Browsing by Author "Williams, Ryan K."

Now showing 1 - 20 of 67
  • Thumbnail Image
    3D Deep Learning for Object-Centric Geometric Perception
    Li, Xiaolong (Virginia Tech, 2022-06-30)
    Object-centric geometric perception aims at extracting the geometric attributes of 3D objects. These attributes include shape, pose, and motion of the target objects, which enable fine-grained object-level understanding for various tasks in graphics, computer vision, and robotics. With the growth of 3D geometry data and 3D deep learning methods, it becomes more and more likely to achieve such tasks directly using 3D input data. Among different 3D representations, a 3D point cloud is a simple, common, and memory-efficient representation that could be directly retrieved from multi-view images, depth scans, or LiDAR range images. Different challenges exist in achieving object-centric geometric perception, such as achieving a fine-grained geometric understanding of common articulated objects with multiple rigid parts, learning disentangled shape and pose representations with fewer labels, or tackling dynamic and sequential geometric input in an end-to-end fashion. Here we identify and solve these challenges from a 3D deep learning perspective by designing effective and generalizable 3D representations, architectures, and pipelines. We propose the first deep pose estimation for common articulated objects by designing a novel hierarchical invariant representation. To push the boundary of 6D pose estimation for common rigid objects, a simple yet effective self-supervised framework is designed to handle unlabeled partial segmented scans. We further contribute a novel 4D convolutional neural network called PointMotionNet to learn spatio-temporal features for 3D point cloud sequences. All these works advance the domain of object-centric geometric perception from a unique 3D deep learning perspective.
  • Thumbnail Image
    Advances in the Use of Finite-Set Statistics for Multitarget Tracking
    Jimenez, Jorge Gabriel (Virginia Tech, 2021-10-27)
    In this dissertation, we seek to improve and advance the use of the finite-set statistics (FISST) approach to multitarget tracking. We consider a subsea multitarget tracking application that poses several challenges due to factors, such as, clutter/environmental noise, joint target and sensor state dependent measurement uncertainty, target-measurement association ambiguity, and sub-optimal sensor placement. The specific application that we consider is that of an underwater mobile sensor that measures the relative angle (i.e., bearing angle) to sources of acoustic noise in order to track one or more ships (targets) in a noisy environment. However, our contributions are generalizable for a variety of multitarget tracking applications. We build upon existing algorithms and address the problem of improving tracking performance for multiple maneuvering targets by incorporation several target motion models into a FISST tracking algorithm known as the probability hypothesis density filter. Moreover, we develop a novel method for associating measurements to targets using the Bayes factor, which improves tracking performance for FISST methods as well as other approaches to multitarget tracking. Further, we derive a novel formulation of Bayes risk for use with set-valued random variables and develop a real-time planner for sensor motion that avoids local minima that arise in myopic approaches to sensor motion planning. The effectiveness of our contributions are evaluated through a mixture of real-world and simulated data.
  • Thumbnail Image
    Beyond LiDAR for Unmanned Aerial Event-Based Localization in GPS Denied Environments
    Mayalu Jr, Alfred Kulua (Virginia Tech, 2021-06-23)
    Finding lost persons, collecting information in disturbed communities, efficiently traversing urban areas after a blast or similar catastrophic events have motivated researchers to develop intelligent sensor frameworks to aid law enforcement, first responders, and military personnel with situational awareness. This dissertation consists of a two-part framework for providing situational awareness using both acoustic ground sensors and aerial sensing modalities. Ground sensors in the field of data-driven detection and classification approaches typically rely on computationally expensive inputs such as image or video-based methods [6, 91]. However, the information given by an acoustic signal offers several advantages, such as low computational needs and possible classification of occluded events including gunshots or explosions. Once an event is identified, responding to real-time events in urban areas is difficult using an Unmanned Aerial Vehicle (UAV) especially when GPS is unreliable due to coverage blackouts and/or GPS degradation [10]. Furthermore, if it is possible to deploy multiple in-situ static intelligent acoustic autonomous sensors that can identify anomalous sounds given context, then the sensors can communicate with an autonomous UAV that can navigate in a GPS-denied urban environment for investigation of the event; this could offer several advantages for time-critical and precise, localized response information necessary for life-saving decision-making. Thus, in order to implement a complete intelligent sensor framework, the need for both an intelligent static ground acoustic autonomous unattended sensors (AAUS) and improvements to GPS-degraded localization has become apparent for applications such as anomaly detection, public safety, as well as intelligence surveillance and reconnaissance (ISR) operations. Distributed AAUS networks could provide end-users with near real-time actionable information for large urban environments with limited resources. Complete ISR mission profiles require a UAV to fly in GPS challenging or denied environments such as natural or urban canyons, at least in a part of a mission. This dissertation addresses, 1) the development of intelligent sensor framework through the development of a static ground AAUS capable of machine learning for audio feature classification and 2) GPS impaired localization through a formal framework for trajectory-based flight navigation for unmanned aircraft systems (UAS) operating BVLOS in low-altitude urban airspace. Our AAUS sensor method utilizes monophonic sound event detection in which the sensor detects, records, and classifies each event utilizing supervised machine learning techniques [90]. We propose a simulated framework to enhance the performance of localization in GPS-denied environments. We do this by using a new representation of 3D geospatial data using planar features that efficiently capture the amount of information required for sensor-based GPS navigation in obstacle-rich environments. The results from this dissertation would impact both military and civilian areas of research with the ability to react to events and navigate in an urban environment.
