Variational Autoencoders (VAEs) generally learn latent representations without direct oversight regarding the structure of the latent space, which can hinder their ability to differentiate meaningful, data-driven variations from those residing outside the data manifold. In this paper, we propose an enhanced VAE framework that enforces a systematic decomposition of the latent space into two distinct components: a manifold subspace that captures valid generative factors of variation, and an orthogonal null subspace that corresponds to directions that intentionally distort the decoded outputs. This decomposition provides explicit control over both on-manifold and off-manifold generation, enabling the model to produce realistic samples alongside systematically perturbed deviations driven by the null space. Experimental results on the MNIST dataset demonstrate that our model successfully learns a stable, low-dimensional manifold while effectively isolating directions that lead to reliable off-manifold outputs.

This paper investigates the planning of a minimum-time tour for a three-dimensional (3D) Dubins airplane model to visually inspect a series of static targets in an urban environment. The inspection tour contains path segments that continuously observe each target (without occlusion) for a predetermined dwell time. Such tours are relevant to applications in urban surveillance and infrastructure inspection. The proposed approach overcomes gaps in existing methods by simultaneously considering 3D visibility constraints, different camera models (body-fixed or gimbaled), dwell times enforced along straight segments, the inspection of multiple targets concurrently, and cost savings gained via altitude changes. The approach presented is underpinned by the computation of visibility volumes as triangular meshes, which encode admissible viewing locations, and a 3D Dubins traveling salesperson problem with overlapping neighborhoods (DTSPN). Several sampling strategies to generate vehicle configurations with feasible dwell segments are proposed and used to formulate the DTSPN graph. Additionally, a heuristic approach is proposed to improve computation time by approximating DTSPN edge costs with a lower bound. Simulation results demonstrate that using the proposed 3D methods leads to a reduction in tour cost compared to two-dimensional constant-altitude methods. A flight test experiment illustrates the practicality of the approach in a simple inspection scenario.

This article describes an RF Direction Finding (DF) payload developed for UAV systems. DF payloads sense RF signals using an antenna array and process the received signals at each antenna location to estimate the number of transmitting RF sources and their bearing relative to the payload. This article uses an open source Software Defined Radio (SDR) known as the KrakenSDR which senses transmitted RF data with (5) antennas. A new deep learning architecture is proposed for estimating the azimuthal Direction of Arrival (DoA) of RF signals from the sensed KrakenSDR antenna data. The recent availability of the compact and comparatively lightweight KrakenSDR hardware for DF applications make academic investigation of this sensor for UAS possible. DF payloads are used in a wide variety of important applications including search-and-rescue, signal intelligence, RF source geolocation, spectrum monitoring, spectrum enforcement and disaster management contexts. This article describes results for a new DoA estimation algorithm and includes discussion on integration challenges, mechanical and electromagnetic design considerations and the payload Size Weight and Power-Cost (SWaP-C) metrics using the KrakenSDR hardware.

This article describes deep learning models for peak detection for FMCW radar depth estimation. Peak detection is a common problem for many fields, where in radar it is used for distance and speed estimation by analyzing the in-phase and quadrature components of the signal. Traditional methods for peak detection focus on local maxima and local derivative information to determine the location of a peak, however, because of this it is fine-tuned for specific peak shapes. We seek to approach this problem using a custom deep neural network to find peaks of various shapes and sizes that would not be found using traditional methods. Using this approach, it is demonstrated that it can outperform commonly used peak detection algorithms such as Scipy's find_peaks. For this article, we have: (1) a new AI algorithm for peak detection for FMCW radar depth estimation, (2) viewed the problem as a regression vs classification problem, and (3) Considered the pros and con of multiple loss functions for both problems.

This article describes a new suite of simulation plugins for the Gazebo 3D simulator to facilitate realistic simulation of time-varying 3D wind fields and gusts. The plugins integrate with ROS and Pixhawk PX4 Software-In-The-Loop (SITL) firmware to aid in the development of robust UAS control systems. Our approach features two main components: (1) real-time plugins for simulating environmental and sensed versions of complex, spatially varying wind velocity fields, using a Fourier-based compression of large CFD datasets, and (2) real-time plugins for modeling environmental and sensed versions of short-duration windblasts. By building on opensource Gazebo and ROS software, the developed framework provides high-accuracy physics simulation with support for multiple vehicles, fostering improved flight controller design and testing in cluttered or challenging atmospheric conditions.

