Synthesis of Artificial Network Based Flight Controller Using Genetic Algorithms

Abstract

Quadcopters are versatile unmanned aerial vehicles that operate in a variety of environments. In this paper we present the results of creating aflight controller for quadcopters based on an artificial neural network trained with genetic algorithms. The controller learns to stabilize a model of the craft and finds an optimal solution based on the custom scoring function. Experiments show better performance compared to manually tuned PID controllers, as well as faster convergence compared to traditional backpropagation learning methods.