  • Thumbnail Image
    Collaborative Multi-Robot Multi-Human Teams in Search and Rescue
    Williams, Ryan K.; Abaid, Nicole; McClure, James; Lau, Nathan; Heintzman, Larkin; Hashimoto, Amanda; Wang, Tianzi; Patnayak, Chinmaya; Kumar, Akshay (2022-04-30)
    Robots such as unmanned aerial vehicles (UAVs) deployed for search and rescue (SAR) can explore areas where human searchers cannot easily go and gather information on scales that can transform SAR strategy. Multi-UAV teams therefore have the potential to transform SAR by augmenting the capabilities of human teams and providing information that would otherwise be inaccessible. Our research aims to develop new theory and technologies for field deploying autonomous UAVs and managing multi-UAV teams working in concert with multi-human teams for SAR. Specifically, in this paper we summarize our work in progress towards these goals, including: (1) a multi-UAV search path planner that adapts to human behavior; (2) an in-field distributed computing prototype that supports multi-UAV computation and communication; (3) behavioral modeling that yields spatially localized predictions of lost person location; and (4) an interface between human searchers and UAVs that facilitates human-UAV interaction over a wide range of autonomy.
  • Thumbnail Image
    Collaborative Path Planning and Control for Ground Agents Via Photography Collected by Unmanned Aerial Vehicles
    Wood, Sami Warren (Virginia Tech, 2022-06-24)
    Natural disasters damage infrastructure and create significant obstacles to humanitarian aid efforts. Roads may become unusable, hindering or halting efforts to provide food, water, shelter, and life-saving emergency care. Finding a safe route during a disaster is especially difficult because as the disaster unfolds, the usability of roads and other infrastructure can change quickly, rendering most navigation services useless. With the proliferation of cheap cameras and unmanned aerial vehicles [UAVs], the rapid collection of aerial data after a natural disaster has become increasingly common. This data can be used to quickly appraise the damage to critical infrastructure, which can help solve navigational and logistical problems that may arise after the disaster. This work focuses on a framework in which a UAV is paired with an unmanned ground vehicle [UGV]. The UAV follows the UGV with a downward-facing camera and helps the ground vehicle navigate the flooded environment. This work makes several contributions: a simulation environment is created to allow for automated data collection in hypothetical disaster scenarios. The simulation environment uses real-world satellite and elevation data to emulate natural disasters such as floods. The environment partially simulates the dynamics of the UAV and UGV, allowing agents to ex- plore during hypothetical disasters. Several semantic image segmentation models are tested for efficacy in identifying obstacles and creating cost maps for navigation within the environ- ment, as seen by the UAV. A deep homography model incorporates temporal relations across video frames to stitch cost maps together. A weighted version of a navigation algorithm is presented to plan a path through the environment. The synthesis of these modules leads to a novel framework wherein a UAV may guide a UGV safely through a disaster area.
  • Thumbnail Image
    Collision Avoidance Using a Low-Cost Forward-Looking Sonar for Small AUVs
    Morency, Christopher Charles (Virginia Tech, 2024-03-22)
    In this dissertation, we seek to improve collision avoidance for autonomous underwater vehicles (AUVs). More specifically, we consider the case of a small AUV using a forward-looking sonar system with a limited number of beams. We describe a high-fidelity sonar model and simulation environment that was developed to aid in the design of the sonar system. The simulator achieves real-time visualization through ray tracing and approximation, and can be used to assess sonar design choices, such as beam pattern and beam location, and to evaluate obstacle detection algorithms. We analyze the benefit of using a few beams instead of a single beam for a low-cost obstacle avoidance sonar for small AUVs. Single-beam systems are small and low-cost, while multi-beam sonar systems are more expensive and complex, often incorporating hundreds of beams. We want to quantify the improvement in obstacle avoidance performance of adding a few beams to a single-beam system. Furthermore, we developed a collision avoidance strategy specifically designed for the novel sonar system. The collision avoidance strategy is based on posterior expected loss, and explicitly couples obstacle detection, collision avoidance, and planning. We demonstrate the strategy with field trials using the 690 AUV, built by the Center for Marine Autonomy and Robotics at Virginia Tech, with a prototype forward-looking sonar comprising of nine beams.
