terrestrial deployments of Unmanned Aerial Vehicles (UAVs) for inspection of large-scale infrastructures. The research thus pursues two main directions: Firstly, a novel formation control strategy is developed to enable autonomous agents to perform proximity operations safely, efficiently, and accurately in order to map the surface of Small Celestial Bodies (SCBs). This investigation encompasses control and coordination strategies, leveraging a realistic astrodynamic model of the orbital environment to navigate safely around SCBs. Along this direction, the contributions focus on enabling a distributed autonomy framework in the form of a cooperative stereo configuration between two spacecraft, allowing acquisition of 3D topological information of the candidate SCB. The framework employs a Leader-Follower approach, treating the maintenance of the desired stereo-formation as a 6 Degree-of-Freedom (DoF) nonlinear model predictive control (NMPC) problem. The second research direction focuses on addressing the problem of enabling robotic inspection for terrestrial applications. With the growing demand for efficient and reliable inspection techniques to improve in-situ situational awareness, the research concentrates on addressing the problem of obtaining detailed visual scan of available structures without any a priori knowledge of either the environment nor the structures. Thus, the key contributions of the presented work reside in the implementation of a unified autonomy, with the unification drawing it's root from the merging of two distinct research perspectives: Inspection and Exploration planning. The contribution establishes a novel solution by introducing a map-independent approach with a synergistic formulation of a reactive profile-adaptive view-planner coupled with a hierarchical exploration strategy and an environment-invariant scene recognition module. By integrating exploration and inspection methodologies, this research seeks to enhance the capabilities of UAVs in navigating and inspecting unknown structures in unfamiliar environments.  Through theoretical developments, extensive simulations and experimental validations, this thesis contributes to the advancement of the state-of-the-art in visual inspection with autonomous robots. Moreover, the findings extend current capabilities of autonomous agents in the field of space exploration as well as in disaster response and complex infrastructure inspection.

This thesis presents a novel architecture for human-in-the-loop control of multiple drones. The design of such systems must address several challenges at the same time. The drones must avoid collisions with each other and with obstacles in their task environment while following operator's command as closely as possible to navigate their environment. To this end, they should be able to adjust their pre-defined desired formation and, if needed, transition to alternative formations to ensure collision-free operation in their task environment while following the operator's commands. The proposed control strategy is a central algorithm with multiple stages and relies on formulating and solving convex optimization problems in real time to achieve the control objectives. The operator provides reference velocity commands for the flock of drones to move them in the task environment. The algorithm creates linear collision avoidance constraints and distributes the operator's commands among the drones through a number of intermediate steps. It generates reference trajectories for the drones motion by solving a model-based optimization problem over a receding horizon. Conventional trajectory controllers generate the control inputs for individual drones. Prospective formation shapes are obtained for the drones by formulating and solving parallel convex optimizations, considering the operator's reference command and the obstacle-free space. While keeping the convexity of the optimization problem, the proposed algorithm allows for the presence of obstacles in the middle of the formation. This is achieved by properly assigning obstacle-free regions to each agent separately in the formation. In addition, safe convex regions in the form of linear inequality constraints are generated in the direction of the operator's commanded velocity. Moreover, constraints are introduced to avoid inter-drone collisions at each step. Trajectory optimization is formulated as a quadratic programming problem similar to model predictive control schemes to minimize deviation from human operator's command. The effectiveness of the proposed control algorithm is initially verified by simulating two different operational scenarios. Furthermore, the algorithm is implemented on actual hardware to operate a flock of three drones in a laboratory setting. The implementation of the algorithm in C++ utilizes high-performance computation techniques to achieve sufficiently high real-time control update rates for smooth and stable operation of the drones. / Thesis / Master of Applied Science (MASc) / The rise of unmanned aerial vehicle technology and the increase in their accessibility have made them viable solutions for serious missions such as search and rescue operations. Complex cooperative tasks can be conducted via a collection of drones which can show higher levels of robustness and agility as a system. Although repetitive and simple actions can be easily automated, real-world problems are unpredictable in which complex decision-making is involved. Such scenarios can be tackled by the presence of a human supervisor to empower the system with strong cognitive capabilities. This thesis presents a multi-layer control framework for human-in-the-loop operation of a flock of unmanned aerial vehicles. This method continuously optimizes the drones trajectories to adhere as closely as possible to operator's motion commands while avoiding collisions among them and with obstacles in their task environment. This new control framework is successfully validated in both simulations and experiments in a laboratory environment.

