Fuzzy logic and image based autonomous navigation

Disponível somente no TrabalhosFeitos
  • Páginas : 16 (3965 palavras )
  • Download(s) : 0
  • Publicado : 28 de agosto de 2012
Ler documento completo
Amostra do texto
Fuzzy Logic and Image based Autonomous Navigation
Ana Paula Abrantes de Castro, Leandro Toss Hoffmann, José Demisio Simões da Silva 1 Instituto Nacional de Pesquisas Espaciais – INPE, Laboratório de Computação e Matemática Aplicada – LAC, 12227-970 São José dos Campos, Brasil.
{apaula, hoffmann, demisio}@lac.inpe.br

Abstract. This paper presents a computational model for adaptive autonomousnavigation based on visual information from the environment and fuzzy logic control as a continuation of the initial developed in Castro et al [1]. A robot moves on a track environment from which it acquires images with the necessary information to guide itself as to the direction and speed to follow from each track position, based on a fuzzy logic decision system. The experiments were conductedin a controlled environment, but the positions of the robot varied depending upon the initial position on the track. The fuzzy logic system made decision based on the acquired track information related to the left and right strips of the track. An automaton based operator determined high contrast regions where the strips were located, leading to the angular direction of the strips, which consistedon the primary navigation information. Both direction and speed were inferred by the fuzzy logic system. In the paper, the performance of the robot navigation is shown. Keywords: autonomous navigation, fuzzy logic control, computer vision



Autonomous navigation is a well-studied topic in Artificial Intelligence [2][3][4], from which different paradigms try to approach theproblem using different sensors to recover environmental information that may lead to a safe and efficient navigation. The autonomous navigation task is related to the ability of a vehicle to move itself without human interaction and reach a goal, in a known or unknown environment. The robot is guided by on-line information during navigation. These tasks require different abilities to reach a point,to react to real time non-determined and unpredictable track situations, to construct and maintain a map of the environment, to determine the robot's position in the map, to create object oriented navigation plans, and to adapt to environmental changes, thus, autonomous navigation is a multidisciplinary area. The different autonomous navigation architectures may be categorized as: hierarchical,at a high-level the models and plans are considered and at low-level action detection and execution are considered; behavioral based, in which complex sequences of actions may be found by combining several simple units; and hybrid ar-

chitectures combining layered organizations and their execution performances. The implemented systems usually use a set of sensors to extract information such as,object position, direction to follow, and/or exiting objects in the environment, thus requiring different knowledge and technologies. However, in designing an autonomous navigation system that resembles the human navigation behavior, under artificial intelligence paradigms, it is necessary to strongly consider the use of information provided by images since visual cues form the fundamentalprimary information for humans to navigate in a known or unknown real world environment. Image based autonomous navigation are found in the literature [1][3][6][7][8]. In general, the acquired images are submitted to computer vision operators to provide 3D real world information from 2D data (images) [9]. This paper presents further developments of a computational model for image based adaptiveautonomous navigation with fuzzy logic control inferences [1]. A robot acquires visual information while it moves in an environment (a track). The fuzzy logic based system automatically corrects its trajectory by determining the direction and speed to be followed from the current position of the robot, based on the directions of the horizontal strips on the track provided by the images of the track,...
tracking img