Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
METAHEURISTICS APPLIED TO A FUZZY-CONTROLLED VEHICLE NAVIGATION OPTIMIZATION
Submission Author:
Luiza Scapinello Aquino , PR
Co-Authors:
Luiza Scapinello Aquino, Yan Lieven, José Henrique Kleinübing Larcher, Viviana Mariani, Leandro dos Santos Coelho
Presenter: José Henrique Kleinübing Larcher
doi://10.26678/ABCM.COBEM2021.COB2021-0029
Abstract
In this paper, a metaheuristic approach is used in the optimization process of a fuzzy control method for a vehicle navigation simulation. The core objective is, given a map containing obstacles with different sizes and shapes, to guide the vehicle to a set target. Since the only data available during the course is the distance from the nearest obstacles and the angle to the destination, it is only possible to control the vehicle’s velocity in each possible direction. Unlike more simple control techniques, fuzzy logic controllers (FLCs) allow to design a relevant control structure, while at the same time having great efficiency. This is due to the fact that fuzzy logic is capable of representing information from complex or uncertain contexts, like non-linear applications such as the one in this work. However, a deficiency in FLCs is the fact that it is a difficult task to tune the hyperparameters without previous experience in the manipulation of the model. The optimization of the FLC parameters can be complex and time-consuming due to the non-linearities present in the model. Evolutionary algorithms and swarm intelligence paradigms are metaheuristic approaches that can be appropriate to overcome this lack of model knowledge, since they are well known for diversifying the range of solutions, including appropriate solution in the search space, as well as having the ability to use available information from the solution candidates to intensify a search. The contribution of this paper is to compare the performance of three metaheuristics including Particle Swarm Optimization, Differential Evolution and Jaya Optimization to improve the fuzzy logic controller parameters of a vehicle’s navigation. The obtained results were promising in terms of the performance indexes in the simulations, they also reveal that all three metaheuristics are capable of achieving favorable performances for the off-line tuning of the FLC.
Keywords
Fuzzy Logic Control, Particle Swarm Optimization, Differential Evolution, Jaya Optimization, Vehicle Navigation
