Fully Tuned Radial Basis Function Neural Networks for Flight Control presents the use of the Radial Basis Function (RBF) neural networks for adaptive control of nonlinear systems with emphasis on flight control applications A Lyapunov synthesis approach is used to derive the tuning rules for the RBF controller parameters in order to guarantee the stability of the closed loop system Unlike previous methods that tune only the weights of the RBF network, this book presents the derivation of the tuning law for tuning the centers, widths, and weights of the RBF network, and compares the results with existing algorithms It also includes a detailed review of system identification, including indirect and direct adaptive control of nonlinear systems using neural networks

Fully Tuned Radial Basis Function Neural Networks for Flight Control

A robust neuro-control scheme is presented for aircraft auto-landing under severe wind conditions and partial loss of control surfaces. In the scheme, a dynamic radial basis function network (RBFN) called minimal resource allocating network (MRAN), that incorporates a growing and pruning strategy, is utilize to aid an H/sub /spl infin// controller using a feedback-error-learning mechanism. The neural network uses only online learning and is not trained "a priori". Specifically, the performance of this neuro-controller for aircraft auto-landing in a microburst along with a partial loss of control effectiveness is analyzed and compared with other control schemes. Simulation studies show that the performance obtained by the neuro-H/sub /spl infin// control scheme is better than the other control schemes under failure and extreme wind conditions.

Robust neuro-H/sub /spl infin// controller design for aircraft auto-landing

The objective of this document Is to show the capabilities of parallel hardware-based on-line learning neural networks (NNs). This specific application is related to an on-line estimation problem for sensor validation purposes. Neural-network-based microprocessors are starting to be commercially available. However, most of them feature a learning performed with the classic back-propagation algorithm (BPA). To overcome this lack of flexibility a customized motherboard with transputers was implemented for this investigation, The extended BPA (EBPA), a modified and more effective BPA, was used for the on-line learning, These parallel hardware-based neural architectures were used to implement a sensor failure detection, identification, and accommodation scheme in the model of a night control system assumed to be without physical redundancy in the sensory capabilities. The results of this study demonstrate the potential for these neural schemes for implementation in actual flight control systems of modern high performance aircraft, taking advantage of the characteristics of the extended back-propagation along with the parallel computation capabilities of NN customized hardware.

Sensor validation using hardware-based on-line learning neural networks

A fuzzy logic based F/A-18 automatic carrier landing system

Performance comparisons of three nonlinear dynamic inversion (NDI) implementation schemes, namely numerical, neural network (NN), and adaptive fuzzy system (AFS), are presented. NDI is known to be a simple yet powerful control strategy, but its full implementation is problematic due to requirements to invert a known nonlinear model. In this paper we examine three different solutions. The baseline NDI controller uses numerical inversion based on the NewtonRhapson technique. The NN controller utilizes a singlehidden-layer, 25-neuron NN running the backpropagation algorithm. The AFS controller executes fuzzy rules that were dynamically created during training based on the input-output relations and the predefined membership sets. Both the NN and the AFS controllers were trained off-line until reasonable convergence was achieved. These NDI controls were applied to a longitudinal maneuver of the X29 research aircraft and they were evaluated in terms of complexity in training and ability to follow a particular trajectory. The results obtained indicate that, although each technique has its own strengths, NN controller was the best both in terms of training ease, reasonable data mapping, model-following accuracy, and control effort.

Comparison of Neural-Network-Based, Fuzzy-Logic-Based, and Numerical Nonlinear Inverse Flight Controls

Summary form only given. Neural networks have the potential to overcome some of the most difficult problems that occur in the design and implementation of modern flight control systems (FCS). An ongoing comprehensive program to develop and assess this technology for Naval FCS applications was discussed. Currently, this program is focused on the development of a neural network FCS design tool, a neural network flight control computer emulator, a hybrid neural network/fuzzy logic automatic aircraft carrier landing system, and a neural network flight control configuration management system. For the neural flight control design tool, some initial favorable results were obtained. >

M. Steinberg

Papers

Applying neural network technology to future generation military flight control systems