|
|
 |
|
David Song with North Carolina A&T State University |
|
Date: October 20, 2004
Time: 10:30am
Location: NIA, Rm 404
Speaker: David Song with North Carolina A&T State University
Subject:* "On High-Confidence Control Systems"
As control methods make their way into standard practice, they have openedthe door for a wide spectrum of complex applications. Such applications could involve significant system parameter/dynamics uncertainties, changing operational environments, or even unexpected sub-system failures. This fact calls for the development of high confidence and reliable control systems capable of tolerating hardware failures and adapting to changing environments.
Fundamental issues related to designing high confidence control systems will be discussed in this talk. High confidence in control theory means robust, adaptive and reliable control. As the first portion of this talk, highly robust partial state feedback distributed/decentralized control technique and "fail-safe" neuro-adaptive control will be examined.
The talk will also stress the fact that current engineering methodologies do not easily tolerate errors or unpredictable failures, nor do they efficiently adapt to changing environments. Adding such adaptation and fault-tolerance via traditional means is rather costly. By contrast, biological systems are adaptable to new environments, can handle a large amount of uncertainty in their perception and processing of the surrounding world, and can collaborate with other biological systems to solve complex problems. Such systems also have sophisticated protection and repairing (self-healing) mechanisms. These facts tend to suggest that biologically inspired methods may offer a fundamentally new domain in addressing system robustness and reliability.
At the closure to this talk, biologically inspired method for developing high confidence control systems will be discussed. The primary focus will be system adaptation and reconfiguration using biological systems as a source of inspiration. This method involves an autonomous decision, which, upon the identification of some failures in a system, can determine a proper control reconfiguration to maintain certain level of system performance. Such a decision is made difficult by the presence of uncertainties in thesystem and by exogenous signal models. The problem is further complicated by the heavy computations involved in fault detection and accommodation (FDA), and by the vagueness in the definition of a failure in the context of system performance. The underlying issues range from those defined at the biological systems level to those formulated at the overall engineering systems level.
|
|
|
Home