This dissertation focuses on the constrained control of multi-body systems which are actuated by vectorized thrusters. A general control framework is proposed to stabilize the task configuration while ensuring constraints satisfaction at all times. For this purpose, the equations of motion of the system are derived using the Euler-Lagrange method. It is seen that under some reasonable conditions, the system dynamics are decoupled. This property is exploited in a cascade control scheme to stabilize the points of equilibrium of the system. The control scheme is composed of an inner loop, tasked to control the attitude of the vectorized thrusters, and an outer loop which is tasked to stabilize the task configuration of the system to a desired configuration. To prove stability, input-to-state stability and small gain arguments are used. All stability properties are derived in the absence of constraints, and are shown to be local. The main result of this analysis is that the proposed control scheme can be directly applied under the assumption that a suitable mapping between the generalized force and the real inputs of the system is designed. This thesis proposes to enforce constraints by augmenting the control scheme with two types of Reference Governor units: the Scalar Reference Governor, and the Explicit Reference Governor. This dissertation presents two case studies which inspired the main generalization of this thesis: (i) the control of an unmanned aerial and ground vehicle manipulating an object, and (ii) the control of a tethered quadrotor. Two further case studies are discussed afterwards to show that the generalized control framework can be directly applied when a suitable mapping is designed. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

In this work, research to develop algorithms for a package exchange maneuver between two quad-rotor aircraft is presented. First, the development of tools used for this research is discussed. Second, a controller is designed that synchronizes the flight paths and motion of two quad-rotor robots. The controller is used to guide a designated follower quad-rotor to follow a leader aircraft’s position and orientation. The follower aircraft is equipped with a simple mechanical manipulator to compensate for limitations in the aircrafts maneuverability. finally, a sensor architecture study for relative navigation of Unmanned Aerial Vehicles (UAV) is presented. The architecture study presents typical navigation solutions, considers each solution’s appropriateness for close-proximity missions, and compares performance.

La planification de mouvement est le domaine de l’informatique qui a trait au développement de techniques algorithmiques permettant la génération automatique de trajectoires pour un système mécanique. La nature d’un tel système varie selon les champs d’application. En animation par ordinateur il peut s’agir d’un avatar humanoïde. En biologie moléculaire cela peut être une protéine. Le domaine d’application de ces travaux étant la robotique aérienne, le système est ici un UAV (Unmanned Aerial Vehicle: véhicule aérien sans pilote) à quatre hélices appelé quadrirotor. Le problème de planification de mouvements consiste à calculer une série de mouvements qui amène le système d’une configuration initiale donnée à une configuration finale souhaitée sans générer de collisions avec son environnement, la plupart du temps connu à l’avance. Les méthodes habituelles explorent l’espace des configurations du système sans tenir compte de sa dynamique. Cependant, la force de poussée qui permet à un quadrirotor de voler est par construction parallèle aux axes de rotation des hélices, ce qui implique que certains mouvements ne peuvent pas être effectués. De plus, l’intensité de cette force de poussée, et donc l’accélération linéaire du centre de masse, sont limitées par les capacités physiques du robot. Pour toutes ces raisons, non seulement la position et l’orientation doivent être planifiées, mais les dérivées plus élevées doivent l’être également si l’on veut que le système physique soit en mesure de réellement exécuter le mouvement. Lorsque c’est le cas, on parle de planification kinodynamique de mouvements. Une distinction est faite entre le planificateur local et le planificateur global. Le premier est chargé de produire une trajectoire valide entre deux états du système sans nécessairement tenir compte des collisions. Le second est l’algorithme principal qui est chargé de résoudre le problème de planification de mouvement en explorant l’espace d’état du système. Il fait appel au planificateur local. Nous présentons