UAV Design for Autonomy: Challenges and Enabling Technologies
Tuesday, June 21, 2016, 9:00-13:00
George J. Vachtsevanos, Professor Emeritus
School of Electrical and Computer Engineering
Georgia Institute of Technology
Kimon P. Valavanis
John Evans Professor
Director, DU Unmanned Systems Research Institute (DU2SRI)
University of Denver
This workshop / tutorial presents a survey of the state-of-the-art technologies aimed to improve UAV autonomy attributes, reliability and resilience to stresses / extreme operating conditions, environment uncertainty and other disturbances. It summarizes methods, techniques and tools required to: detect on-board the UAV incipient failures; predict the remaining useful life (RUL) of failing components (and/or sub-systems); take appropriate corrective action(s) in the event of a contingency. Human-machine interface (HMI) issues are emphasized as they constitute a major concern due to numerous mishaps attributed to human errors, and a dynamic case-based reasoning (DCBR) paradigm is followed to reduce operator workload. As reliability analysis and life cycle management of unmanned systems constitute significant research thrusts, design for resilience is also covered with emphasis in modeling, prognostics and optimum complex system design in the presence of large-grain disturbances.
Tutorial Outline – Topics to be Covered
- A Framework for UAV Design for Autonomy
- Condition Based Maintenance (CBM), Prognostics and Health Management (PHM)
- Human – Machine Interface (HMI)
- Reliability Analysis and Life Cycle Management
- Fault Tolerant Control
- Design for Resilience
- Discussion, Q&A
A Framework for UAV Design for Autonomy: An architectural framework is presented to evaluate platform autonomy/autonomy attributes/autonomous operation based on the fundamental concepts of trust, risk and confidence. The platform reconfiguration problem is considered as an essential component of assured autonomy, which coupled with fault diagnosis, failure prognosis, decision making and contingency management, completes the proposed framework for design for assured autonomy. The design for autonomy framework depends on its constituent modules and requires integration with the aim to assist the designer and/or system operator functioning as a decision support tool and providing urgently needed capabilities to improve safety, reliability, survivability and maintainability of critical assets.
Condition Based Maintenance and Prognostics and Health Management: A new paradigm is emerging related to the way we repair, maintain and overhaul critical aerospace and industrial processes/systems. The old paradigm of “fix it when broken”, or time-based maintenance, is being replaced with maintenance practices that are executed only when needed. The problem areas are discussed emphasizing novel technologies that build upon concepts from monitoring, sensing, data mining and diagnostics/prognostics. Enabling technologies include such data mining tools for feature extraction and selection as vibration signal analysis and optimum choice of the “best” features or condition indicators; a “smart” knowledge base is exploited as the reasoning paradigm and diagnostic methods are introduced that guarantee performance in terms of fault declaration confidence and given false alarm rate. A number of case studies is presented from the aerospace and industrial arenas to demonstrate the efficacy of the proposed technologies.
Human-Machine Interface: Recent events associated with aircraft mishaps strongly enforce what is called the automation paradox – progressively move a human’s tasks to a machine (automation). The aim is to enhance human reliance on automated (autonomous) systems that are designed and implemented on-board or off-board an aircraft with the expressed goal to monitor critical system components/subsystems, assess their health status and report on incipient (possibly catastrophic) failure modes. The pilot/sensor in-the-loop framework presents major challenges that need to be addressed if issues about the aircraft’s integrity are to be effectively managed, including conflicts between the pilot’s intent/commands and fault-tolerant control commands/advisories. A rigorous systems engineering process is suggested to analyze and design tools and techniques for platform health management, human-automation interfaces and conflict resolution between the human and automated systems. Innovative technologies are introduced to define and quantify concepts of risk, confidence and trust as essential to maximize and rely on automated system (autonomy) attributes. The output of the automated system is viewed as a decision support system supported by appropriate explanation modules.
