One of the main challenges of the integration of unmanned aerial vehicles (UAVs) into today’s society is the risk of in-flight failures, such as motor failure, occurring in populated areas that can result in catastrophic accidents. We propose a framework to manage the consequences of an in-flight system failure and to bring down the aircraft safely without causing any serious accident to people, property, and the UAV itself. This can be done in three steps: a) Detecting a failure, b) Finding a safe landing spot, and c) Navigating the UAV to the safe landing spot. In this thesis, we will look at part b. Specifically, we are working to develop an active system that can detect landing sites autonomously without any reliance on UAV resources. To detect a safe landing site, we are using a deep learning algorithm named "You Only Look Once" (YOLO) that runs on a Jetson Xavier NX computing module, which is connected to a camera, for image processing. YOLO is trained using the DOTA dataset and we show that it can detect landing spots and obstacles effectively. Then by avoiding the detected objects, we find a safe landing spot. The effectiveness of this algorithm will be shown first by comprehensive simulations. We also plan to experimentally validate this algorithm by flying a UAV and capturing ground images, and then applying the algorithm in real-time to see if it can effectively detect acceptable landing spots.

One of the vital processes for the maintenance of infrastructure is the collection of information about the inventory and current state of the infrastructure. Such activities are mostly done manually by the inspector in the field. However, Unmanned Aerial Vehicles (UAV) offer the possibilities to improving the accuracy, precision, and efficiency of those tasks. The present dissertation focusses on the evaluation of the requirements and possibilities for the incorporation of UAV in the assessment of port infrastructure, with an emphasis on pavement infrastructure. The first step to reach the goal of the research was the elaboration of an extensive literature review where the leading practices and trends for the use of Unmanned Aerial Systems (UAS) were identified. Based on the literature review, it was possible to propose a roadmap for the implementation of the UAS in the assessment of port infrastructure. The roadmap was implemented in a case study for the Norvik port in Stockholm while the restrictions and information allowed. This research produced several key findings. First, it was possible to recognize the lack of precise definitions in the pavement assessment, the faults in the current manual collection of pavement distresses and the voids in an investigation regarding the recognition of pavement defects different than cracking as some of the critical problems in the area. Additionally, the current applications like bridge and structural inspection, and available technologies like LiDAR or visual sensors were identified along with its improvement opportunities and restrictions. The key steps for the implementation of a UAS for assessing infrastructure were formulated as the identification of the needs and critical parameters, the selection of the UAS components, mainly the UAV and sensor, and the postprocessing of the data. The main conclusion drawn from the research is that it is possible to use UAS to assess the state of the infrastructure. However, not all UAS are suitable for all situations or necessities. The selection of the UA, according to the needs and limitations of the project, plays a vital role regarding the viability of implementation of a UAS for monitoring port infrastructure. The sufficiency of a UAS is closely related on its capability to acquire the information of the selected structures, with the required quality, and overcome the limitations, challenges, and restrictions of the site of application. As a way forward, the most important element to address is the implementation of Machine Learning (ML) techniques and Artificial Intelligence (AI) to extract the relevant features of the data automatically.

This study explores the potential benefits, limitations and drawbacks of usingLoRa with unmanned aerial vehicles (UAVs) and autonomous mobile robots (AMRs)together for warehouse management. Three research concerns are looked at in order toprovide a full analysis of the topic. First, the benefits, limitations and drawbacks of utilizingLoRa with UAVs and AMRs are discussed along with suggestions for how to get over thechallenges within warehouse management. Second, potential security risks with the usage ofLoRa in conjunction with UAVs and AMRs for warehouse management are recognized, andmitigation strategies are recommended. The study's last portion considers how warehousemanagement company Proton Finishing AB may utilize LoRa in combination with UAVs andAMRs to promote environmental sustainability.

