More and more UAVs are developed for various purposes and their flight controllers are required to have sufficient robustness and performance to realize their versatile missions. To design these sophisticated controller is pretty much time-consuming task by traditional design method. Neural network based adaptive control with dynamic inversion is applied to solve this problem. This method has two advantages. One is that the gain scheduling is not necessary because nonlinear dynamic inversion is applied to control nonlinear systems. The other is that neural network improves the controller performance by estimating parameters of the actual plant. Numerical examples show its effectiveness and its ability to adapt to modeling errors. This paper concludes that proposed method reduces the workload of controller design task and it has ability to adapt various errors of nonlinear systems.
[1]
Michael I. Jordan,et al.
Forward Models: Supervised Learning with a Distal Teacher
,
1992,
Cogn. Sci..
[2]
Anthony J. Calise,et al.
Adaptive Autopilot Designs for an Unmanned Aerial Vehicle
,
2005
.
[3]
Hirokazu Suzuki,et al.
Evaluation of Guidance and Control Systems of a Balloon-Launched Drop-Test Vehicle
,
2006
.
[4]
Toshikazu Motoda,et al.
Longitudinal Landing Control Law for an Autonomous Reentry Vehicle
,
1999
.
[5]
Anthony J. Calise,et al.
Neural-Network Augmentation of Existing Linear Controllers
,
2001
.