Search for: All records

In this thesis, I present a decentralized sparse Gaussian process regression (DSGPR) model with event-triggered, adaptive inducing points. This DSGPR model brings the advantages of sparse Gaussian process regression to a decentralized implementation. Being decentralized and sparse provides advantages that are ideal for multi-agent systems (MASs) performing environmental modeling. In this case, MASs need to model large amounts of information while having potential intermittent communication connections. Additionally, the model needs to correctly perform uncertainty propagation between autonomous agents and ensure high accuracy on the prediction. For the model to meet these requirements, a bounded and efficient real-time sparse Gaussian process regression (SGPR) model is needed. I improve real-time SGPR models in these regards by introducing an adaptation of the mean shift and fixed-width clustering algorithms called radial clustering. Radial clustering enables real-time SGPR models to have an adaptive number of inducing points through an efficient inducing point selection process. I show how this clustering approach scales better than other seminal Gaussian process regression (GPR) and SGPR models for real-time purposes while attaining similar prediction accuracy and uncertainty reduction performance. Furthermore, this thesis addresses common issues inherent in decentralized frameworks such as high computation costs, inter-agent message bandwidth restrictions, and data fusion integrity. These challenges are addressed in part through performing maximum consensus between local agent models which enables the MAS to gain the advantages of decentralization while keeping data fusion integrity. The inter-agent communication restrictions are addressed through the contribution of two message passing heuristics called the covariance reduction heuristic and the Bhattacharyya distance heuristic. These heuristics enable user to reduce message passing frequency and message size through the Bhattacharyya distance and properties of spatial kernels. The entire DSGPR framework is evaluated on multiple simulated random vector fields. The results show that this framework effectively estimates vector fields using multiple autonomous agents. This vector field is assumed to be a wind field; however, this framework may be applied to the estimation of other scalar or vector fields (e.g., fluids, magnetic fields, electricity, etc.). Keywords: Sparse Gaussian process regression, clustering, event-triggered, decentralized, sensor fusion, uncertainty propagation, inducing points 

View Video Presentation: https://doi.org/10.2514/6.2022-3218.vid The ability to accurately and rapidly assess unsteady interactional aerodynamics is a shortcoming and bottleneck in the design of various next-generation aerospace systems: from electric vertical takeoff and landing (eVTOL) aircraft to airborne wind energy (AWE) and wind farms. In this study, we present a meshless CFD framework based on the reformulated vortex particle method (rVPM) for the analysis of complex interactional aerodynamics. The rVPM is a large eddy simulation (LES) solving the Navier-Stokes equations in their vorticity form. It uses a meshless Lagrangian scheme, which not only avoids the hurdles of mesh generation, but it also conserves the vortical structure of wakes over long distances with minimal numerical dissipation, while being 100x faster than conventional mesh-based LES. Wings and rotating blades are introduced in the computational domain through actuator line and actuator surface models. Simulations are coupled with an aeroacoustics solver to predict tonal and broadband noise radiated by rotors. The framework, called FLOWUnsteady, is hereby released as an open-source code and extensively validated. Validation studies published in previous work by the authors are summarized, showcasing rotors across operating conditions with a rotor in hover, propellers, a wind turbine, and two side-by-side rotors in hover. Validation of rotor-wing interactions is presented simulating a tailplane with tip-mounted propellers and a blown wing with propellers mounted mid-span. The capabilities of the framework are showcased through the simulation of a tiltwing eVTOL vehicle and an AWE wind-harvesting aircraft, featuring rotors with variable RPM, variable pitch, tilting of wings and rotors, non-trivial flight paths, and complex aerodynamic interactions. 

Winged eVTOL aircraft’s ability to generate aerodynamic lift with wings and to create upward thrust with upward-facing rotors makes these vehicles capable of the kind of versatile flight needed in urban environments. Because of these vehicles’ aerodynamic complexities and their unique methods of producing thrusts and torques, control allocation is needed to determine how to distribute force and torque efforts across the aircraft’s actuators. However, current control allocation methods fail to properly represent the actuators’ complex dynamics and are unable to harness the full potential of these over-actuated vehicles. Current shortcomings include modeling rotors as linear effectors while the wide range of airspeeds experienced by eVTOL aircraft leads to significant nonlinearities in the thrust and torque achieved by each rotor. This means linear control allocation methods may consistently fail to produce desired thrusts and torques, which can inhibit the vehicle from tracking a trajectory at best, and at worst can cause the vehicle to stall and lose control. Additionally, current control allocation methods are often unable to prioritize low-energy actuators resulting in shorter battery life. We present a nonlinear control allocation method that considers a nonlinear rotor model, allows for prioritization of low-energy control surfaces over rotors, and reliably accounts for actuator saturation. Simulation results show a 90% reduction in high-airspeed trajectory tracking position error from a typical, linear least-squares pseudoinverse control allocation method while maintaining comparable energy use. 

