 |
||
|
Advanced Search |
||
|
|||||
|
Formal Methods < Verification & Validation < Results Home
|
|
Click here to complete a short survey. The results of this survey will be used to help us improve the research program and this website.
Click here to view research projects that had new research results added in the last 90 days.
|
| Bayesian Verification & Validation Tools for Adaptive Systems | |
| Point of Contact |
Johann Schumann Schumann@email.arc.nasa.gov |
| Dates | January 2004 - December 2006 |
| Problem | Reliable operation in a complex, changing and uncertain environment is an important requirement for NASA missions. Changing conditions during the mission prohibit an accurate estimation of the system behavior during design phase. Here, adaptive control architectures (e.g., with neural networks) are applied and play an important role. We will develop and mature a V&V software process and tools for performance analysis of adaptive systems, which make sure that any action of the system under changing conditions is safe and accurate. We focus on a Bayesian approach that provides probabilistic estimates about the system performance and a safe operation envelope. These V&V and monitoring tools will be evaluated in NASA relevant case studies. |
| Objective | Safety is a major priority in all NASA missions. One way to increase reliability is to ensure that any action taken is safe and accurate in the current state of the system, e.g., by using an adaptive controller. The factor limiting the use of adaptive systems is the inability to provide a theoretically sound and practical V&V approach. In this research, we will develop techniques and tools with the ultimate aim of guaranteeing system robustness for a wide variety of NASA missions. In particular, we will work on tools for V&V, analysis and monitoring purposes, which enable the estimation of control system/model performance, safety envelope and accuracy, and the refinement of a V&V Software Process for infusion of such tools in NASA missions. |
| Results |
SAS 05 Executive Presentation.ppt SAS 05 Technical Presentation.ppt Report on Case Study I.pdf Prototypical implementation of Bayesian Performance Modeling tool for system identification and report on initial experiments.pdf Report on principle of operation and prototypical implementation of Bayesian Envelope tool for Neural Networks.pdf Report on Case Study II.pdf Report on approach to extend tools toward other model representation methods.pdf |
| Keywords | Bayesian methods, adaptive systems, neural networks, V&V tools, V&V software process |
| Categories |
Software Reliability Software Safety Domain-Specific Analysis Formal Methods |
|
| |
|
Curator: Josh Stonestreet NASA Official: Lisa Montgomery |
|
NASA Privacy, Security, Notices |