Fly Simulation

Authors:

Aleksandar Rašković, Gymnasium, Subotica

Tadej Ristić, Computer Gymnasium, Belgrade

Mentor:

Mladen Bašić, School of Electrical Engineering, University of Belgrade

Abstract

This paper presents an implementation of a physical simulation of a fly’s flight aimed at learning stable flying behavior using artificial intelligence. The fly model includes kinematic and dynamic simulation in three-dimensional space, with emphasis on realistic wing rotation and air drag force calculation. The wing movements are controlled by a neural network whose parameters are optimized using a genetic algorithm. By applying different definitions of the fitness function, the influence of training criteria on flight stability and efficiency was analyzed. Experimental results show that the model successfully learns realistic wing-flapping motion, maintains altitude, and reduces undesired lateral deviations.

Full Paper