Eventos Anais de eventos
DINAME2019
DINAME2019
Quadrotor Black-Box System Identification using Metaheuristics
Submission Author:
Ewerton Cristhian Lima de Oliveira , PA
Co-Authors:
Ewerton Cristhian Lima de Oliveira, Jasmine Araujo, Orlando Fonseca Silva, Antonio Silveira, Juan Ferreira Vidal, Anderson de França Silva
Presenter: Anderson de França Silva
doi://10.26678/ABCM.DINAME2019.DIN2019-0078
Abstract
The complexity of nonlinear and multivariable dynamic systems, such as Unmanned Aerial Vehicle (UAV), have been aim of several researches, mainly directed to identification model and controller design, because the drones have several applications in trade, industry, military, etc. However, the low error and high precision in identification of nonlinear multivariable systems is a great challenge for traditional techniques. Therefore, this paper has the objective to present the applicability of metaheuristics in quadrotor systems identification, further a comparative study of three metaheuristics, Particle Swarm Optimization (PSO), Adaptive Particle Swarm Optimization (APSO) and Cuckoo Search (CS), in black-box system identification from real data of an unmanned quadrotor model AR.Drone 2.0, where was used four uncoupled NARX structs, and the performance of each metaheuristic is evaluated according to mean squared error (MSE), precision and average processing time (APT) after 30 simulations in parameter estimation using Matlab software. The results show that PSO had better performance in precision and APT, and CS reached minor MSE.
Keywords
parameter estimation, drone, PSO, APSO, Cuckoo Search
