Roman Osorio

SVM and ANN Application to Multivariate Pattern Recognition Using Scatter Data

Several methods of Statistical Process Control (SPC) are used to analyze process measurements with the purpose to detect faults that affect the process stability. SPC has a major drawback because it indicates the presence of faults without explaining which ones and where are the faults. In practical applications, SPC just analyses univariate signals limiting the study of multiple measures. Nowadays, novel methods have been developed for fault analysis based on pattern recognition in control charts. However, the majority of these studies follow a univariate approach. This article proposes a multivariate pattern recognition approach using machine learning algorithms in conjunction with a scatter diagram as the proposed method. In particular the aim of this approach is to monitor quality characteristics of a product in a multivariate environment considering states in control and out of control without the constraints of statistical conditions with the possibility of its application in real time. Results using Support Vector Machines (SVM) and the FuzzyARTMAP neural network showed that multivariate patterns can be recognized successfully in 81% of the cases.

Integration of Algorithms for Maps Construction and Simultaneous Localization in a Mobile Robot

In this paper it is presented an integration of algorithms that permits maps construction and navigation of mobile robots. Simultaneous Localization and Mapping (SLAM) algorithm is used based on FastSLAM method. Navigation system is based on VHF to avoid obstacle and a spiral way trajectory method. To three different complex simulation maps are used to evaluate the system.

Obstacle Avoidance Behaviors for Mobile Robots Using Genetic Algorithms and Recurrent Neural Networks

Abstract This paper discusses how to generate mobile robots’ behaviors using genetic algorithms (GA). The behaviors are built using state machines implemented in recurrent neural networks (RNN), controlling the movements of a humanoid mobile robot. The weights of the RNN are found using a GA, these are evaluated according to a fitness function that grades their performance. Basically, this function evaluates the robots performance when it goes from an origin to a destination, and the grading of the robot evaluates also that the robots behavior using RNN is similar to the behavior generated by a potential fields approach for navigation. Our objective was to prove that GA is a good option as a method for finding behaviors for mobile robots’ navigation and also that these behaviors can be implemented using RNN.

Using Background and Segmentation Algorithms Applied in Mobile Robots

Abstract In this article a segmentation algorithm for detecting moving objects is presented. The aim of the research is to integrate the algorithm in applications such as car parking video surveillance systems. One of the techniques used in this paper to detect motion in a sequence of images is the use of the background model, which is widely used. The technique allows to detect which objects are moving (without identification) which is the first stage for further processing in tasks such as tracking and object recognition. The results from the segmentation algorithm using several parameters are presented that validate the approach.

Fuzzy Logic for Omnidirectional Mobile Platform Control Based in FPGA and Bluetooth Communication

An Omni directional mobile platform control using the intelligence technique of Fuzzy Logic is showed in the article, the control allows a practical and reliable driving control of 4 Omni directional wheels, implemented in FPGA allowing to have an independent and autonomous single chip system out of a central computer dependence in order to be used with different applications like service robots platforms. An additional feature is performed by using Bluetooth communication with a cellular phone based on a smartphone OS Android as the handset control device. Driving movement for the mobile platform is limited for 8 directions, a Fuzzy Logic module controls the travelling of the platform with independent movement for each wheel, physical feedback is implemented by using electronic decoders.

Surveillance System Mobile Object Using Segmentation Algorithms

In this paper a segmentation algorithm is used to detect moving objects and to integrate it to a supervision and surveillance systems, in a parking lot, as a first step. One of the way to moving detection in image sequences is the moving object segmentation by background model, very well-known technique that it permit to know what objects are moving. This can be employed, in the second stage, to identify and to follow objects.

Movement and Color Detection of a Dynamic Object: An application to a Mobile Robot

The paper describes the integration of several image processing algorithms necessary to recognize a particular color and the movement of an object. The main objective is to detect the object by its color and track it by a mobile robot. Mean filter is applied to soften and sharpen the input image. Then, RGB filter is applied to calculate the center of mass and area of the object and to locate its position in a real environment to develop the robot motion. These algorithms are applied to a mobile robot, in a tested scenario, tracking an object.

Real Time Object Recognition Methodology

This paper shows a methodology for on-line recognition and classification of pieces in robotic assembly tasks and its application into an intelligent manufacturing cell. The performance of industrial robots working in unstructured environments can be improved using visual perception and learning techniques The object recognition is accomplished using a neuronal network with FuzzyARTMAP architecture for learning and recognition purposes, which receives a descriptor vector called CFD&POSE as the input. This vector represents an innovative methodology for classification and identification of pieces in robotic tasks, every single stage of the methodology, is described step by step and the proposed algorithms explained. The vector compresses 3D object data from assembly parts and is invariant to scale, rotation and orientation. The approach in combination with the fast learning capability of ART networks indicates the suitability for industrial robot applications as it is shown in experimental results and the possibility to add concatenated information into the descriptor vector to achieve a much more robust methodology.

A learning approach for on-line robotic assembly tasks

Abstract The acquisition of assembly skills by robots is greatly supported by the efective use of contact force sensing and object recognition vision systems. In this paper, we describe the ability to invariantly recognize assembly parts at different scale, rotation and orientation within the work space. The paper shows a methodology for on-line recognition and classification of pieces in robotic assembly tasks and its application into an intelligent manufacturing cell. The performance of industrial robots working in unstructured environments can be improved using visual perception and learning techniques. In this sense, the described technique for object recognition is accomplished using an Artificial Neural Network (ANN) architecture which receives a descriptive vector called CFD&POSE as the input. This vector represents an innovative methodology for classification and identification of pieces in robotic tasks. The vector compresses 3D object data from assembly parts and it is invariant to scale, rotation and orientation, and it also supports a wide range of illumination levels. The approach in combination with the fast learning capability of ART networks indicates the suitability for industrial robot applications as it is demonstrated through experimental results.

Mapping visual behavior to robotic assembly tasks

This paper shows a methodology for on-line recognition and classification of pieces in robotic assembly tasks and its application into an intelligent manufacturing cell. The performance of industrial robots working in unstructured environments can be improved using visual perception and learning techniques. The object recognition is accomplished using an Artificial Neural Network (ANN) architecture which receives a descriptive vector called CFD&POSE as the input. This vector represents an innovative methodology for classification and identification of pieces in robotic tasks, every stage of the methodology is described and the proposed algorithms explained. The vector compresses 3D object data from assembly parts and it is invariant to scale, rotation and orientation, and it also supports a wide range of illumination levels. The approach in combination with the fast learning capability of ART networks indicates the suitability for industrial robot applications as it is demonstrated through experimental results.