Flávio Neves

Autonomous navigation system using Event Driven-Fuzzy Cognitive Maps

This study developed an autonomous navigation system using Fuzzy Cognitive Maps (FCM). Fuzzy Cognitive Map is a tool that can model qualitative knowledge in a structured way through concepts and causal relationships. Its mathematical representation is based on graph theory. A new variant of FCM, named Event Driven-Fuzzy Cognitive Maps (ED-FCM), is proposed to model decision tasks and/or make inferences in autonomous navigation. The FCM’s arcs are updated from the occurrence of special events as dynamic obstacle detection. As a result, the developed model is able to represent the robot’s dynamic behavior in presence of environment changes. This model skill is achieved by adapting the FCM relationships among concepts. A reinforcement learning algorithm is also used to finely adjust the robot behavior. Some simulation results are discussed highlighting the ability of the autonomous robot to navigate among obstacles (navigation at unknown environment). A fuzzy based navigation system is used as a reference to evaluate the proposed autonomous navigation system performance.

Startup of a distillation column using intelligent control techniques

Abstract This work proposes the development of an intelligent predictive controller. Recurrent neural networks are used to identify the process, providing predictions about its behavior, based on control actions applied to the system. These information are then used by fuzzy controllers to accomplish a better control performance. Moreover, the fuzzy controller membership functions are evolved by Genetic algorithms (GAs) allowing an automatic tune of controllers. The combined use of these techniques make possible the control of multi-variable processes using several fuzzy controllers where the coupling among controlled variables are modeled by neural networks, and control objectives can be inserted into the GA fitness function. The methodology was applied to a simulation of the startup of a continuous distillation column. This process is chosen due to their characteristics, such as inertia, large accommodation time and conflicting control objectives that make these processes hard to control with traditional methods.

A dynamic fuzzy cognitive map applied to chemical process supervision

This work develops an intelligent tool based on fuzzy cognitive maps to supervisory process control. Fuzzy cognitive maps are a neuro-fuzzy methodology that can accurate model complexly system using a causal-effect fuzzy reasoning. In the proposed approach, new types of concept and relation, not restricted to cause-effect ones, are added to the model resulting in a dynamic fuzzy cognitive map (D-FCM). In this sense, a supervisory system is developed in order to control a fermentation process. This process has a non-linear behavior and presents several problems, such as non-minimum phase and large accommodation time. The supervisor goal is to operate the process in normal and critical conditions. The expert knowledge about the process behavior in both conditions is used to build the D-FCM supervisor. Simulation results are presented in order to validate the proposed intelligent supervisor.

Adhesion Force Control and Active Gravitational Compensation for Autonomous Inspection in LPG Storage Spheres

This paper presents a climbing robot based in wheel locomotion and magnetic adherence, a common mechanical topology applicable to a wide range of industrial tasks. The robot is applied to perform internal/external inspection in liquefied petroleum gas (LPG) storage spheres, hence, some severe operation features like adherence and force balance impose necessary conditions to the robot. In order to satisfy these conditions, a dynamic control system is developed in two steps, active gravitational compensation and adhesion force control.

Environment Identification and Path Planning for Autonomous NDT Inspection of Spherical Storage Tanks

This paper presents a novel approach to inspection planning in spherical storage tanks by an autonomous climbing robot. The objective is the automatic extraction of some environment characteristics, by robot, to predict the tank dimensions and robot localization. Three distinct perception sources (long range laser rangefinder, light detection and ranging, and depth camera) are used to predict a 3D occupancy grid wrapping calculated tank. From this grid, a path for tank inspection is computed that ensuring a complete icon at the entire tank surface. This scanning must consider kinematic constraints of magnetic wheels and NDT standard. The approach is evaluated in four LPGs spherical tanks virtually designed with same characteristics that real tank projects.

