Optimization of Gain Tuning in Quadrotors for Attitude Stability using Best Mean Random Metaheuristic

Abstract

Drones are a type of Unmanned Aerial Vehicles (UAV) come under multicopters. UAV, in certain applications require low altitude and short flight distance characteristics. In such instances, quadrotors are most suitable than the fixed wing UAVs. Quadrotors are controlled by the stabilization process in real time. It requires sensors including accelerometers, gyroscope, altitude sensors to estimate the air frame position and velocity. For efficient flight control, different variables affect the output characteristics. Quadrotors stable control is a main concern in all the applications. Attitude restoration in quadrotors should be done by control techniques. Usually, PID Controllers are designed with respect to the flight motion by providing trail value of the gain constants. In the present work, optimal values of control gains are obtained using surrogate optimization strategy with the use of neural networks and best mean random metaheuristic optimization. The objective is to minimise the settling time in roll motion of a quadrotor during the hovering. Initially, by varying the Kp, Kd & Ki (gain) values over a range, the controlled time responses of trajectory are recorded and the corresponding amplitude and settling time are tabulated. The set of values are trained using Radial Basis Function (RBF) neural network to evaluate the settling time for a given control gains Kp, Kd & Ki. Further, a newly developed BMR optimization algorithm is employed along with trained neural network as a surrogate model for estimating the function evaluations (settling time values). For a given hovering path, settling time in roll direction is minimized by proper selection of control gain parameters. The results are presented with a simulated data of the quadrotor.