Jesus Savage

Followers

Following

Public Views

Interests

Uploads

Papers by Jesus Savage

Feature detection using Hidden Markov Models for 3D-visual recognition

2019 IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC)

In this work, we present a novel implementation for visual recognition using probabilistic models... more In this work, we present a novel implementation for visual recognition using probabilistic models. Given a scene view, we first propose a 3D feature extraction from a point cloud as a series of observations for a Hidden Markov Model; then, we evaluate the Profile HMM in the place recognition task using a publicly available dataset. Furthermore, we evaluated a classical HMM in the object recognition task in the context of anthropomorphic service robots. Results show that our approach performs well in the aforementioned tasks with high recognition rates.

Generating Reactive Robots’ Behaviors using Genetic Algorithms

Proceedings of the 13th International Conference on Agents and Artificial Intelligence, 2021

In this paper, we analize and benchmark three genetically-evolved reactive obstacle-avoidance beh... more In this paper, we analize and benchmark three genetically-evolved reactive obstacle-avoidance behaviors for mobile robots. We buit these behaviors with an optimization process using genetic algorithms to find the one allowing a mobile robot to best reactively avoid obstacles while moving towards its destination. We compare three approaches, the first one is a standard method based on potential fields, the second one uses on finite state machines (FSM), and the last one relies on HMM-based probabilistic finite state machines (PFSM). We trained the behaviors in simulated environments to obtain the optimizated behaviors and compared them to show that the evolved FSM approach outperforms the other two techniques.

Download

A SLAM system based on Hidden Markov Models

Informatics and Automation, 2021

We present a graph SLAM system based on Hidden Markov Models (HMM) where the sensor readings are ... more We present a graph SLAM system based on Hidden Markov Models (HMM) where the sensor readings are represented with different symbols using a number of clustering techniques; then, the symbols are fused as a single prediction, to improve the accuracy rate, using a Dual HMM. Our system’s versatility allows to work with different types of sensors or fusion of sensors, and to implement, either active or passive, graph SLAM. The Toyota HSR (Human Support Robot) robot was used to generate the data set in both real and simulated competition environments. We tested our system in the kidnapped robot problem by training a representation, improving it online, and, finally, solving the SLAM problem.

Download

Mass Estimation in Manipulation Tasks of Domestic Service Robots using Fault Reconstruction Techniques

ArXiv, 2020

Manipulation is a key capability in domestic service robots, as can be seen in the rulebooks of l... more Manipulation is a key capability in domestic service robots, as can be seen in the rulebooks of last Robocup@Home editions. Currently, object recognition is performed based mostly on visual information. Some robots use also 3D information such as point clouds or laser scans but, to the knowledge of authors, robots don't use physical properties to improve object recognition. Estimation of an object's weight during a manipulation task is something new in the @Home league and such ability can improve performance of domestic service robots. In this work we propose to estimate the weight of the grasped object using Sliding Mode Observers. If we consider the manipulator without load as the nominal system and object's weight as a fault signal, we can estimate such weight by an appropriate filtering of the output error injection term of the sliding mode observer. To implement our proposal we used MATLAB and Simulink Robotics System Toolbox, ROS Toolbox and Simscape. To improve c...

Download

Recognition of Static Gestures using Three-Dimensional Chain Codes using Dominant Direction Vectors

This paper presents a novel method for static gesture recognition based on three-dimensional chai... more This paper presents a novel method for static gesture recognition based on three-dimensional chain codes which are computed from three-dimensional skeletons acquired by three-dimensional vision sensors, such as Microsoft Kinect. The method has two stages: a digitization stage and a recognition stage. The digitization stage is based on the orthogonal direction change chain code, which represents the changes of direction on segments of a three-dimensional curve which was fitted in a three-dimensional grid; these changes of direction are invariant to rotation and translation. The recognition stage is based in the detection of dominant changes of direction on segments of a three-dimensional curve which was fitted in a three-dimensional grid. The experiments for testing this method of static gesture recognition involved recording the pose of the arms of a subject who was standing at increasing distances and the body was oriented in frontal and three-quarters angles; the poses were matche...