The senseFly eBeePlus fixed-wing drone is a market available UAV compatible with a range of sensors that includes the Parrot Sequoia+ multispectral camera. Commercial applications of the drone predominantly focus on agriculture, environmental management, and engineering applications. The Sequoia 4-band multispectral sensor with bands optimised for plant health analysis, has a spectral range that coincides with the absorption features of iron. Previous studies with the use of hyperspectral sensors on multicopter UAVs have proven successful in the detection and delineation of hydroxides and sulphates associated with weathering of sulphides at the surface. This study aims to evaluate the ability of the eBeePlus drone equipped with a Parrot Sequoia+ sensor to effectively detect and delineate surficial sulphide mineral expressions by testing its capability on a known nickel-copper mineralisation occurrence at Illukunnguaq, on the north-western coast of Disko Island, West Greenland. Formally hosting a 28-tonne nickeliferous pyrrhotite massive sulphide boulder, many companies have sought this region for a possible extension of the mineralisation or another local mineral occurrence. Iron-feature band ratios and Spectral Angle Mapping (SAM) are two methods tested to first characterise the known occurrence, then search the wider region for other features with a similar signature as the Illukunnguaq dyke. To assist the evaluation and fine tune the Sequoia sensor, it will be compared against the trialled and trusted Rikola hyperspectral sensor, proven to map iron features. In addition, eigen maxima as one of many geomorphological indices that utilise the co-product Digital Surface Model (DSM) of the spectral survey, is employed to assess whether the Illukunnguaq dyke and other features are structurally mappable.  Results show that the Sequoia multispectral sensor, albeit less spectrally resolved than the Rikola hyperspectral sensor was able to detect surficial sulphide mineral expressions both by applying iron-feature band ratios and SAM. The latter was performed using laboratory measured and open-access library spectra. To fine-tune the tools compatible with the Sequoia sensor, in-depth investigations into iron-feature band ratio index values and best-fit library spectra for SAM was conducted. Confidence was increased by the blind detection of another known exposure and permitted a regional search to find additional features with spectral similarities to the Illukunnguaq dyke for future ground truthing. This study demonstrates that the eBeePlus drone can be used for mineral exploration when iron-sulphides are a part of the mineral system and outcropping at the surface. Leading field programs with detailed multispectral mapping can improve the efficiency of geologists by generating or verifying targets prior to ‘boots-on-the-ground’ geological sampling or mapping.

This diploma thesis solves the analysis of the influence of calibration and the method of strips alignment on the geometric accuracy of a point cloud acquired by UAV lidar scanning. The aim was to find out the influence of individual used methods, respectively various combinations. The effect of the design of the cross-flights has also been added. The evaluation was performed using standard deviations of the distances corresponding to the areas scanned in different point bands. Furthermore, verification was performed by comparing checkpoints. The results show that there is no dependence between the individual combinations. The only case was a larger displacement of the point cloud at the edge of the scanned strip in the case of cross-flights.

This work demonstrates the use of genetic algorithms as a stochastic optimization technique for developing a camera network design and the flight path for photogrammetricapplications using Small Unmanned Aerial Vehicles. This study develops a Virtual Optimizer for Aerial Routes (VOAR) as a new photogrammetric mapping tool for acquisition of images to be used in 3D reconstruction. 3D point cloud models provide detailed information on infrastructure from places where human access may be difficult. This algorithm allows optimized flight paths to monitor infrastructure using GPS coordinates and optimized camera poses ensuring that the set of images captured is improved for 3D point cloud development. Combining optimization techniques, autonomous aircraft and computer vision methods is a new contribution that this work provides.This optimization framework is demonstrated in a real example that includes retrieving the coordinates of the analyzed area and generating autopilot coordinates to operate in fully autonomous mode. These results and their implications are discussed for future work and directions in making optical techniques competitive with aerial or ground based LiDAR systems.

