Adaptive verification for an on-line learning neural-based flight control system

This paper presents a complex adaptive systems approach for the verification of an adaptive, online learning, sigma-pi neural network that is used for the intelligent flight control system (IFCS) that has the potential of commercial aviation application. This paper reports on the partial completion of my doctoral dissertation proposal at Nova Southeastern University, in the Graduate School of Computer and Information Sciences. The most significant shortcoming of the prior and current approaches to verifying adaptive neural networks is the application of linear approaches to a non-linear problem. The project will use a MatLab simulation of the sigma-pi adaptive neural network and an aircraft simulation to fly a series of simulated flight tests. As a result of the flight simulations, a statistical analysis of the neural network weights is performed as input to both a complexity analysis and a neural network rule extraction analysis. Complex adaptive methods are a novel approach to overcome previous linear analysis limitations. Future work will be required to analyze emergent behavior of the neural network weights to show stability and convergence characteristics. Advances in computational power and neural network techniques for estimating aerodynamic stability and control derivatives provide opportunity for real-time adaptive control. New verification techniques are needed that substantially increases trustworthiness in the use of these neural network systems in life critical systems. Verification of neural-based IFCS is currently an urgent and significant research and engineering topic since these systems are being looked upon as a new approach for aircraft survivability, for both commercial and military.