Reliability Analysis and Life Cycle Management: Lifecycle management excellence involves all events and operations occurring during the system lifetime such as design, manufacturing, testing, operation, degradation, inspection, maintenance, repair, and failure. Lifecycle management implies not only the optimized design of engineered systems, but mainly degradation handling through monitoring, inspection, and maintenance intervention. Optimization comes from the real needs to balance the risks, the costs, and the benefits of engineering activities by searching for the best compromise between conflicting requirements such as performance, safety, and reliability. In a wider scope, lifecycle optimization should address not only the cost and safety of engineered systems themselves but also parameters related to design, use, and operation. Recent advances in lifecycle engineering, preventive maintenance strategies, reliability, and optimization techniques provide a rich ground for interdisciplinary research in engineering lifecycle management. A novel framework is introduced for lifecycle management of engineered systems typically found on a UAV focusing on reliability concepts and initially employing a battery system for proof of concept. The objective is to develop a system-based architecture that builds upon a suitable system model, innovative prognostic routines that estimate the remaining useful life (RUL) of systems subject to fault/failure modes, rigorous reliability analysis tools, and appropriate optimization methods that capitalize on optimization and reliability findings and ascertain improvements in system design and/or maintenance. Because reliability and maintenance management plays an essential role throughout the entire engineered system lifecycle, the emphasis of is on reliability-centered lifecycle management. The efficacy of the proposed approach is demonstrated via an application to an onboard lithium-ion (Li-ion) battery system in a UAV.
Fault Tolerant Control: The emergence of complex unmanned aerospace systems is driving the development and implementation of new design and control technologies that are aimed to accommodate incipient failures and improve the reliability and safety of such complex systems; there is documented need to enhance the operational integrity of aircraft via fundamental fault/failure accommodation strategies. Modern technological systems rely on sophisticated control techniques to meet increased performance and safety requirements. A fault-tolerant control strategy typically involves control systems that possess the ability to accommodate system component failures automatically [while] maintaining overall system stability and acceptable performance. It is true though those incipient failure conditions of critical aircraft components / subsystems may lead to a catastrophic event endangering the life of the vehicle and resulting in failure of mission completion. It is essential, therefore, for successful mission completion to estimate on-line in real-time the remaining useful life of failing components/subsystems and take appropriate corrective action to assure that the aircraft will survive in the presence of severe fault modes until it completes its mission. A novel approach to fault-tolerance is introduced by considering prognostic results as inputs to a Model Based Control strategy that trades off system performance with control activity in order to extend the Remaining Useful Life of the platform so that a detrimental event does not occur within the mission profile.
Design for Resilience: Concepts of intelligence/intelligent control were advanced over the past years aiming to endow complex systems with the ability to sense their internal and external environment, detect abnormal or incipient failures, predict the remaining useful time and even take corrective action to remedy potentially detrimental events (control reconfiguration, failure mitigation). Design for resilience and reliability builds upon the lessons learned from early successes and failures of this interplay between animal and machine. It relies upon characteristic attributes of the biological world such as immunity and self-healing to withstand and absorb severe disturbances. Significant advances in computing, communications, sensing and control are contributing to migrate these concepts from “science fiction” to reality. We are venturing into unknown territory for the design of resilient systems focusing on fundamental theoretical notions aiming to design a system of systems whose emergent behavior cannot usually be inferred from those of the constituent systems/subsystems. We are addressing the design for resilience of cyber physical systems, such Unmanned Aerial Vehicles (UAVs) that are subjected to severe disturbances exploiting important biologically inspired properties of immunity and self-healing. We are seeking answers to such questions: Can we make a quantum improvement to the design of cyber physical systems so that they can accomplish challenging tasks and maintain their operational integrity when subjected to severe internal/external disturbances that currently available technologies fail to achieve? It is well documented that unmanned aerial vehicles comprise more than 40% of all class A air mishaps attributed primarily to mechanical failures, human error, environmental stresses, etc. It is anticipated that the design of resilient unmanned autonomous systems will reduce significantly the catastrophic events due to severe disturbances. The objective is to introduce new methods for situational awareness, modeling of complex systems, self-organization, fault tolerant control that enable targeted systems with properties of immunity and self-healing.
- Graduate students in electrical engineering, mechanical and aerospace engineering;
- Scientists, researchers, practitioners, UAV system designers and engineers;
- UAV practitioners, researchers and developers.
Tutorial Date: Tuesday, June 21, 2016
Tutorial Duration: Half Day (9:00-13:00), 4 hours
Tutorial Material: To be delivered to participants via Dropbox invitation and USB jump drive swap.