un planificateur local qui interpole deux états comprenant un nombre arbitraire de degrés de liberté ainsi que leurs dérivées premières et secondes. Compte tenu d’un ensemble de limites sur les dérivées des degrés de liberté jusqu’au quatrième ordre (snap), il produit rapidement une trajectoire en temps minimal quasi optimale qui respecte ces limites. Dans la plupart des algorithmes modernes de planification de mouvements, l’exploration est guidée par une fonction de distance (ou métrique). Le meilleur choix pour celle-ci est le cost-to-go, c.a.d. le coût associé à la méthode locale. Dans le contexte de la planification kinodynamique de mouvements, il correspond à la durée de la trajectoire en temps minimal. Le problème dans ce cas est que calculer le cost-to-go est aussi difficile (et donc aussi coûteux) que de calculer la trajectoire optimale elle-même. Nous présentons une métrique qui est une bonne approximation du cost-to-go, mais dont le calcul est beaucoup moins coûteux. Le paradigme dominant en planification de mouvements aujourd’hui est l’échantillonnage aléatoire. Cette classe d’algorithmes repose sur un échantillonnage aléatoire de l’espace d’état afin de l’explorer rapidement. Une stratégie commune est l’échantillonnage uniforme. Il semble toutefois que, dans notre contexte, ce soit un choix assez médiocre. En effet, une grande majorité des états uniformément échantillonnés ne peuvent pas être interpolés. Nous présentons une stratégie d’échantillonnage incrémentale qui diminue considérablement la probabilité que cela ne se produise. / Motion planning is the field of computer science that aims at developing algorithmic techniques allowing the automatic computation of trajecto- ries for a mechanical system. The nature of such a system vary according to the fields of application. In computer animation it could be a humanoid avatar. In molecular biology it could be a protein. The field of application of this work being aerial robotics, the system is here a four-rotor UAV (Unmanned Aerial Vehicle) called quadrotor. The motion planning problem consists in computing a series of motions that brings the system from a given initial configuration to a desired final configuration without generating collisions with its environment, most of the time known in advance. Usual methods explore the system’s configuration space regardless of its dynamics. By construction the thrust force that allows a quadrotor to fly is tangential to its attitude which implies that not every motion can be performed. Furthermore, the magnitude of this thrust force and hence the linear acceleration of the center of mass are limited by the physical capabilities of the robot. For all these reasons, not only position and orientation must be planned, higher derivatives must be planned also if the motion is to be executed. When this is the case we talk of kinodynamic motion planning. A distinction is made between the local planner and the global planner. The former is in charge of producing a valid trajectory between two states of the system without necessarily taking collisions into account. The later is the overall algorithmic process that is in charge of solving the motion planning problem by exploring the state space of the system. It relies on multiple calls to the local planner. We present a local planner that interpolates two states consisting of an arbitrary number of degrees of freedom (dof) and their first and second derivatives. Given a set of bounds on the dof derivatives up to the fourth order (snap), it quickly produces a near-optimal minimum time trajectory that respects those bounds. In most of modern global motion planning algorithms, the exploration is guided by a distance function (or metric). The best choice is the cost-to-go, i.e. the cost associated to the local method. In the context of kinodynamic motion planning, it is the duration of the minimal-time trajectory. The problem in this case is that computing the cost-to-go is as hard (and thus as costly) as computing the optimal trajectory itself. We present a metric that is a good approximation of the cost-to-go but which computation is far less time consuming. The dominant paradigm nowadays is sampling-based motion planning. This class of algorithms relies on random sampling of the state space in order to quickly explore it. A common strategy is uniform sampling. It however appears that, in our context, it is a rather poor choice. Indeed, a great majority of uniformly sampled states cannot be interpolated. We present an incremental sampling strategy that significantly decreases the probability of this happening.