This paper presents a novel simulation framework designed for high-fidelity multi-physics simulation of shock waves due to blast phenomena. The framework includes simulation of the physical pressure wave and the acoustic and visual phenomena associated with the blast event using the Gazebo environment. The framework integrates advanced technologies, including the Robot Operating System (ROS), QGroundControl, and PX4 Software-In-The-Loop (SITL), to synchronize visual, acoustic, and dynamic pressure data, ensuring realistic and efficient simulations. A key innovation in this framework is the use of a client-server architecture, which enables realtime adjustments and precise multimedia data synchronization, effectively minimizing latency and improving overall simulation quality and allowing multi-vehicle simulations in a single virtual scene. The specialized plugins employed for rendering and acoustic modeling capture the intricate dynamics of explosions, enhancing the realism of visual and auditory representations. Creation of this technology allows development of control algorithms that improve autonomous vehicle control algorithms in the presence of extreme perturbations with impacts in autonomous vehicle safety and defense sectors. The proposed framework offers a robust solution for interactive simulations, demonstrating significant advancements in both the fidelity and applicability of blast effect modeling.

Rapid detection of dynamic blast events are critical for enhancing response strategies in both civilian and military applications. This paper introduces a groundbreaking ultra-low-latency blast detection system leveraging the unique capabilities of event cameras, which capture pixel-level changes in intensity at microsecond resolution. Unlike traditional video-based systems, this approach utilizes a real-time algorithm that processes events as they are captured, significantly reducing the detection latency. The proposed system combines histogram-based analysis with a novel diffusion model to accurately detect and characterize blast events through spatial and temporal event data aggregation. Experimental results demonstrate that this method achieves detection latencies on the order of milliseconds, substantially faster than conventional methods. The system's efficacy is validated through rigorous testing in simulated environments, showcasing its potential to significantly enhance real-time surveillance and safety monitoring systems. The implementation promises sub-stantial improvements in operational response times, crucial for environments where rapid reaction is paramount.

This article presents an approach for recreating experimental Unmanned Aerial Vehicle (UAV) flight in the state-of-the-art 3D simulation software. Through the use of the Unreal Engine, AirSim simulator, and the Cesium for Unreal plugin with Google Maps, we demonstrate replicating an experiment of a real-world flight in the digital twin environment of the same location. Work investigates the viability of replicating real-world experiments by assessing the similarity between the experimental results of the real-world and digital twin experiments. The experiments involve analyzing the image telemetry and map generated of the real-world and digital twin images using the Direct Sparse Odometry (DSO) algorithm. The results have shown that replicating the real-world experiment in the digital environment produces similar results to those seen in the real-world. This will allow researchers to explore the impact of sensor, vehicle, and algorithm parameters in a controlled, repeatable environment before real-world deployment.

UAV missions often require specific geometric constraints to be satisfied between ground locations and the vehicle location. Such requirements are typical for contexts where line-of-sight must be maintained between the vehicle location and the ground control location and are also important in surveillance applications where the UAV wishes to be able to sense, e.g., with a camera sensor, a specific region within a complex geometric environment. This problem is further complicated when the ground location is generalized to a convex 2D polygonal region. This article describes the theory and implementation of a system which can quickly calculate the 3D volume that encloses all 3D coordinates from which a 2D convex planar region can be entirely viewed; referred to as a visibility volume. The proposed approach computes visibility volumes using a combination of depth map computation using GPU-acceleration and geometric boolean operations. Solutions to this problem require complex 3D geometric analysis techniques that must execute using arbitrary precision arithmetic on a collection of discontinuous and non-analytic surfaces. Post-processing steps incorporate navigational constraints to further restrict the enclosed coordinates to include both visibility and navigation constraints. Integration of sensing visibility constraints with navigational constraints yields a range of navigable space where a vehicle will satisfy both perceptual sensing and navigational needs of the mission. This algorithm then provides a synergistic perception and navigation sensitive solution yielding a volume of coordinates in 3D that satisfy both the mission path and sensing needs.

