Image & Robotics Network

Swarm Intelligence & Robot Teams

Juan Manuel Ahuactzin — 2010-05-22T15:45:41Z

Part 1 agent introduction
Part 2 application to inverse kinematics
Part 3 emergence
Part 4 spatial coordination
Part 5 reconfigurable robot

Swarm Intelligence & Robot Teams -

Evolutionary Computer Vision

Juan Manuel Ahuactzin — 2009-12-17T21:11:55Z

Evolutionary Computer Vision -

Service Robots

Juan Manuel Ahuactzin — 2009-12-17T20:57:45Z

Autonomous Vehicles

Juan Manuel Ahuactzin — 2009-12-17T19:16:07Z

Autonomous Vehicles (AVs) are automated vehicles to carry out a specific task, without direct human intervention. AVs are mobile robotic applications that have generated great interest in recent years due to their capacity to perform repetitive tasks in remote or harmful environments with extreme operating conditions. The applications and tasks of these devices vary from the transportation of material, to the exploration of planet's surfaces. AVs consist of selecting a vehicle originally designed for human drive, and installing the necessary components and systems to carry out the required tasks with autonomy. The Tutorial describes the Navigation system of AVs, composed by two components: the Trajectory Planning, computed off-line, defines the trajectory the AVs has to follow to reach its goal; and the Trajectory Following, executed on-line, allowing to follow the planned trajectory by optimizing some requirements like time or minimum error, while avoiding collisions. The Navigation system takes into account the physical properties of AVs, translated into non-holonomic constraints. At same time, it supposes the existence of control architecture, to structure the AV operation, including high level functions like planning, also with the command generation for sensors and actuators.

Autonomous Vehicles -

Robot Learning

Juan Manuel Ahuactzin — 2009-12-14T18:08:28Z

Eduardo Morales

National Institute for Astrophysiscs, Optics and Electronics, Tonantzintla, Puebla, Mexico.

Robot Learning is a currently very active research area as it promises to create more flexible, independent and autonomous robots. It also opens the possibility, for non-experts, to instruct a robot how to perform different tasks without the need of programming. In this short tutorial we will review some of the current main techniques used in robot learning and show some of their applications. In particular, we will focused this tutorial in reinforcement learning and programming by demonstration, but we will also review some techniques used in visual concept learning and other machine learning techniques that can be used in robotics.

Robot Learning -

Bayesian Perception

Juan Manuel Ahuactzin — 2009-12-03T22:40:54Z

Olivier Aycard

LIG Laboratory - Grenoble University - INRIA Grenoble, France

L. Enrique Sucar

Robotics Laboratory - Department of Computer Science - INAOE, Mexico

The goal of this lecture is to give an overview of Bayesian techniques and their applications to perception for mobile robots and intelligent vehicles. After a brief review of the fundamentals of probability and graph theory, we introduce the different Bayesian models & algorithms generally used in perception, including: Bayesian Classifiers, Bayesian Networks, Hidden Markov Models, Kalman Filters, and Particle Filters. In the second part we present some applications of these models & algorithms for perception solutions such as: gesture recognition, multi-object tracking, intelligent vehicles and medical applications.

MegaCrowd simulation

Juan Manuel Ahuactzin — 2009-12-03T21:32:04Z

Issac Rudomin

ITESM-CEM Estado de México, Mexico

We will discuss our Crowd simulation systems, and compare CPU and our GPU implementation in GLSL. We discuss issues such as behavior, collision detection and avoidance, level of detail, as well as our results, and limitations. We discuss how we are using CUDA instead of GLSL to overcome some of these limitations. Another issue is how much character variety is needed for realistic crowds, and how to genetrate facial and body variety.

MegaCrowd simulation -

Visual servoing

Juan Manuel Ahuactzin — 2009-11-24T15:10:00Z

Philippe Martinet

LASMEA-CNRS & IFMA, Campus des Cezeaux, 63177 Aubière, France

This lecture will present recent advances in visual servoing.

First we will present a global introduction to visual servoing. This will include some knowledge on geometrical vision (camera model, camera calibration, 3D reconstruction), some knowledge on sensor based control, the state of the art and bibliography issue, and some applications with video illustration.

Second we will proceed on modelling issue. We will address the case of 3D features and 2D features. If time allow it, I will make a demonstration using the visual servoing toolbox using matlab/simulink.

Nonholonomic motion planning

Juan Manuel Ahuactzin — 2009-11-17T16:48:44Z

Jean-Paul Laumond

LAAS-CNRS, Toulouse, France

This lecture introduces differential geometry tools (vector field Lie algebra) to address kinematic constraints (such as "rolling without sliding") into the general motion planning framework. The worked-out examples will be cars and mobile robots with trailers. Then we will present recent results showing that the human locomotion naturally obeys some nonholonomic constraints. This last example opens perspectives in motion planning for humanoid robots.

The lecture will include: - basic differential geometry concepts - the notion of small-time controllability - the vector field Lie algebra and the LARC theorem - effective methods to steer nonholonomic systems (and open problems) - a general scheme for nonholonomic motion planning - on the human locomotion and humanoid robotics.

Bayesian programming Lab

Juan Manuel Ahuactzin — 2009-11-17T16:46:03Z

Juan-Manuel Ahuactzin Larios

Probayes SAS. 345, rue Lavoisier, Inovallée 38330 Saint Ismier Cedex, France

This is the practical sessions of Bayesian Programming where we teach a second level of expertise: how to develop and implement a computer program effectively that includes some Bayesian computation. Programs will be developed in ProBT© a C++ multi-platform library used to automate probabilistic calculus. The ProBT® library has two main components: (i) a friendly Application Program Interface (API) for building Bayesian models and (ii) a high-performance Bayesian Inference Engine (BIE) allowing execution of the probability calculus in exact or approximate modes.

CONTENT

Variables
Distributions
Bayesian networks
Types of reasoning
Mixture models
Naïve Bayes
- Parameter estimation
- Classification
- Sensor fusion
HMM
Bayesian filter
Particles filter
Learning