Tetherbot

built by Scott Preston

*
Click on an image to enlarge it
* *
Target Environment Locomotion Method
Indoors 2 Wheels
Sensors / Input Devices Actuators / Output Devices
Sharp IR hacked futaba servos
Control Method Power Source
Tethered Battery
CPU Type Operating System
Intel 80x86 Windows 95/98/NT/2K/CE...
Programming Lanuage Weight
BASIC N/A
Time to build Cost to build
N/A N/A
URL for more information
http://preston.etowns.net/science/robots/tether1.asp
Comments
Tetherbot is a technology demonstrator, like his predecessor Followbot. The main technologies that he demonstrated were: IR Object Avoidance. Servo Control from the PC (Windows NT 4.0). and Algorithm Testbed. All were very successful and Tetherbot will continue to be used in Genetic Algorithm Research.

Subsystems Description
Chassis - consist of a Radio Shack 6"x4" Project Box. The tether is a 25' RJ-45 cable connected to my PC's Parallel Port.

Power - since it's tethered I was able to use an old PC power supply. It supplies +5VDC and ground via two unused wires in the RJ-45.

Locomotion - x2 "Hacked" Futaba 3003 DC Servo Motors. The hack I used was to replace the potentiometer with a voltage divider consisting of two 1/4 watt 2.2k resistors.

Sensors - 2 Hi-Output IR LED's from Radio Shack and 1 Sharp 40 kHz IR receiver Module. A schematic of the circuit can be found here.

Control - PC Parallel Port. Because this is under Windows NT and the OS does not have direct access to the I/O. I am using DriverLINX by Scientific Software Tools, Inc. I also used a bit of code that allows me to measure how long my PC takes to perform For-Next operations. This was very helpful in sending the correct Microsecond pulse to the servo motors, mainly since VB's timer is only Millisecond resolution.

Genetic Algorithms
So far I have only tested one genetic algorithm. It's goal is to teach Tetherbot how to get out of it's playpen in a certain amount of time. If I turn on the "avoid object" algorithm it will eventually be able to escape but I wanted the robot to be able to repeat what it has learned.

To do this I have employed a genetic algorithm that consist of one gene. That gene contains a sequence of directions that tetherbot needs to move. The genes are tested for fitness by the whether Tetherbot realizes he has espaped or the directions from the gene have been exausted.

Email: scott@preston.dyndns.com

Sorted by Robot
<< PreviousNext >>

Sorted by Builder
<< PreviousNext >>

Sponsored Ad