Self-Organizing Radial Basis Function Networks for Adaptive Flight Control

摘要

The performance of nonlinear flight-control algorithms, such as feedback linearization and dynamic inversion, isnheavily dependent on the fidelity of the dynamic model being inverted. Incomplete or incorrect knowledge of thenaircraft dynamics results in reduced performance and may lead to instability. A self-organizing parametrizationnstructure is developed to augment the baseline dynamic inversion controller for a high-performance militarynaircraft. This algorithmis proven to be stable and can guarantee arbitrary tracking error performance. The trainingnalgorithm to grow the network and adapt the parameters is derived from Lyapunov theory. In addition to growingnthe network of basis functions, a pruning strategy is incorporated to keep the size of the network as small as possible.nThe controller is simulated for different situations, including control surface failures, modeling errors, and externalndisturbances. A performance measure of maximum tracking error is specified for the controllers a priori. Excellentntracking errorminimization to a prespecified level using the adaptive component is achieved.The performance of thenself-organizing radial-basis-function network-based controller is also compared with a fixed radial-basis-functionnnetwork-based adaptive controller.While the fixed radial-basis-function network-based controller,which is tuned toncompensate for control surface failures, fails to achieve the same performance under modeling uncertainty andndisturbances, the self-organizing radial-basis-function network is able to achieve good tracking convergence undernall specified error conditions.