Optimization of waypoint-guided potential field navigation using evolutionary algorithms

2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (IEEE Cat. No.04CH37566)

This paper describes a method for optimization of waypoint selection for potential field navigati... more This paper describes a method for optimization of waypoint selection for potential field navigation in autonomous robots. In the method presented here, a genetic algorithm (GA) is used for optimizing the potential field. The chromosome of each individual encodes parametrizations for the potential field generated by waypoints, obstacles, and goals. The waypoints themselves are obtained through a Voronoi tessellation of the environment in which the robot is operating. It is demonstrated that the algorithm allows a robot to navigate safely and efficiently through spaces with many obstacles, even in cases where these are placed in a strongly unfavorable way. Furthermore, the results from simulations were implemented successfully in an actual Khepera robot. Using a slightly simplified navigation procedure, in which the robot comes to a standstill between successive steps in the navigation, the Khepera robot managed to navigate through one of the most difficult environments used in the simulations. Finally, the paper briefly describes a different implementation of potential field navigation, in the path planning adaptation submodule of a more advanced simulated mobile robot (VirBot).

Download

Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications

Lecture Notes in Computer Science, 2013

The use of general descriptive names, registered names, trademarks, service marks, etc. in this p... more The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein.

Download

Multiview Object and View Sequence Recognition using Hidden Markov Models

2022 IEEE 18th International Conference on Automation Science and Engineering (CASE)

Combination of Two Methods for Isolated Word Speech Recognition

We present a method for isolated word speech recognition combining Hidden Markov Models (HMM) and... more We present a method for isolated word speech recognition combining Hidden Markov Models (HMM) and Multi Section Vector Quantization (MSVQ) Code Books techniques. The use of these two techniques together enables the system to correct half of the wrong decisions made when only one technique is used. This system was implemented on a digital signal microprocessor (DSP), and is operating in real time. This system will be used in a Virtual Reality environment. In the first part of the report we make an overview of the MSVQ and the HMM techniques. Then we describe the system components. At the end we report the results obtained. Vector Quantization In recent years Vector Quantization (VQ) techniques have been used for data compression. Speech coding using VQ is based on linear prediction coding (LPC). The vector quantizer is a set of LPC vectors that represents the shape of the signals to be encoded. The LPC parameters used are the inverse filter gain squared oe 2 and the linear predictive...

Download

Intelligent flat-and-textureless object manipulation in Service Robots

This work introduces our approach to the flat and textureless object grasping problem. In particu... more This work introduces our approach to the flat and textureless object grasping problem. In particular, we address the tableware and cutlery manipulation problem where a service robot has to clean up a table. Our solution integrates colour and 2D and 3D geometry information to describe objects, and this information is given to the robot action planner to find the best grasping trajectory depending on the object class. Furthermore, we use visual feedback as a verification step to determine if the grasping process has successfully occurred. We evaluate our approach in both an open and a standard service robot platform following the RoboCup@Home international tournament regulations.

Download

Teleoperated Service Robot with an Immersive Mixed Reality Interface

Informatics and Automation

Teleoperated service robots can perform more complex and precise tasks as they combine robot skil... more Teleoperated service robots can perform more complex and precise tasks as they combine robot skills and human expertise. Communication between the operator and the robot is essential for remote operation and strongly affects system efficiency. Immersive interfaces are being used to enhance teleoperation experience. However, latency or time delay can impair the performance of the robot operation. Since remote visualization involves transmitting a large amount of video data, the challenge is to decrease communication instability. Then, an efficient teleoperation system must have a suitable operation interface capable of visualizing the remote environment, controlling the robot, and having a fast response time. This work presents the development of a service robot teleoperation system with an immersive mixed reality operation interface where the operator can visualize the real remote environment or a virtual 3D environment representing it. The virtual environment aims to reduce the lat...

Download

Knowledge Acquisition and Automatic Generation of Rules for the Inference Machine CLIPS

Abstract. A hierarchical representation of objects is dynamically generated from the input of a v... more Abstract. A hierarchical representation of objects is dynamically generated from the input of a virtual vision system. It is used to analyze a sequence of actions and extract behavior rules that can be utilized by the inference machine CLIPS. The vision system is assumed to provide simplified positional and shape information about visible 3D silhouettes in a frame per frame basis. A virtual agent, attempts to keep track of every image, without any previous knowledge about the object it represents. The hierarchy is restructured as necessary, to include new perceived images, in such a way that it also reflects factual relationships amongst them. Modifications between consecutive frames are internally interpreted and represented as functions which take the original world description and transform it into the next frame. A partial order is defined while looking for the satisfaction of domain/codomain requirements in functions composition, thus leading to the CLIPS rules. 1

Animal-Machine Interfaces

| During the last decades too much e ort have been done to develop human-machine interfaces that ... more | During the last decades too much e ort have been done to develop human-machine interfaces that has lead to several devices that help humans to use them. Our main motivation for this project is that those traditional human-machine interfaces and their advantages can be extended to other living beings. In order to provide them comfort and also to enhance human-animal communication. In this paper we present the results of the rst two-years of a research project named UNAM-CAN, which is a joint e ort of the Engineering and Zoological faculties at UNAM University. The project involves a service dog that carries a computer (Wearable Computer), that gives simple commands through a speaker. The main idea is that a complex task can be decomposed into simple sequential tasks that the dog can execute. One of the main problems to solve is that the dog needs to perform the requested tasks without the presence of a person. We did some experiments that shows that this is feasible. I.