Opportunities to use synthetic aperture radar (SAR) in scientific studies and military operations are expanding with the development of small SAR systems that can be operated on small unmanned air vehicles (UAV)s. While the nimble nature of small UAVs make them an attractive platform for many reasons, small UAVs are also more prone to deviate from a linear course due autopilot errors and external forces such as turbulence and wind. Thus, motion compensation and improved processing algorithms are required to properly focus the SAR images. The work of this dissertation overcomes some of the challenges and addresses some of the opportunities of operating SAR on small UAVs. Several contributions to SAR backprojection processing for UAV SARs are developed including: 1. The derivation of a novel SAR backprojection algorithm that accounts for motion during the pulse that is appropriate for narrow or ultra-wide-band SAR. 2. A compensation method for SAR backprojection to enable radiometrically accurate image processing. 3. The design and implementation of a real-time backprojection processor on a commercially available GPU that takes advantage of the GPU texture cache. 4. A new autofocus method that improves the image focus by estimating motion measurement errors in three dimensions, correcting for both amplitude and phase errors caused by inaccurate motion parameters. 5. A generalization of factorized backprojection, which we call the Dually Factorized Backprojection method, that factorizes the correlation integral in both slow-time and fast-time in order to efficiently account for general motion during the transmit of an LFM-CW pulse. Much of this work was conducted in support of the Characterization of Arctic Sea Ice Experiment (CASIE), and the appendices provide substantial contributions for this project as well, including: 1. My work in designing and implementing the digital receiver and controller board for the microASAR which was used for CASIE. 2. A description of how the GPU backprojection was used to improved the CASIE imagery. 3. A description of a sample SAR data set from CASIE provided to the public to promote further SAR research.

Landing an Unmanned Aerial Vehicle (UAV) is a non-trivial problem. Removing the ability to cooperate with the landing site further increases the complexity. This thesis develops a multi-stage process that allows a UAV to locate the safest landing site, and then land without a georeference. Machine vision is the vehicle sensor used to locate potential landing hazards and generate an estimated UAV position. A description of the algorithms, along with validation results, are presented. The thesis shows that software-simulated landing performs adequately, and that future hardware integration looks promising.

The transition from a mathematical algorithm to a physical hardware implementation is non-trivial. This thesis discusses the issues involved in the transition from the theory of real-time trajectory generation all the way through a hardware experiment. Documentation of the validation process as well as modifications to the existing theory as a result of hardware testing are treated at length. The results of hardware experimentation show that trajectory generation can be done in real-time in a manner facilitating coordination of multiple small UAVs.

This thesis addresses the improvement of cooperative task allocation to vehicles in multiple-vehicle, multiple-target scenarios through the use of more effective preplanned tours. Effective allocation of vehicles to targets requires knowledge of both the team objectives and the contributions that individual vehicles can make toward accomplishing team goals. This is primarily an issue of individual vehicle path planning --- determining the path the vehicles will follow to accomplish individual and team goals. Conventional methods plan optimal point-to-point path segments that often result in lengthy and suboptimal tours because the trajectory neither considers future tasks nor the overall path. However, cooperation between agents is improved when the team selects vehicle assignments based on the ability to complete immediate and subsequent tasks. This research demonstrates that planning more efficient tour paths through multiple targets results in better use of individual vehicle resources, faster completion of team objectives, and improved overall cooperation between agents. This research presents a method of assigning unmanned aerial vehicles to targets to improve cooperation. A tour path planning method was developed to overcome shortcomings of traditional point-to-point path planners, and is extended to the specific tour-planning needs of this problem. The planner utilizes an on-line learning heuristic search to find paths that accomplish team goals in the shortest flight time. The learning search planner uses the entire sensor footprint, more efficiently plans tours through closely packed targets, and learns the best order for completion of the multiple tasks. The improved planner results in assignment completion times that range on average between 1.67 and 2.41 times faster, depending on target spread. Assignments created from preplanned tours make better use of vehicle resources and improve team cooperation. Path planning and assignment selection are accomplished in near real-time through the use of path heuristics and assignment cost estimates to reduce the problem size to tractable levels. Assignments are ordered according to estimated or predicted value. A reduced number of ordered assignments is considered and evaluated to control problem growth. The estimates adequately represent the actual assignment value, effectively reduce problem size, and produce near-optimal paths and assignments for near-real-time applications.