Presents the application of the minimal radial basis function neural networks called minimal resource allocation neural networks (MRAN) for fault-tolerant flight controller design. Based on a study of different architectures for neural control, a simple architecture in which the MRAN controller is aiding a conventional controller is proposed. The main advantage in this scheme is that it requires no off-line training for the neural network and the scheme has good fault tolerant capabilities. The MRAN controller is illustrated for a F8 fighter aircraft longitudinal control in an autopilot mode for following velocity and pitch rate pilot commands under large parameter variations and sudden variations in actuator time constants. Results indicate that MRAN controller exhibits better performance than an earlier suggested feed forward inverse neural controller using gradient learning scheme.
Fault tolerant flight controller using minimal resource allocating neural networks (MRAN)
American Control Conference, 1999 ; 4 ; 2605-2609
1999
5 Seiten, 5 Quellen
Conference paper
English
Active engine mounting controller using extended minimal resource allocating networks
Tema Archive | 2008
|Active engine mounting controller using extended minimal resource allocating networks
Automotive engineering | 2008
|A Fault-tolerant Flight Controller Design for Actuator Faults
British Library Conference Proceedings | 2001
|Fault-Tolerant Flight Computer
Online Contents | 1996