  • Thumbnail Image
    Communication-Aware, Scalable Gaussian Processes for Decentralized Exploration
    Kontoudis, Georgios Pantelis (Virginia Tech, 2022-01-25)
    In this dissertation, we propose decentralized and scalable algorithms for Gaussian process (GP) training and prediction in multi-agent systems. The first challenge is to compute a spatial field that represents underwater acoustic communication performance from a set of measurements. We compare kriging to cokriging with vehicle range as a secondary variable using a simple approximate linear-log model of the communication performance. Next, we propose a model-based learning methodology for the prediction of underwater acoustic performance using a realistic propagation model. The methodology consists of two steps: i) estimation of the covariance matrix by evaluating candidate functions with estimated parameters; and ii) prediction of communication performance. Covariance estimation is addressed with a multi-stage iterative training method that produces unbiased and robust results with nested models. The efficiency of the framework is validated with simulations and experimental data from field trials. The second challenge is to perform predictions at unvisited locations with a team of agents and limited inter-agent information exchange. To decentralize the implementation of GP training, we employ the alternating direction method of multipliers (ADMM). A closed-form solution of the decentralized proximal ADMM is provided for the case of GP hyper-parameter training with maximum likelihood estimation. Multiple aggregation techniques for GP prediction are decentralized with the use of iterative and consensus methods. In addition, we propose a covariance-based nearest neighbor selection strategy that enables a subset of agents to perform predictions. Empirical evaluations illustrate the efficiency of the proposed methods
  • Thumbnail Image
    Communication-Driven Robot Learning for Human-Robot Collaboration
    Habibian, Soheil (Virginia Tech, 2024-07-25)
    The growing presence of modern learning robots necessitates a fundamental shift in design, as these robots must learn skills from human inputs. Two main components close the loop in a human-robot interaction: learning and communication. Learning derives robot behaviors from human inputs, and communication conveys information about the robot's learning to the human. This dissertation focuses on methods that enable robots to communicate their internal state clearly while learning precisely from human inputs. We first consider the information implicitly communicated by robot behavior during human interactions and whether it can be utilized to form human-robot teams. We investigate behavioral economics to identify biases and expectations in human team dynamics and incorporate them into human-robot teams. We develop and demonstrate an optimization approach that relates high-level subtask allocations to low-level robot actions, which implicitly communicates learning to encourage human participation in robot teams. We then study how communication helps humans teach tasks to robots using active learning and interactive imitation learning algorithms. Within the active learning approach, we develop a model that forms a belief over the human's mental model about the robot's learning. We validate that our algorithm enables the robot to balance between learning human preferences and implicitly communicating its learning through questions. Within the imitation learning approach, we integrate a wrapped haptic display that explicitly communicates representations from the robot's learned behavior to the user. We show that our framework helps the human teacher improve different aspects of the robot's learning during kinesthetic teaching. We then extend this system to a more comprehensive interactive learning architecture that provides multi-modal feedback through augmented reality and haptic interfaces. We present a case study with this closed-loop system and illustrate improved teaching, trust, and co-adaptation as the measured benefits of communicating robot learning. Overall, this dissertation demonstrates that bi-directional communication helps robots learn faster and adapt better, while humans experience a more intuitive and trust-based interaction.
