Department of Computer Science
San Jose State University
Special Talk:
Verification and Validation of Neuro-adaptive Aircraft Control Systems
By Johann
Schumann
Senior Scientist
Research Institute for Advanced Computer Science
NASA/Ames
Thursday, February 7, 2008, 3 p.m.
MacQuarrie Hall Room 225
Traditional fixed-gain control has proven to be unsuccessful
to deal with complex, strongly nonlinear, or changing systems such as a
damaged aircraft. Control systems, which use a neural network that can
adapt toward changes in the plant, have been actively investigated as
they offer many advantages. An online adaptive controller has been
implemented and successfully test-flown on a manned F-15 aircraft at
NASA DFRC within the Intelligent Flight Control System (IFCS) project.
In this talk, I will briefly introduce adaptive flight control and will
discuss the specific challenges for the verification and validation
(V&V) of such systems. As performance and safety guarantees cannot
be provided at development time, novel tools and approaches to support
V&V and certification had to be developed. I will present our
Confidence tool, which dynamically estimates the neural network
performance, and discuss our V&V approach.
Dr. Johann Schumann (PhD
1991, Dr. habil 2000, Munich, Germany) is a Senior Scientist with RIACS
and working in the Robust Software Engineering Group at NASA Ames. He
is engaged in research on verification and validation of autonomy
software and adaptive controllers, and on automatic generation of
reliable code for data analysis and state estimation. Dr. Schumann is
author of a book on theorem proving in software engineering and has
published more than 70 articles on automated deduction, automatic
program generation, and neural network oriented topics.