Adriano Joaquim de Oliveira Cruz

Analysis of the performance of a fuzzy controller developed to guide a simulated robot

The main goal of this work is to compare two different configurations defined for a fuzzy controller that is used to guide a simulated robot through a virtual world populated with randomly placed obstacles. In the first configuration the system controls a single output, the rotation angle of the robot. In the second configuration there is an additional output that controls the robot step. In order to compare the performance of both controllers we studied the robot stability based on the removal of rules. We also measured two parameters: processing time and the amount of steps necessary to reach the final goal. The simplicity, easiness of design and robustness of both controllers suggests that for many control tasks, it is possible to create a fuzzy system that performs well under real conditions. The results show that an improved performance was obtained when the fuzzy module processed more variables from the environment, without compromising the simplicity of the design.

Supermarket Game: An Adaptive Intelligent Computer Game for Attention Deficit/Hyperactivity Disorder Diagnosis

The This work reports the research and the application of an adaptive intelligent computer game with a fuzzy decision system developed specifically to help in the diagnosis process of the Attention Deficit/Hyperactivity Disorder (ADHD). ADHD is a neuropathology, its main signal is the severe performance impairment of activities that involves executive functions: planning, volition, problem resolution and aim oriented works. The specification, implementation and other relevant aspects of The Supermarket Game are here discussed.

Analysis of the Stability of a Fuzzy Control System Developed to Control a Simulated Robot

This work presents a fuzzy system designed to guide a simple simulated robot through a virtual world populated with randomly placed obstacles. The environment simulates a situation where the robot has to move in a complex world and has to calculate not only its direction but also its speed. The main goal was to investigate the performance of a fuzzy controller studying its sensitivity and robustness. The sensors used by the robot were very simple, so its knowledge of the environment was limited. The system was quickly engineered and proved to be very successful reaching the goal in all designed tests. The tests also proved that the controller is very robust. The simplicity, easiness of design and robustness of the controller, suggests that for many control tasks, it is possible to create a fuzzy system that can perform very well under real conditions

MULTIPLUS: A modular high-performance multiprocessor

Abstract The MULTIPLUS project which is currently under development at NCE/UFRJ, Brazil, aims at the study of parallel processing problems in MIMD environments. The project includes the development of a parallel shared-memory architecture and a UNIX[+]-like operating system called MULPLIX. The MULTIPLUS architecture uses an inverted n-cube multistage network to interconnect clusters of processing nodes designed around a double-bus system. As a consequence, the architecture is partitionable and modular. It can easily and efficiently support configurations ranging from workstations to powerful parallel supercomputers with up to 2048 processing nodes. The MULPLIX operating system provides MULTIPLUS with an efficient computing environment for parallel scientific applications. MULPLIX uses the concept of thread, implements busy-waiting synchronization primitives very efficiently and carefully considers data locality and scientific processing requirements in the policies adopted for memory management and thread scheduling.

ATM: a New Heuristic Algorithm Based on Genetic Algorithm and Betting Theory

This paper proposes a new heuristic algorithm for combinatorial optimization. This algorithm is based on genetic algorithm and betting theory. It applies probabilistic recombination to a solution given as input and returns a new solution much more optimized. Experiments using arrays from the library of the Travel Salesman Problem have been done and the obtained results have demonstrated that this algorithm is very efficient, mainly for dense arrays

New Solutions for Old Problems: Use of Interfaces Human/Computer to Assist People with Visual and/or Motor Impairment in the Use of DOSVOX and microFênix

The article describes two solutions to improve the interface to software tools developed at to Tercio Pacitti Institute – Federal University of Rio de Janeiro, for people with disabilities. The first is an interface that allows people with visual impairments who also have the inability to use hands, have access to DOSVOX system, one of the most used Assistive Technologies in Brazil. DOSVOX enables the interaction between the visually impaired and the computer, whose operation is originally performed solely by keyboard, now can be operated via a gyroscope coupled to the user’s head, whose movement is translated into a series of keystrokes. The other interface is a BCI – Brain Computer Interface to microFenix, software designed for helping people with severe motor impairment to interact with a microcomputer.

An new model for robot navigation based on AprioriAll algorithm

This paper proposes a new model for robot navigation in dynamic environments. The robot learns the motions of the objects, does motion ahead predictions, and plans a collision-free path to the goal. In this paper two new predicting algorithms, based on the AprioriAll algorithm, and a planning algorithm, using a heuristic search, are proposed. A novel framework and a simulation environment were used to validate the model. The results have shown that the robot succeeds in reaching the goal via collision-free paths at an average success of about 94%.

Diesel engines diagnosis through analysis of lubricating oil

Expert systems can be applied to solve complex problems in different areas of knowledge. It have been used by specialists to help diagnose problems in areas such as medicine and engineering. More recently techniques from computational intelligence area, such as neural networks and fuzzy logic were used to build expert systems. These systems can learn from examples provided by problems solved in the past. The success of the expert system depends on how good the model is defined. This paper presents the modeling of an expert system used to classify the internal combustion diesel about the level of corrosion of its metal parts on a three-level scale. The system bases its decision on the results of used lubricating oil analysis. It uses a neuro-fuzzy model, called ANFIS, trained from past diagnoses performed by specialists. The paper discusses how important parameters of the system were defined, such as which variables obtained from the analysis of the oil should be used, the number of training epochs and the number of membership functions used to map the inputs.

Avaliação Exploratória de Conceito do Sistema NÉBULA em Jogos Psicopedagógicos

Uma das muitas aplicacoes de jogos eletronicos, que empregam tecnicas de inteligencia computacional no processamento de suas informacoes, constitui a area de jogos psicopedagogicos, virtualizados em sistemas informatizados. Nesse caso, os algoritmos de inteligencia computacional sao aplicados dentro de processos pedagogicos para analise, intervencao e reabilitacao cognitiva. No entanto, a modelagem de algoritmos matematicos envolvidos no processo de avaliacao cognitiva pode ser complexa para especialistas de outras areas. Este trabalho apresenta o estudo exploratorio de conceito de NEBULA, uma plataforma integrada para o desenvolvimento de sistemas de inferencias nebulosas de maneira simplificada e desacoplada da construcao dos jogos psicopedagogicos, permitindo que as equipes multidisciplinares participem ativamente da construcao da logica auxiliar a analise cognitiva.

A dynamic fluid level monitoring application using hough transform and edge enhancement

Fluid level monitoring is a common control task in chemical processes. The on-line level monitoring of fluids with different densities is a complex task to be accomplished using traditional sensors, as floats for example. Most of the classic sensors have parameters that need to be recalibrated several times when monitoring liquids with different densities. To accomplish the recalibration, the process must be frequently paused to suit the new liquid characteristics. Besides, some traditional level monitoring methods are not safe for use with flammable liquids. In this paper we propose a new density invariant and non contact approach for level monitoring that is based on image processing Hough transform and image normalization. A prototype was implemented and an experiment in laboratory was carried out to test the performance of the proposed model.