  • Thumbnail Image
    Considerations of Reinforcement Learning within Real-Time Wireless Communication Systems
    Jones, Alyse M. (Virginia Tech, 2022-06-15)
    Afflicted heavily by spectrum congestion, the unpredictable, dynamic conditions of the radio frequency (RF) spectrum has increasingly become a major obstacle for devices today. More specifically, a significant threat existing within this kind of environment is interference caused by collisions, which is increasingly unavoidable in an overcrowded spectrum. Thus, these devices require a way to avoid such events. Cognitive radios (CR) were proposed as a solution through its transmission adaptability and decision-making capabilities within a radio. Through spectrum sensing, CRs are able to capture the current condition of the RF spectrum and based on its decision-making strategy, interpret these results to make an informed decision on what to do next to optimize its own communication. With the emergence of artificial intelligence, one such decision-making strategy CRs can utilize is Reinforcement Learning (RL). Unlike standard adaptive radios, CRs equipped with RL can predict the conditions of the RF spectrum, and using these predictions, understand what it must do in the future to operate optimally. Recognizing the usefulness of RL in hard-to-predict environments, such as the RF spectrum, research of RL within CRs have become more popular over the past decade, especially for interference mitigation. However, the existing literature neglects to confront the possible limitations that pose a threat to the proper implementation of RL in RF systems. Therefore, this thesis is motivated to investigate what limitations in real-time communication systems can hinder the performance of RL, and as a result of these limitations, emphasize the considerations that should be a focus in the design and implementation of radio frequency reinforcement learning (RFRL) systems. The effects of latency, power, wireless channel impairments, different transmission protocols, and different spectrum sensing detectors are among the possible limitations simulated and analyzed within this work that are not typically considered within simulation-based prior art. To perform this investigation, a representative real-time OFDM transmit/receive chain is implemented within the GNU Radio framework. The system, operating over-the-air through USRPs, leverages reinforcement learning, e.g. Q-Learning, in order to avoid interference with other spectrum users. Performance analysis of this representative system provides a systematic approach for helping to predict limiting factors within an implemented real-time system and thus, aim to provide guidance on how to design these systems with these practical limitations in mind.
  • Thumbnail Image
    Constrained Clustering for Frequency Hopping Spread Spectrum Signal Separation
    White, Parker Douglas (Virginia Tech, 2019-09-16)
    Frequency Hopping Spread Spectrum (FHSS) signaling is used across many devices operating in both regulated and unregulated bands. In either situation, if there is a malicious device operating within these bands, or more simply a user operating out of the required specifications, the identification this user important to insure communication link integrity and interference mitigation. The identification of a user involves the grouping of that users signal transmissions, and the separation of those transmission from transmissions of other users in a shared frequency band. Traditional signal separation methods often require difficult to obtain hardware fingerprinting characteristics or approximate geo-location estimates. This work will consider the characteristics of FHSS signals that can be extracted directly from signal detection. From estimates of these hopping characteristics, novel source separation with classic clustering algorithms can be performed. Background knowledge derived from the time domain representation of received waveforms can improve these clustering methods with the novel application of cannot-link pairwise constraints to signal separation. For equivalent clustering accuracy, constraint-based clustering tolerates higher parameter estimation error, caused by diminishing received signal-to-noise ratio (SNR), for example. Additionally, prior work does not fully cover the implications of detecting and estimating FHSS signals using image segmentation on a Time-Frequency (TF) waterfall. This work will compare several methods of FHSS signal detection, and discuss how each method may effect estimation accuracy and signal separation quality. The use of constraint-based clustering is shown to provide higher clustering accuracy, resulting in more reliable separation and identification of active users in comparison to traditional clustering methods.
  • Thumbnail Image
    Control Design for Long Endurance Unmanned Underwater Vehicle Systems
    Kleiber, Justin Tanner (Virginia Tech, 2022-05-24)
    In this thesis we demonstrate a technique for robust controller design for an autonomous underwater vehicle (AUV) that explicitly handles the trade-off between reference tracking, agility, and energy efficient performance. AUVs have many sources of modeling uncertainty that impact the uncertainty in maneuvering performance. A robust control design process is proposed to handle these uncertainties while meeting control system performance objectives. We investigate the relationships between linear system design parameters and the control performance of our vehicle in order to inform an H∞ controller synthesis problem with the objective of balancing these tradeoffs. We evaluate the controller based on its reference tracking performance, agility and energy efficiency, and show the efficacy of our control design strategy.
  • Thumbnail Image
    Cooperative human-robot search in a partially-known environment using multiple UAVs
    Chourey, Shivam (Virginia Tech, 2020-08-28)
    This thesis details out a system developed with objective of conducting cooperative search operation in a partially-known environment, with a human operator, and two Unmanned Aerial Vehicles (UAVs) with nadir, and front on-board cameras. The system uses two phases of flight operations, where the first phase is aimed at gathering latest overhead images of the environment using a UAV’s nadir camera. These images are used to generate and update representations of the environment including 3D reconstruction, mosaic image, occupancy image, and a network graph. During the second phase of flight operations, a human operator marks multiple areas of interest for closer inspection on the mosaic generated in previous step, displayed via a UI. These areas are used by the path planner as visitation goals. The two-step path planner, which uses network graph, utilizes the weighted-A* planning, and Travelling Salesman Problem’s solution to compute an optimal visitation plan. This visitation plan is then converted into Mission waypoints for a second UAV, and are communicated through a navigation module over a MavLink connection. A UAV flying at low altitude, executes the mission plan, and streams a live video from its front-facing camera to a ground station over a wireless network. The human operator views the video on the ground station, and uses it to locate the target object, culminating the mission.