More than 70% of Earth's surface is covered by water. Earth's underwater world holds many exciting forms of life and undiscovered possibilities. It is sometimes referred to as "The Unexplored Frontier." We still do not fully understand the entirety of what happens in this mysterious world. The field of underwater sensor networks is a means of monitoring these environments. However, underwater sensor networks are still fraught with challenges; one of the main challenges being communication. In this thesis we look to improve communication in underwater sensor networks. We expand a simulation environment that models node to node communication in an underwater sensor network that utilizes AquaNodes. We address issues with the first iteration of the environment, expand it to include packet-loss for acoustic communication, and make the addition of three dimensional topologies. We found that acoustic packet-loss had a larger impact on the energy consumption of the communication algorithms with more acoustic communication and three dimensional topologies do not affect the communication algorithms. In addition to expanding the simulation environment we also explore using UAVs as a means of extracting data out of underwater sensor network. We conduct field experiments to characterize radio communication, develop an energy model to understand the energy limitations of an UAV, and develop overall policies for using an UAV with an underwater sensor network that utilizes AquaNodes. We learned that node to node radio communication range on the surface of the water had shorter ranges than on land. We also learned that node to UAV communication range was dependant on the altitude of the UAV. Overall, we found that using an UAV as a data mule was a viable method of extracting data out of certain underwater sensor network configurations.

Le but de cette thèse est d'étendre l'état de l'art par des contributions sur le comportement collectif d'un groupe de robots volants, à savoir des quadrirotors UAV. Afin de pouvoir sûrement naviguer dans un environnement, ces derniers peuvent se reposer uniquement sur leurs capacités à bord et non sur des systèmes centralisés (e.g., Vicon ou GPS). Nous réalisons cet objectif en offrant une possible solution aux problèmes de contrôle en formation et de localisation à partir de mesures à bord et via une communication locale. Nous abordons ces problèmes exploitant différents concepts provenant de la théorie des graphes algébriques et de la théorie de la rigidité. Cela nous permet de résoudre ces problèmes de façon décentralisée et de proposer des algorithmes décentralisés capables de prendre en compte également des limites sensorielles classiques. Les capacités embarquées que nous avons mentionnées plus tôt sont représentées par une caméra monoculaire et une centrale inertielle (IMU) auxquelles s'ajoute la capacité de chaque robot à communiquer (par RF) avec certains de ses voisins. Cela est dû au fait que l'IMU et la caméra représentent une possible configuration économique et légère pour la navigation et la localisation autonome d'un quadrirotor UAV. / The aim of this Thesis is to give contributions to the state of the art on the collective behavior of a group of flying robots, specifically quadrotor UAVs, which can only rely on their onboard capabilities and not on a centralized system (e.g., Vicon or GPS) in order to safely navigate in the environment. We achieve this goal by giving a possible solution to the problems of formation control and localization from onboard sensing and local communication. We tackle these problems exploiting mainly concepts from algebraic graph theory and the so-called theory of rigidity. This allows us to solve these problems in a decentralized fashion, and propose decentralized algorithms able to also take into account some typical sensory limitations. The onboard capabilities we referred to above are represented by an onboard monocular camera and an inertial measurement unit (IMU) in addition to the capability of each robot to communicate (through RF) with some of its neighbors. This is due to the fact that an IMU and a camera represent a possible minimal, lightweight and inexpensive configuration for the autonomous localization and navigation of a quadrotor UAV.

