Octorotor flight control system design with stochastic optimal tuning, deep learning and differential morphing

In this paper, simultaneous longitudinal and lateral flight control is investigated for an octorotor by using stochastic optimal tuning and deep learning under differential morphing. Octorotor models for differential morphing were drawn in SOLIDWORKS drawing program. Arm lengths are randomly estimated in the algorithm. Moments of inertia changing according to morphing ratios are estimated with deep neural network. In addition, the proportional–integral–derivative controller coefficients required for both longitudinal and lateral flight according to the morphing ratios are estimated by simultaneous perturbation stochastic approximation. Considering the design performance criteria, 49.95% improvement was achieved in the total cost. The estimation of unknown parameters by optimization method and deep learning was tested in simulations, and the octorotor successfully followed the given reference angle.