Resilient Design of Unmanned Autonomous Systems
Professor George Vachtsevanos
Georgia Institute of Technology
Resilience is defined as the ability of a system to absorb or accommodate extreme/severe disturbances without an adverse effect on the system’s operational integrity; ability of system to maintain an acceptable level of performance in the face of large-grain fault/failure modes. Typically, resilience is measured by the magnitude of the disturbance that can be accommodated. Maximum resilience is defined as the resilience to maximum or worst disturbance that can be accommodated, i.e. a disturbance that can have a significant negative impact on the system’s integrity. In the design for resilience framework, resilience is quantified considering two system-level properties: self-organization and control reconfiguration or fault-tolerance. The first level of protection against disturbances involves monitoring the system state, i.e. correlating internal and external events to temporal, spatial and functional constraints. Then the design for resilience of complex physical systems imposes a crucial requirement: Faults/incipient failure modes, caused from severe disturbances must be detected reliably. We introduce next a systematic methodology to resilience design starting with a thorough modeling approach assisted by concepts from Complexity Theory.
In the life degradation modeling module, we begin by introducing a prototypical unmanned autonomous vehicle structure and associated function subjected to fault/failure mechanisms as well as large-grain disturbances. We exploit complexity theory concepts to understand the system’s behavioral patterns with specific emphasis on its “emergent” behaviors. Self-organization of autonomous systems begins with spectral and epidemic spreading models to represent the behaviors of the system as a network. The system self-organizes via an optimization method seeking to derive the “best” compensating mechanism to achieve set objectives while maintaining system stability. Reconfigurable design of systems centers on incorporating autonomy and resilience, flexibility, interoperability, sustainment and reliability under changing operational requirements and uncertain/dynamic environments or mission profiles, without major changes to the system’s initial design. Resilience and robustness refer to complex manned and unmanned systems maintaining consistently an acceptable level of performance in the presence of unexpected large-grain internal and external disturbances (fault/failures, etc.) in unstructured environments. Reconfigurable design and operation is formulated as an optimization problem where new or reconfigured designs and their operational characteristics are optimized to perform as designed/desired and traditional engineering focuses on top-down system optimization to meet set objectives, typically a set of engineering performance measures. We are introducing a novel approach to fault-tolerance 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 (RUL) objectives, typically a set of engineering performance measures. The proposed fault-tolerant control architecture performs one of three actions, low-level control reconfiguration at the component level, mid-level control redistribution at the sub-system level, and high-level mission adaptation at the highest echelon of the architecture. We illustrate the efficacy of the self-organization and control reconfiguration strategies with typical autonomous system examples.
Brief Biography of the Speaker:
Dr. George Vachtsevanos is currently serving as Professor Emeritus at the Georgia Institute of Technology. He served as Professor of Electrical and Computer Engineering at the Georgia Institute of Technology from 1984 until September, 2007. Dr Vachtsevanos directs at Georgia Tech the Intelligent Control Systems laboratory where faculty and students began research in diagnostics in 1985 with a series of projects in collaboration with Boeing Aerospace Company funded by NASA and aimed at the development of fuzzy logic based algorithms for fault diagnosis and control of major space station subsystems. His work in Unmanned Aerial Vehicles dates back to 1994 with major projects funded by the U.S. Army and DARPA. He has served as the Co-PI for DARPA’s Software Enabled Control program over the past six years and directed the development and flight testing of novel fault-tolerant control algorithms for Unmanned Aerial Vehicles. He has represented Georgia Tech at DARPA’s HURT program where multiple UAVs performed surveillance, reconnaissance and tracking missions in an urban environment. Under AFOSR sponsorship, the Impact/Georgia Team is developing a biologically-inspired micro aerial vehicle. His research work has been supported over the years by ONR, NSWC, the MURI Integrated Diagnostic program at Georgia Tech, the U,S. Army’s Advanced Diagnostic program, General Dynamics, General Motors Corporation, the Academic Consortium for Aging Aircraft program, the U.S. Air Force Space Command, Bell Helicopter, Fairchild Controls, among others. He has published over 300 technical papers and is the recipient of the 2002-2003 Georgia Tech School of ECE Distinguished Professor Award and the 2003-2004 Georgia Institute of Technology Outstanding Interdisciplinary Activities Award. He is the lead author of a book on Intelligent Fault Diagnosis and Prognosis for Engineering Systems published by Wiley in 2006.