ABSTRACT Virtual Reality for Archeological Maya Cities

important ancient Maya metropolis set deep in the tropical forest of a National Reserve called Ca... more important ancient Maya metropolis set deep in the tropical forest of a National Reserve called Calakmul’s Biosphere. Partially covered by the untouchable forest, the buildings stand almost hidden from each other, making it very difficult for a large number of visitors to grasp the architecture and length of the site. The solitude of this distant archeological site represents an additional problem for the safekeeping of its priceless findings, so none of them are exhibited and no museum in site has been settled on yet. The participants have foreseen the building of a site museum with virtual reproductions, to alleviate the difficulty the visitors face for appreciating the urban design of the city as well as the impossibility of exhibition of its priceless findings. In order to accomplish a presentation for the better understanding of Calakmul’s culture, the participants will use Virtual Reality (VR), Augmented Reality (AR) and VRML interactive 3D objects.

Comparison of Two Objects Classification Techniques using Hidden Markov Models and Convolutional Neural Networks

Informatics and Automation

This paper presents a comparison between discrete Hidden Markov Models and Convolutional Neural N... more This paper presents a comparison between discrete Hidden Markov Models and Convolutional Neural Networks for the image classification task. By fragmenting an image into sections, it is feasible to obtain vectors that represent visual features locally, but if a spatial sequence is established in a fixed way, it is possible to represent an image as a sequence of vectors. Using clustering techniques, we obtain an alphabet from said vectors and then symbol sequences are constructed to obtain a statistical model that represents a class of images. Hidden Markov Models, combined with quantization methods, can treat noise and distortions in observations for computer vision problems such as the classification of images with lighting and perspective changes.We have tested architectures based on three, six and nine hidden states favoring the detection speed and low memory usage. Also, two types of ensemble models were tested. We evaluated the precision of the proposed methods using a public do...

Download

P o S ( A I S I S 2 0 1 9 ) 0 4 3 Robotics , AI and Machine Vision

This paper presents a compilation of the accumulated development in the Bio-Robotics laboratory o... more This paper presents a compilation of the accumulated development in the Bio-Robotics laboratory of the Engineering Faculty at UNAM. The techniques used to build robotic service systems are briefly mentioned. In particular, a line of research has been developed using most of them interdependently in what is called in a hybrid system ViRBot (VIrtual and Real roBOt sysTem); a system to operate mobile robots. ViRBot is an abstraction system integrated into four specialized layers; Input, planning, knowledge management, and execution to provide intelligence to a mechatronic agent to execute the service in domestic environments. This project has been developed incrementally with the participation of students and profesors. Currently, the ViRBot system inter-operates with ROS and together, they form an intelligent systems development platform for computer vision, digital signal processing, automatic planning, automatic control and human-robot interaction.

Download

Advanced Robotics (ICAR), 2013 16th International Conference on

Progress in Pattern Recognition, Image Analysis and Applications (CIARP)

Download

Robot Soccer World Cup XV

Pumas@Home 2012 Team Description Paper

This paper describes an autonomous service robot called Justina built by the team Pumas in the la... more This paper describes an autonomous service robot called Justina built by the team Pumas in the laboratory of Biorobotics at the National University of Mexico. The robot is based on the ViRbot architecture for autonomous mobile robots operation. ViRbot defines a human interaction interface based on Natural Language Processing for understanding voice and gesture commands. After this, a Conceptual De- pendence (CD) representation is generated, and an action planner infers a set of rules to perform the commands based on the sensors information and the state of the robot. The aforementioned rules become subtasks that are performed by different submodules that control the hardware and share information by TCP/IP sockets through a Blackboard in or- der to achieve the given command. By using a good representation of the problem domain through CD's and a rule-based system, such as an inference engine, the operation of the robot becomes a more tractable problem and easier to implement.

Jesus Savage

Uploads

Papers by Jesus Savage

Log In