  • Thumbnail Image
    Cooperative Payload Transportation by UAVs: A Model-Based Deep Reinforcement Learning (MBDRL) Application
    Khursheed, Shahwar Atiq (Virginia Tech, 2024-08-20)
    We propose a Model-Based Deep Reinforcement Learning (MBDRL) framework for collaborative paylaod transportation using Unmanned Aerial Vehicles (UAVs) in Search and Rescue (SAR) missions, enabling heavier payload conveyance while maintaining vehicle agility. Our approach extends the single-drone application to a novel multi-drone one, using the Probabilistic Ensembles with Trajectory Sampling (PETS) algorithm to model the unknown stochastic system dynamics and uncertainty. We use the Multi-Agent Reinforcement Learning (MARL) framework via a centralized controller in a leader-follower configuration. The agents utilize the approximated transition function in a Model Predictive Controller (MPC) configured to maximize the reward function for waypoint navigation, while a position-based formation controller ensures stable flights of these physically linked UAVs. We also developed an Unreal Engine (UE) simulation connected to an offboard planner and controller via a Robot Operating System (ROS) framework that is transferable to real robots. This work achieves stable waypoint navigation in a stochastic environment with a sample efficiency following that seen in single UAV work. This work has been funded by the National Science Foundation (NSF) under Award No. 2046770.
  • Thumbnail Image
    Coverage Planning for Unmanned Aerial Vehicles
    Yu, Kevin Li (Virginia Tech, 2021-06-08)
    This dissertation investigates how to plan paths for Unmanned Aerial Vehicles (UAV) for the task of covering an environment. Three increasingly complex coverage problems based on the environment that needs to be covered are studied. The dissertation starts with a 2D point coverage problem where the UAV needs to visit a set of sites on the ground plane by flying on a fixed altitude plane parallel to the ground. The UAV has limited battery capacity which may make it infeasible to visit all the points. A novel symbiotic UAV and Unmanned Ground Vehicle (UGV) system where the UGV acts as a mobile recharging station is proposed. A practical, efficient algorithm for solving this problem using Generalized Traveling Salesperson Problem (GTSP) solver is presented. Then the algorithm is extended to a coverage problem that covers 2D regions on the ground with a UAV that can operate in fixed-wing or multirotor mode. The algorithm is demonstrated through proof-of-concept experiments. Then this algorithm is applied to covering 2D regions, not all of which lie on the same plane. This is motivated by bridge inspection application, where the UAV is tasked with visually inspecting planar regions on the bridge. Finally, a general version of the problem where the UAV is allowed to fly in complete 3D space and the environment to be covered is in 3D as well is presented. An algorithm that clusters viewpoints on the surface of a 3D structure and has an UAV autonomously plan online paths to visit all viewpoints is presented. These online paths are re-planned in real time as the UAV obtains new information on the structure and strives to obtain an optimal 3D coverage path.
  • Thumbnail Image
    DeepPaSTL: Spatio-Temporal Deep Learning Methods for Predicting Long-Term Pasture Terrains Using Synthetic Datasets
    Rangwala, Murtaza; Liu, Jun; Ahluwalia, Kulbir Singh; Ghajar, Shayan; Dhami, Harnaik Singh; Tracy, Benjamin F.; Tokekar, Pratap; Williams, Ryan K. (MDPI, 2021-11-05)
    Effective management of dairy farms requires an accurate prediction of pasture biomass. Generally, estimation of pasture biomass requires site-specific data, or often perfect world assumptions to model prediction systems when field measurements or other sensory inputs are unavailable. However, for small enterprises, regular measurements of site-specific data are often inconceivable. In this study, we approach the estimation of pasture biomass by predicting sward heights across the field. A convolution based sequential architecture is proposed for pasture height predictions using deep learning. We develop a process to create synthetic datasets that simulate the evolution of pasture growth over a period of 30 years. The deep learning based pasture prediction model (DeepPaSTL) is trained on this dataset while learning the spatiotemporal characteristics of pasture growth. The architecture purely learns from the trends in pasture growth through available spatial measurements and is agnostic to any site-specific data, or climatic conditions, such as temperature, precipitation, or soil condition. Our model performs within a 12% error margin even during the periods with the largest pasture growth dynamics. The study demonstrates the potential scalability of the architecture to predict any pasture size through a quantization approach during prediction. Results suggest that the DeepPaSTL model represents a useful tool for predicting pasture growth both for short and long horizon predictions, even with missing or irregular historical measurements.