The main characteristic attribute of the rotorcraft is the use of rotary wings to produce the thrust force necessary for motion. Therefore, rotorcraft have an advantage relative to fixed wing aircraft because they do not require any relative velocity to produce aerodynamic forces. Rotorcraft have been used in a wide range of missions of civilian and military applications. Particular interest has been concentrated in applications related to search and rescue in environments that impose restrictions to human presence and interference. The main representative of the rotorcraft family is the helicopter. Small scale helicopters retain all the flight characteristics and physical principles of their full scale counterpart. In addition, they are naturally more agile and dexterous compared to full scale helicopters. Their flight capabilities, reduced size and cost have monopolized the attention of the Unmanned Aerial Vehicles research community for the development of low cost and efficient autonomous flight platforms. Helicopters are highly nonlinear systems with significant dynamic coupling. In general, they are considered to be much more unstable than fixed wing aircraft and constant control must be sustained at all times. The goal of this dissertation is to investigate the challenging design problem of autonomous flight controllers for small scale helicopters. A typical flight control system is composed of a mathematical algorithm that produces the appropriate command signals required to perform autonomous flight. Modern control techniques are model based, since the controller architecture depends on the dynamic description of the system to be controlled. This principle applies to the helicopter as well, therefore, the flight control problem is tightly connected with the helicopter modeling. The helicopter dynamics can be represented by both linear and nonlinear models of ordinary differential equations. Theoretically, the validity of the linear models is restricted in a certain region around a specific operating point. Contrary, nonlinear models provide a global description of the helicopter dynamics. This work proposes several detailed control designs based on both dynamic representations of small scale helicopters. The controller objective is for the helicopter to autonomously track predefined position (or velocity) and heading reference trajectories. The controllers performance is evaluated using X-Plane, a realistic and commercially available flight simulator.

A forced landing for an Unmanned Aerial Vehicle (UAV) is required if there is an emergency on board that requires the aircraft to land immediately. Piloted aircraft in the same scenario have a human on board that is able to engage in the complex decision making process involved in the choice of a suitable landing location. If UAVs are to ever fly routinely in civilian airspace, then it is argued that the problem of finding a safe landing location for a forced landing is an important unresolved problem that must be addressed. This thesis presents the results of an investigation into the feasibility of using machine vision techniques to locate candidate landing sites for an autonomous UAV forced landing. The approach taken involves the segmentation of the image into areas that are large enough and free of obstacles; classification of the surface types of these areas; incorporating slope information from readily available digital terrain databases; and finally fusing these maps together using a high level set of simple linguistic fuzzy rules to create a final candidate landing site map. All techniques were evaluated on actual flight data collected from a Cessna 172 flying in South East Queensland. It was shown that the use of existing segmentation approaches from the literature did not provide the outputs required for this problem in the airborne images encountered in the gathered dataset. A simple method was then developed and tested that provided suitably sized landing areas that were free of obstacles and large enough to land. The advantage of this novel approach was that these areas could be extracted from the image directly without solving the difficult task of segmenting the entire image into the individual homogenous objects. A number of neural network classification approaches were tested with the surface types of candidate landing site regions extracted from the aerial images. A number of novel techniques were developed through experimentation with the classifiers that greatly improved upon the classification accuracy of the standard approaches considered. These novel techniques included: automatic generation of suitable output subclasses based on generic output classes of the classifier; an optimisation process for generating the best set of input features for the classifier based on an automated analysis of the feature space; the use of a multi-stage classification approach; and the generation of confidence measures based on the outputs of the neural network classifiers. The final classification result of the system performs significantly better than a human test pilot's classification interpretation of the dataset samples. In summary, the algorithms were able to locate candidate landing site areas that were free of obstacles 92.3 ±2.6% (99% confidence in the result) of the time, with free obstacle candidate landing site areas that were large enough to land in missed only 5.3 ±2.2% (99% confidence in the result) of the time. The neural network classification networks developed were able to classify the surface type of the candidate landing site areas to an accuracy of 93.9 ±3.7% (99% confidence in the result) for areas labelled as Very Certain. The overall surface type classification accuracy for the system (includes all candidate landing sites) was 91.95 ±4.2% (99% confidence in the result). These results were considered to be an excellent result as a human test pilot subject was only able to classify the same data set to an accuracy of 77.24 %. The thesis concludes that the techniques developed showed considerable promise and could be used immediately to enhance the safety of UAV operations. Recommendations include the testing of algorithms over a wider range of datasets and improvements to the surface type classification approach that incorporates contextual information in the image to further improve the classification accuracy.