Improved Fault Tolerance for Autolanding Using Adaptive Backstepping Neural Controller

This paper presents a neural-aided controller that enhances the fault tolerant capabilities of a high performance fighter aircraft during the landing phase when subjected to severe winds and failures such as stuck control surfaces. The neural network scheme herein is called adaptive backstepping neural controller (ABNC). In this study we have combined ABNC along with the classical controller to enhance the failure tolerance of the latter using the feedback error-learning paradigm due to Gomi and Kawato. The ABNC controller uses radial basis function neural networks with on-line learning without prior training. Information about actuator failures is not available to the controller for use in reconfiguration. The baseline controller required for the feedback error learning has been designed using a classical design approach to achieve the desired autonomous landing profile with tight touchdown dispersions called therein as the pillbox. The baseline design is capable of meeting touchdown requirements in severe winds but is not specifically designed for failures. The performance of the classical design augmented with the ABNC based neural controller is studied in detail for predefined failure scenarios. The failures considered in this study are: i. Single faults of either aileron or elevator stuck at certain deflections and ii. Combination fault for both one aileron and one elevator stuck at different deflections. A hard over failure of particular interest in this study is that of both the ailerons being stuck at particular deflections. Simulation studies indicate that the neural controller aids the baseline controller, significantly enhancing the fault-tolerance envelope.