  • Thumbnail Image
    Directional Perception of Force in a Virtual Reality Environment
    Long, Zihao (Virginia Tech, 2020-05-08)
    Force feedback during teleoperation and in Virtual Reality (VR) environments is becoming increasingly common. We are interested in understanding the impact of motion on the directional accuracy of force perception, as observed in a VR environment. We used a custom force-feedback system that pulled a handle with a force of 1.87N at various angles in front of N=14 subjects. The virtual environment showed a curved wall, which corresponded to the locations from which the force could physically originate. Subjects selected where they perceived the force to originate from with a virtual laser pointer and by orienting their head. We compared several conditions: the subject held the handle still; the subject moved the handle back and forth toward the center of the wall; the subject moved the handle back and forth across their body; and the subject moved the handle back and forth toward where they thought the force was originating. Subjects were able to localize the force with an average accuracy of 1-10 degrees depending on the force's location, which is better than previous studies. All conditions had similiar accuracies. Subjects had the best precision when they followed the force as compared to either of the other conditions with movement.
  • Thumbnail Image
    Distributed Feedback Control Algorithms for Cooperative Locomotion: From Bipedal to Quadrupedal Robots
    Kamidi, Vinaykarthik Reddy (Virginia Tech, 2022-03-25)
    This thesis synthesizes general and scalable distributed nonlinear control algorithms with application to legged robots. It explores both naturally decentralized problems in legged locomotion, such as the collaborative control of human-lower extremity prosthesis and the decomposition of high-dimensional controllers of a naturally centralized problem into a net- work of low-dimensional controllers while preserving equivalent performance. In doing so, strong nonlinear interaction forces arise, which this thesis considers and sufficiently addresses. It generalizes to both symmetric and asymmetric combinations of subsystems. Specifically, this thesis results in two distinct distributed control algorithms based on the decomposition approach. Towards synthesizing the first algorithm, this thesis presents a formal foundation based on de- composition, Hybrid Zero Dynamics (HZD), and scalable optimization to develop distributed controllers for hybrid models of collaborative human-robot locomotion. This approach con- siders a centralized controller and then decomposes the dynamics and parameterizes the feedback laws to synthesize local controllers. The Jacobian matrix of the Poincaré map with local controllers is studied and compared with the centralized ones. An optimization problem is then set up to tune the parameters of the local controllers for asymptotic stability. It is shown that the proposed approach can significantly reduce the number of controller parameters to be optimized for the synthesis of distributed controllers, deeming the method computationally tractable. To evaluate the analytical results, we consider a human amputee with the point of separation just above the knee and assume the average physical parameters of a human male. For the lower-extremity prosthesis, we consider the PRleg, a powered knee-ankle prosthetic leg, and together, they form a 19 Degrees of Freedom (DoF) model. A multi-domain hybrid locomotion model is then employed to rigorously assess the performance of the afore-stated control algorithm via numerical simulations. Various simulations involving the application of unknown external forces and altering the physical parameters of the human model unbeknownst to the local controllers still result in stable amputee loco- motion, demonstrating the inherent robustness of the proposed control algorithm. In the later part of this thesis, we are interested in developing distributed algorithms for the real-time control of legged robots. Inspired by the increasing popularity of Quadratic programming (QP)-based nonlinear controllers in the legged locomotion community due to their ability to encode control objectives subject to physical constraints, this thesis exploits the idea of distributed QPs. In particular, this thesis presents a formal foundation to systematically decompose QP-based centralized nonlinear controllers into a network of lower-dimensional local QPs. The proposed approach formulates a feedback structure be- tween the local QPs and leverages a one-step communication delay protocol. The properties of local QPs are analyzed, wherein it is established that their steady-state solutions on periodic orbits (representing gaits) coincide with that of the centralized QP. The asymptotic convergence of local QPs' solutions to the steady-state solution is studied via Floquet theory. Subsequently, to evaluate the effectiveness of the analytical results, we consider an 18 DoF quadrupedal robot, A1, as a representative example. The network of distributed QPs mentioned earlier is condensed to two local QPs by considering a front-hind decomposition scheme. The robustness of the distributed QP-based controller is then established through rigorous numerical simulations that involve exerting unmodelled external forces and intro- ducing unknown ground height variations. It is further shown that the proposed distributed QPs have reduced sensitivity to noise propagation when compared with the centralized QP. Finally, to demonstrate that the resultant distributed QP-based nonlinear control algorithm translates equivalently well to hardware, an extensive set of blind locomotion experiments on the A1 robot are undertaken. Similar to numerical simulations, unknown external forces in the form of aggressive pulls and pushes were applied, and terrain uncertainties were introduced with the help of arbitrarily displaced wooden blocks and compliant surfaces. Additionally, outdoor experiments involving a wide range of terrains such as gravel, mulch, and grass at various speeds up to 1.0 (m/s) reiterate the robust locomotion observed in numerical simulations. These experiments also show that the computation time is significantly dropped when the distributed QPs are considered over the centralized QP.
