Hypervelocity's World: January 2009

Wednesday, January 28, 2009

Projects have been done in Samsung

[2004] SHADOW Project - Graphics Library for Digital TV
[2005 ~ 2006] DSP Project - Digital TV Software Platform
[2007] MultiCube Project - Asynchronous File System for Flash Memory
[2008 ~ ] Robo-Ray Project - Software Platform for Humanoid Robots
[2009 ~ ] TAME Project - An Affective Component for Robots and CE Devices

Back to Portfolio

Tuesday, January 27, 2009

Q-Learning

I applied Q-Learning to path planning and Hanoi Tower. Q-Learning is more clear than Back-Propagation and easier to use compared with other machine learning algorithms.

Back to Portfolio

A* Algorithm

I made a program which could test the A* algorithm. You can build a map and test your own A* algorithm : you can change hueristic equation and add the smoothing algorithm and so on.

Back to Portfolio

Machine Learning and Algrorithm

Neural Network

I used Back-propagation for recognizing objects such as sphere and cube. First, this program trained with desired images. Then, it read images from a camera and analyzed it based on the trained weight functions. Moreover, I implemented a back-propagation program.

This program is Self-Organizing Map(SOM). I made this program for categorizing data.

Q-Learning

This program is a back-propagation training progam. You can add layers and nodes, train your own data, and change some parameters such as steepness, momentum and learning rate. Finally, you can save the result of training to a file. I applied Q-Learning to path planning and Hanoi Tower. Q-Learning is more clear than Back-Propagation and easier to use compared with other machine learning algorithms.

A* Algorithm

I made a program which could test the A* algorithm. You can build a map and test your own A* algorithm: you can change hueristic equations and add the smoothing algorithm and so on.

More

Back to Portfolio

Vector Field Histogram(VFH) Simulator

This simulator can test VFH in various environments and make your own environment using a map editor. I made this simulator and tested my navigation algorithm in many cases before testing the algorithms in a real robot. OpenGL was used to construct 3D environments.

VFH Simulator Architecture

VFH Simulator Map Editor

Test Results

Back to Portfolio

Friday, January 23, 2009

Samsung Software Membership Projects (SSM)

Home Networking

This project was to monitor a house in the Internet. A user could access the camera and get various information of a house on a homepage which we made. This was my first Software Member Ship project.

X-Men Project

In this project, my part was to implement a navigation program. I applied HIMM and VFH to an omni wheel robot and it could send camera images to a PDA of an user and the user could control the robot with the PDA.

Goliath Project

My team designed this robot's appearance and made it. My part was a navigation program and a behavior generation program. I used A* algorithm and Histogram In Motion Mapping(HIMM) for navigation. We could generate a variety of behaviors with the behavior generation program.

Back to Portfolio

Mobile Robot

I implemented a program which could build a map and avoid obstacles by gathering distance data from a robot. Moreover, I make the mobile robot and used PID algorithm to control a DC motor of a RC car.
Spec : 8 Ultrasonic Sensors, 1 Digital Compass, 1 DC Motor, PIC MCU

Back to Portfolio

Robot Intelligence Development Environment(RIDE)

I developed a robot simulator, RIDE. RIDE presents a simple and efficient method for modeling a robot cognition system using synthetic vision in virtual environments. Virtual robots perceive external stimuli through a synthetic vision system and use the perceived data for navigation and action. I made my own graphics engine with OpenGL and used Open Dynamics Engine(ODE). RIDE is composed of two things, RIDE Editor and RIDE Simulator.

RIDE Architecture

Synthetic Vision Results

Robots and Environments

Map-Building Using Synthetic Vision

RIDE Sim RIDE Editor

Face Expression

Goal Driven Agent Test

Back to Portfolio

Eyebot

EyeBot was developed at he University of Western Australia. The robot has differential steering, three infrared position sensitive devices (PSDs), a color CMOS camera, a digital compass, and a wireless communication module. The three distance data values from the PSDs and the robot's current location (delivered by shaft encoders) are sent to the host computer to update the map-building with HIMM in real-time. A new path is generated using the A* algorithm and transmitted to the robot. 3D map building is also constructed with distance data and the current robot position.

Architecture

Robot Control Program

Go to a destination in a unknown environment

Eyebot

Back to Portfolio

Haptic Device

This was a haptic device, detecting the status of your hands. I used neural network algorithm for recognizing the status of a hand. If you express a gesture with this glove, the control program can recognize the gesture, but desired gestures must be trained before using this haptic device.
Spec : 5 Flex Sensors, PIC MCU, 1 Bluetooth

Architecture

BP Training and Recognition Program

Haptic Glove

Back to Portfolio

Four Legged Robot

I made a four legged robot. It could walk, build a map and avoid obstacles. Histogram In Motion Mapping(HIMM) and Vector Field Histogram(VFH) were used for navigation and I implemented the control program which could build the 3D map of environment, generate behaviors of the robot and test the algorithms of map-building and obstacle avoidance.

Spec : 14DOF, AI Motor, 1 LCD, PIC MCU, 5 Ultrasonic Sensors, 1 Bluetooth, 1 Touch Sensor
I won several contests in Korea with this robot.

Architecture of Four Legged Robot

Robot Control Program

Motion Generator

Back to Portfolio

My First Biped Robot

This robot was my first biped robot which I designed all of it. I made a program to generate behaviors of this robot. I could make various behaviors without a real robot and then I could apply the created behaviors to the real robot.

Spec : 19DOF, AI Motor, PIC MCU, 1 Bluetooth