This article describes a novel fusion of a generative formal model for three-dimensional (3D) shapes with deep learning (DL) methods to understand the geometric structure of 3D objects and the relationships between their components, given a collection of unorganized point cloud measurements. Formal 3D shape models are implemented as shape grammar programs written in Procedural Shape Modeling Language (PSML). Users write PSML programs to describe complex objects, and DL networks estimate the configured free parameters of the program to generate 3D shapes. Users write PSML programs to enforce fundamental rules that define an object class and encode object attributes, including shapes, components, size, position, etc., into a parametric representation of objects. This fusion of the generative model with DL offers artificial intelligence (AI) models an opportunity to better understand the geometric organization of objects in terms of their components and their relationships to other objects. This approach allows human-in-the-loop control over DL estimates by specifying lists of candidate objects, the shape variations that each object can exhibit, and the level of detail or, equivalently, dimension of the latent representation of the shape. The results demonstrate the advantages of the proposed method over competing approaches.

This article discusses the use of a simulated environment to predict algorithm results in the real world. Simulators are crucial in allowing researchers to test algorithms, sensor integration, and navigation systems without deploying expensive hardware. This article examines how the AirSim simulator, Unreal Engine, and Cesium plugin can be used to generate simulated digital twin models of real-world locations. Several technical challenges in completing the analysis are discussed and the technical solutions are detailed in this article. Work investigates how to assess mapping results for a real-life experiment using Cesium Tiles provided by digital twins of the experimental location. This is accompanied by a description of a process for duplicating real-world flights in simulation. The performance of these methods is evaluated by analyzing real-life and experimental image telemetry with the Direct Sparse Odometry (DSO) mapping algorithm. Results indicate that Cesium Tiles environments can provide highly accurate models of ground truth geometry after careful alignment. Further, results from real-life and simulated telemetry analysis indicate that the virtual simulation results accurately predict real-life results. Findings indicate that the algorithm results in real life and in the simulated duplicate exhibited a high degree of similarity. This indicates that the use of Cesium Tiles environments as a virtual digital twin for real-life experiments will provide representative results for such algorithms. The impact of this can be significant, potentially allowing expansive virtual testing of robotic systems at specific deployment locations to develop solutions that are tailored to the environment and potentially outperforming solutions meant to work in completely generic environments.

This paper presents a data-set of performance characteristics of nearly two hundred vertical/short takeoff and landing (V/STOL) uncrewed aerial systems (UAS). Characteristics of the UAS that are recorded include maximum gross takeoff weight, endurance, maximum length dimension, speed, payload, and payload fraction. The data-set is restricted to small UAS that weigh under 500 lbs. The results are visualized via scatter plots and statistically characterized. The performance of different UAS design types and UAS Group Numbers are compared. The data-set provides a snapshot of current capabilities of small UAS in the V/STOL category and may be useful to UAS developers and procurement agencies.

This paper describes the hardware design and flight demonstration of a small quadrotor with imaging sensors for urban mapping, hazard avoidance, and target tracking research. The vehicle is equipped with five cameras, including two pairs of fisheye stereo cameras that enable a nearly omnidirectional view and a two-axis gimbaled camera. An onboard NVIDIA Jetson Orin Nano computer running the Robot Operating System software is used for data collection. An autonomous tracking behavior was implemented to coordinate the motion of the quadrotor and gimbaled camera to track a moving GPS coordinate. The data collection system was demonstrated through a flight test that tracked a moving GPS-tagged vehicle through a series of roads and parking lots. A map of the environment was reconstructed from the collected images using the Direct Sparse Odometry (DSO) algorithm. The performance of the quadrotor was also characterized by acoustic noise, communication range, battery voltage in hover, and maximum speed tests.

This article presents a comprehensive review of and analysis of state-of-the-art mapping algorithms for UAV (Unmanned Aerial Vehicle) applications, focusing on canopy-level and high-speed scenarios. This article presents a comprehensive exploration of sensor technologies suitable for UAV mapping, assessing their capabilities to provide measurements that meet the requirements of fast UAV mapping. Furthermore, the study conducts extensive experiments in a simulated environment to evaluate the performance of three distinct mapping algorithms: Direct Sparse Odometry (DSO), Stereo DSO (SDSO), and DSO Lite (DSOL). The experiments delve into mapping accuracy and mapping speed, providing valuable insights into the strengths and limitations of each algorithm. The results highlight the versatility and shortcomings of these algorithms in meeting the demands of modern UAV applications. The findings contribute to a nuanced understanding of UAV mapping dynamics, emphasizing their applicability in complex environments and high-speed scenarios. This research not only serves as a benchmark for mapping algorithm comparisons but also offers practical guidance for selecting sensors tailored to specific UAV mapping applications.