Unmanned aerial vehicles (UAV) enable detailed historical preservation of large-scale infrastructure and contribute to cultural heritage preservation, improved maintenance, public relations, and development planning. Aerial and terrestrial photo data coupled with high accuracy GPS create hyper-realistic mesh and texture models, high resolution point clouds, orthophotos, and digital elevation models (DEMs) that preserve a snapshot of history. A case study is presented of the development of a hyper-realistic 3D model that spans the complex 1.7 km2 area of the Brigham Young University campus in Provo, Utah, USA and includes over 75 significant structures. The model leverages photos obtained during the historic COVID-19 pandemic during a mandatory and rare campus closure and details a large scale modeling workflow and best practice data acquisition and processing techniques. The model utilizes 80,384 images and high accuracy GPS surveying points to create a 1.65 trillion-pixel textured structure-from-motion (SfM) model with an average ground sampling distance (GSD) near structures of 0.5 cm and maximum of 4 cm. Separate model segments (31) taken from data gathered between April and August 2020 are combined into one cohesive final model with an average absolute error of 3.3 cm and a full model absolute error of <1 cm (relative accuracies from 0.25 cm to 1.03 cm). Optimized and automated UAV techniques complement the data acquisition of the large-scale model, and opportunities are explored to archive as-is building and campus information to enable historical building preservation, facility maintenance, campus planning, public outreach, 3D-printed miniatures, and the possibility of education through virtual reality (VR) and augmented reality (AR) tours. 

Unmanned aerial systems (UAS) are effective forsurveillance and monitoring, but struggle with persistent, long-term tracking due to limited flight time. Persistent trackingcan be accomplished using multiple vehicles if one vehiclecan effectively hand off the tracking information to anotherreplacement vehicle. In this paper we propose a solution tothe moving-target handoff problem in the absence of GPS. Theproposed solution uses a nonlinear complimentary filter forself-pose estimation using only an IMU, a particle filter forrelative pose estimation between UAS using a relative rangemeasurement, visual target tracking using a gimballed camerawhen the target is close to the handoff UAS, and track correlationlogic using Procrustes analysis to perform the final target handoffbetween vehicles. We present extensive simulation results thatdemonstrates the effectiveness of our approach and performMonte-Carlo simulations that indicate a 97% successful handoffrate using the proposed methods. 

Remote sensing with unmanned aerial vehicles (UAVs) facilitates photogrammetry for environmental and infrastructural monitoring. Models are created with less computational cost by reducing the number of photos required. Optimal camera locations for reducing the number of photos needed for structure-from-motion (SfM) are determined through eight mathematical set-covering algorithms as constrained by solve time. The algorithms examined are: traditional greedy, reverse greedy, carousel greedy (CG), linear programming, particle swarm optimization, simulated annealing, genetic, and ant colony optimization. Coverage and solve time are investigated for these algorithms. CG is the best method for choosing optimal camera locations as it balances number of photos required and time required to calculate camera positions as shown through an analysis similar to a Pareto Front. CG obtains a statistically significant 3.2 fewer cameras per modeled area than base greedy algorithm while requiring just one additional order of magnitude of solve time. For comparison, linear programming is capable of fewer cameras than base greedy but takes at least three orders of magnitude longer to solve. A grid independence study serves as a sensitivity analysis of the CG algorithms α (iteration number) and β (percentage to be recalculated) parameters that adjust traditional greedy heuristics, and a case study at the Rock Canyon collection dike in Provo, UT, USA, compares the results of all eight algorithms and the uniqueness (in terms of percentage comparisons based on location/angle metadata and qualitative visual comparison) of each selected set. Though this specific study uses SfM, the principles could apply to other instruments such as multi-spectral cameras or aerial LiDAR. 

Unsupervised machine learning algorithms (clustering, genetic, and principal component analysis) automate Unmanned Aerial Vehicle (UAV) missions as well as the creation and refinement of iterative 3D photogrammetric models with a next best view (NBV) approach. The novel approach uses Structure-from-Motion (SfM) to achieve convergence to a specified orthomosaic resolution by identifying edges in the point cloud and planning cameras that “view” the holes identified by edges without requiring an initial model. This iterative UAV photogrammetric method successfully runs in various Microsoft AirSim environments. Simulated ground sampling distance (GSD) of models reaches as low as 3.4 cm per pixel, and generally, successive iterations improve resolution. Besides analogous application in simulated environments, a field study of a retired municipal water tank illustrates the practical application and advantages of automated UAV iterative inspection of infrastructure using 63 % fewer photographs than a comparable manual flight with analogous density point clouds obtaining a GSD of less than 3 cm per pixel. Each iteration qualitatively increases resolution according to a logarithmic regression, reduces holes in models, and adds details to model edges.