  • Thumbnail Image
    Distributed Intelligence for Multi-Agent Systems in Search and Rescue
    Patnayak, Chinmaya (Virginia Tech, 2020-11-05)
    Unfavorable environmental and (or) human displacement may engender the need for Search and Rescue (SAR). Challenges such as inaccessibility, large search areas, and heavy reliance on available responder count, limited equipment and training makes SAR a challenging problem. Additionally, SAR operations also pose significant risk to involved responders. This opens a remarkable opportunity for robotic systems to assist and augment human understanding of the harsh environments. A large body of work exists on the introduction of ground and aerial robots in visual and temporal inspection of search areas with varying levels of autonomy. Unfortunately, limited autonomy is the norm in such systems, due to the limitations presented by on-board UAV resources and networking capabilities. In this work we propose a new multi-agent approach to SAR and introduce a wearable compute cluster in the form factor of a backpack. The backpack allows offloading compute intensive tasks such as Lost Person Behavior Modelling, Path Planning and Deep Neural Network based computer vision applications away from the UAVs and offers significantly high performance computers to execute them. The backpack also provides for a strong networking backbone and task orchestrators which allow for enhanced coordination and resource sharing among all the agents in the system. On the basis of our benchmarking experiments, we observe that the backpack can significantly boost capabilities and success in modern SAR responses.
  • Thumbnail Image
    Distributed, Stable Topology Control of Multi-Robot Systems with Asymmetric Interactions
    Mukherjee, Pratik (Virginia Tech, 2021-06-17)
    Multi-robot systems have recently witnessed a swell in interest in the past few years because of their various applications such as agricultural autonomy, medical robotics, industrial and commercial automation and, search and rescue. In this thesis, we particularly investigate the behavior of multi-robot systems with respect to stable topology control in asymmetric interaction settings. From theoretical perspective, we first classify stable topologies, and identify the conditions under which we can determine whether a topology is stable or not. Then, we design a limited fields-of-view (FOV) controller for robots that use sensors like cameras for coordination which induce asymmetric robot to robot interactions. Finally, we conduct a rigorous theoretical analysis to qualitatively determine which interactions are suitable for stable directed topology control of multi-robot systems with asymmetric interactions. In this regard, we solve an optimal topology selection problem to determine the topology with the best interactions based on a suitable metric that represents the quality of interaction. Further, we solve this optimal problem distributively and validate the distributed optimization formulation with extensive simulations.  For experimental purposes, we developed a portable multi-robot testbed which enables us to conduct multi-robot topology control experiments in both indoor and outdoor settings and validate our theoretical findings. Therefore, the contribution of this thesis is two fold: i) We provide rigorous theoretical analysis of  stable coordination of multi-robot systems with directed graphs, demonstrating the graph structures that induce stability for a broad class of coordination objectives; ii) We develop a testbed that enables validating multi-robot topology control in both indoor and outdoor settings.
  • Thumbnail Image
    Drone Cellular Networks: Fundamentals, Modeling, and Analysis
    Banagar, Morteza (Virginia Tech, 2022-06-23)
    With the increasing maturity of unmanned aerial vehicles (UAVs), also known as drones, wireless ecosystem is experiencing an unprecedented paradigm shift. These aerial platforms are specifically appealing for a variety of applications due to their rapid and flexible deployment, cost-effectiveness, and high chance of forming line-of-sight (LoS) links to the ground nodes. As with any new technology, the benefits of incorporating UAVs in existing cellular networks cannot be characterized without completely exploring the underlying trade space. This requires a detailed system-level analysis of drone cellular networks by taking the unique features of UAVs into account, which is the main objective of this dissertation. We first focus on a static setup and characterize the performance of a three-dimensional (3D) two-hop cellular network in which terrestrial base stations (BSs) coexist with UAVs to serve a set of ground user equipment (UE). In particular, a UE connects either directly to its serving terrestrial BS by an access link or connects first to its serving UAV which is then wirelessly backhauled to a terrestrial BS (joint access and backhaul). We consider realistic antenna radiation patterns for both BSs and UAVs using practical models developed by the third generation partnership project (3GPP). We assume a probabilistic channel model for the air-to-ground transmission, which incorporates both LoS and non-LoS links. Assuming the max-power association policy, we study the performance of the network in both amplify-and-forward (AF) and decode-and-forward (DF) relaying protocols. Using tools from stochastic geometry, we analyze the joint distribution of distance and zenith angle of the closest (and serving) UAV to the origin in a 3D setting. Further, we identify and extensively study key mathematical constructs as the building blocks of characterizing the received signal-to-interference-plus-noise