Manual identification of archaeological features in LiDAR imagery is labor-intensive, costly, and requires archaeological expertise. This paper shows how recent advancements in deep learning (DL) present efficient solutions for accurately segmenting archaeological structures in aerial LiDAR images using the YOLOv8 neural network. The proposed approach uses novel pre-processing of the raw LiDAR data and dataset augmentation methods to produce trained YOLOv8 networks to improve accuracy, precision, and recall for the segmentation of two important Maya structure types: annular structures and platforms. The results show an IoU performance of 0.842 for platforms and 0.809 for annular structures which outperform existing approaches. Further, analysis via domain experts considers the topological consistency of segmented regions and performance vs. area providing important insights. The approach automates time-consuming LiDAR image labeling which significantly accelerates accurate analysis of historical landscapes.

This article presents an analysis of current state-of-the-art sensors and how these sensors work with several mapping algorithms for UAV (Unmanned Aerial Vehicle) applications, focusing on low-altitude and high-speed scenarios. A new experimental construct is created using highly realistic environments made possible by integrating the AirSim simulator with Google 3D maps models using the Cesium Tiles plugin. Experiments are conducted in this high-realism simulated environment to evaluate the performance of three distinct mapping algorithms: (1) Direct Sparse Odometry (DSO), (2) Stereo DSO (SDSO), and (3) DSO Lite (DSOL). Experimental results evaluate algorithms based on their measured geometric accuracy and computational speed. The results provide valuable insights into the strengths and limitations of each algorithm. Findings quantify compromises in UAV algorithm selection, allowing researchers to find the mapping solution best suited to their application, which often requires a compromise between computational performance and the density and accuracy of geometric map estimates. Results indicate that for UAVs with restrictive computing resources, DSOL is the best option. For systems with payload capacity and modest compute resources, SDSO is the best option. If only one camera is available, DSO is the option to choose for applications that require dense mapping results.

This paper presents a software pipeline that enables simulating a quadrotor’s flight in realistic urban wind fields where complex wind phenomena are common and have significant impact on vehicle dynamics. The pipeline integrates the OpenStreetMap database for obtaining real-world building geometry, the OpenFOAM computational fluid dynamics (CFD) solver for computing a three-dimensional, steady, wind field, the Gazebo robotics simulation environment, and the PX4 software-in-the-loop autopilot. A 3D wind plugin is developed to interpolate a pre-computed CFD wind field at runtime during the simulation. The approach is demonstrated by comparing the flight performance of a quadrotor flying over a university campus environment with and without the wind field.

Synthetic Aperture Radar (SAR) systems sense electromagnetic backscatter from scenes generated from a sequence of excitation pulses of RF radiation emitted from the radar antenna varying spatial positions. Focusing the radar returns into coherent images requires highly accurate knowledge of the antenna position for the duration of the pulses. In this article a massively parallel approach is propose to solve the NP-hard problem of focusing radar data collected in the presence of large motion errors. Little research has been dedicated to the development of focusing algorithms capable of image formation when motion error magnitudes exceed the nominal wavelength of the radar excitation signal. This problem has been shown to be non-deterministic polynomial-time hard (NP-hard) to solve and computational challenges are exacerbated by the high computational cost of associated with the SAR focusing algorithms needed to conduct the search. The proposed approach seeks to address these challenges by restricting trajectories to smooth (low-order) curve trajectories and applying an optimized massively parallel GPU implementation of the SAR focusing algorithm to search over candidate trajectories for the trajectory yielding a focused SAR image.

This paper investigates the problem of planning a minimum-length tour for a three-dimensional Dubins airplane model to visually inspect a series of targets located on the ground or exterior surface of objects in an urban environment. Objects are 2.5D extruded polygons representing buildings or other structures. A visibility volume defines the set of admissible (occlusion-free) viewing locations for each target that satisfy feasible airspace and imaging constraints. The Dubins traveling salesperson problem with neighborhoods (DTSPN) is extended to three dimensions with visibility volumes that are approximated by triangular meshes. Four sampling algorithms are proposed for sampling vehicle configurations within each visibility volume to define vertices of the underlying DTSPN. Additionally, a heuristic approach is proposed to improve computation time by approximating edge costs of the 3D Dubins airplane with a lower bound that is used to solve for a sequence of viewing locations. The viewing locations are then assigned pitch and heading angles based on their relative geometry. The proposed sampling methods and heuristics are compared through a Monte-Carlo experiment that simulates view planning tours over a realistic urban environment.