Scheduled Fuzzy Controllers for Omnidirectional Motion of an Autonomous Inspection Robot with Four Fully Steerable Magnetic Wheels

This work presents a scheduled fuzzy controllers for an autonomous inspection robot designed to work inside and outside spherical storage tanks. The robot is designed with four fully steerable magnetic wheels and a mechanical topology which promotes the correct adjustment of adhesion system. The proposed motion control works according with robots specific characteristics to ensure the quasi-omnidirectional motion over strong adhesion with tank surface and minimizes wheels kinematics constraints.

Optical-Ultrasonic Heterogeneous Sensor Based on Soft-Computing Models

A heterogeneous sensor system to determine the ethanol concentration in ethanol-water solution is demonstrated. The system consists of an optical-fiber refractometric transducer based on a long-period grating and a pair of ultrasonic transducers in transmission-reception mode, connected to a stand-alone electronic board for data acquisition, storage, and signal preprocessing. To implement a coherent sensor fusion from both measurement techniques, two soft computing methods (artificial neural network model and neuro-fuzzy model) are studied. A comparative analysis of these models was carried out based on the measured data. The best performance was obtained with the neural-network-based model. This model showed that it was able to correlate the responses of the optical-fiber transducer and the ultrasound system with the ethanol-water concentration. The final performance of the heterogeneous system is better within the whole range of concentrations, even if compared with the best performance of the individual sensors for limited ranges.

Using MOGA to order batches in a real world pipeline network

The scheduling of activities to distribute oil derivate products through a pipe network is a complex combinatorial problem that presents a hard computational solution. This problem could be decomposed on three sub problems according to the key elements of scheduling: assignment of resources, sequencing of activities, and determination of resource timing utilization by these activities. This work develops a model to the sequencing sub-problem. The main objective is to develop a multi-objective genetic algorithm to order oil derivate products batches input into the network. From the operational practice, the batches sequencing has great influence on the final scheduling result. The MOGA model provides a set of solutions that means different options of pipeline operations, in a small computational time. This work contributes to the development of a tool to aid the specialist to solve the batch sequencing problem, which reflects in a more efficient use of the pipeline network.

Quasi-Omnidirectional Fuzzy Control of a Climbing Robot for Inspection Tasks

This work presents a novel method to quasi-omnidirectional control of an intelligent inspection robot designed to work inside and outside spherical storage tanks. The main objective is to promote a stable and smooth navigation during inspection tasks, ensuring the safety motion under adhesion and kinematic constraints. The robot is designed with four independent steerable magnetic wheels and a mechanical topology that allows the correct adjustment of adhesion system. A scheduled Fuzzy control is developed to achieve an optimal behavior and maximize the robot’s maneuverability, considering the magnetic restrictions of adhesion system and kinematic constraints of the inspection robot. The high adaptability of its mechanical topology (i.e., wheel misalignment, magnetic adhesion system, wheel camber and flexibilities in mechanical structure) and gravitational disturbance introduce many nonlinear characteristics in dynamic behavior that cannot be neglected, making the determination of its dynamic model a complex task. The Fuzzy approach allows to project a control system without a depth knowledge of its dynamic properties, to minimize the dynamic disturbances found in robot structure. Thus, the proposed motion control works according to the robot specific characteristics to ensure the quasi-omnidirectional motion over a reliable adhesion to tank surface and to minimize the effects of wheels kinematic constraints.

Enhancing Robot Capabilities of Environmental Perception through Embedded GPU

The use of GPU in point cloud processing usually represents a gain in computation time more than ten times higher then on CPU, especially for large amounts of data. This paper brings an evaluation of the processing time for point cloud fusion in three different systems (PC, Nvidia TX1 and Nvidia TK1) using CPU and GPU. The objective is to find the best way to perform sensor fusion for an autonomous inspection robot. Four different scenarios were considered for the tests, with five different neighbor radius. A table with the processing time found for each experiment was created to allow a quick comparison. As results, a reduction in the number of points of up to 82.3 times was obtained, besides a time difference between CPU and GPU of up to 160 times.