ratio (SINR) distribution. Using these results, we obtain exact mathematical expressions for the coverage probability in both AF and DF relaying protocols. Furthermore, considering the fact that backhaul links could be quite weak because of the downtilted antennas at the BSs, we propose and analyze the addition of a directional uptilted antenna at the BS that is solely used for backhaul purposes. The superiority of having directional antennas with wirelessly backhauled UAVs is further demonstrated via extensive simulations. Second, we turn our attention to a mobile setup and characterize the performance of several canonical mobility models in a drone cellular network in which UAV base stations serve UEs on the ground. In particular, we consider the following four mobility models: (i) straight line (SL), (ii) random stop (RS), (iii) random walk (RW), and (iv) random waypoint (RWP), among which the SL mobility model is inspired by the simulation models used by the 3GPP for the placement and trajectory of UAVs, while the other three are well-known canonical models (or their variants) that offer a useful balance between realism and tractability. Assuming the nearest-neighbor association policy, we consider two service models for the UEs: (i) UE independent model (UIM), and (ii) UE dependent model (UDM). While the serving UAV follows the same mobility model as the other UAVs in the UIM, it is assumed to fly towards the UE of interest in the UDM and hover above its location after reaching there. We then present a unified approach to characterize the point process of UAVs for all the mobility and service models. Using this, we provide exact mathematical expressions for the average received rate and the session rate as seen by the typical UE. Further, using tools from the calculus of variations, we concretely demonstrate that the simple SL mobility model provides a lower bound on the performance of other general mobility models (including the ones in which UAVs follow curved trajectories) as long as the movement of each UAV in these models is independent and identically distributed (i.i.d.). Continuing our analysis on mobile setups, we analyze the handover probability in a drone cellular network, where the initial positions of the UAVs serving the ground UEs are modeled by a homogeneous Poisson point process (PPP). Inspired by the mobility model considered in the 3GPP studies, we assume that all the UAVs follow the SL mobility model, i.e., move along straight lines in random directions. We further consider two different scenarios for the UAV speeds: (i) same speed model (SSM), and (ii) different speed model (DSM). Assuming nearest-neighbor association policy, we characterize the handover probability of this network for both mobility scenarios. For the SSM, we compute the exact handover probability by establishing equivalence with a single-tier terrestrial cellular network, in which the BSs are static while the UEs are mobile. We then derive a lower bound for the handover probability in the DSM by characterizing the evolution of the spatial distribution of the UAVs over time. After performing these system-level analyses on UAV networks, we focus our attention on the air-to-ground wireless channel and attempt to understand its unique features. For that, we first study the impact of UAV wobbling on the coherence time of the wireless channel between UAVs and a ground UE, using a Rician multi-path channel model. We consider two different scenarios for the number of UAVs: (i) single UAV scenario (SUS), and (ii) multiple UAV scenario (MUS). For each scenario, we model UAV wobbling by two random processes, i.e., the Wiener and sinusoidal processes, and characterize the channel autocorrelation function (ACF) which is then used to derive the coherence time of the channel. For the MUS, we further show that the UAV-UE channels for different UAVs are uncorrelated from each other. One key observation that is revealed from our analysis is that even for small UAV wobbling, the coherence time of the channel may degrade quickly, which may make it difficult to track the channel and establish a reliable communication link. Finally, we develop an impairments-aware air-to-ground unified channel model that incorporates the effect of both wobbling and hardware impairments, where the former is caused by random physical fluctuations of UAVs, and the latter by intrinsic radio frequency (RF) nonidealities at both the transmitter and receiver, such as phase noise, in-phase/quadrature (I/Q) imbalance, and power amplifier (PA) nonlinearity. The impact of UAV wobbling is modeled by two stochastic processes, i.e., the canonical Wiener process and the more realistic sinusoidal process. On the other hand, the aggregate impact of all hardware impairments is modeled as two multiplicative and additive distortion noise processes, which is a well-accepted model. For the sake of generality, we consider both wide-sense stationary (WSS) and nonstationary processes for the distortion noises. We then rigorously characterize the ACF of the wireless channel, using which we provide a comprehensive analysis of four key channel-related metrics: (i) power delay profile (PDP), (ii) coherence time, (iii) coherence bandwidth, and (iv) power spectral density (PSD) of the distortion-plus-noise process. Furthermore, we evaluate these metrics with reasonable UAV wobbling and hardware impairment models to obtain useful insights. Similar to our observation above, this work again demonstrates that the coherence time severely degrades at high frequencies even for small UAV wobbling, which renders air-to-ground channel estimation very difficult at these frequencies.