A Neural Adaptive Controller in Flapping Flight

In this paper, we propose a neural adaptive controller for attitude control in a flapping-wing insect model. The model is nonlinear and subjected to periodic force/torque generated by nominal wing kinematics. Two sets of model parameters are obtained from the fruit fly Drosophila melanogaster and the honey bee Apis mellifera. Attitude control is achieved by modifying the wing kinematics on a stroke-by-stroke basis. The controller is based on filtered-error with neural network models approximating system nonlinearities. Lyapunov-based stability analysis shows the asymptotic convergence of system outputs. We present simulation results for angular position stabilization and trajectory tracking. Trajectory tracking is illustrated by two cases: saccadic turning and sinusoidal variation in the yaw angle. The proposed controller successfully regulates flight orientation – roll, pitch and yaw angles – by generating desired torque resulting from tuning parameterized wing motion. Results furthermore show similarities between simulated and observed turning from real insects, suggesting some inherent properties in insect flight dynamics and control. The proposed controller has potential applications in future flapping-wing Micro Air Vehicles (MAVs).