The Line Follower Tests

Design Thoughts

Line following robots are often design with the technique best described as seat-of-the-pants. The DBW (Distance Between Wheels), and FOS (Forward Offset of Sensors) is often based on the size of motors available and what looks good. The number of sensors and motor speed is also designed with what we have on hand.

So as I started on my next line follower I thought about what it would to really "design" a robot. After a few weeks of thought I came up with the testing that follows. The project right now (August 2, 2003) is in the starting stages and I'm looking for input on what else to test and record.

Just to catch your eye I'll put one of the test result pictures here at the start.

Hopefully the picture will interest you enough to read how it was done. As the project is in the early stages I would love to have input on what else to test. So read on and let me know what you think. If you want to jump right to the pretty pictures then go ahead, but come back and read later.

The Robot Base To be able to test a number of configuration I had to have a base that was adjustable. I worked up a set of three bases. Each base had a different DBW, 7", 5-1/2", 4", and had mounting holes to accommodate the sensors places in four different FOS positions. This gave 12 different base setups. Processor Computing power is supplied by a Sumo11 board, this is a mini-sumo sized version of the handy board. The programming is in IC4 an interactive version of C designed for the handy board and Lego RCX. I used the built in PWM and H-bridge to drive the motors, and the thumb wheel built into the board allowed for easily changing the speed of the motors before each test.
Sensors The sensors for the test are built from four Fairchild QRB1134s. Each sensor runs to one input on the Sumo11 board. The sensors are spaced on 5/8" centers. From past tests a spacing of just less than the width of the line to follow seems to work best. Because the sensors are about 7/8" wide I places the sensors on both sides of a circuit board to achieve the 5/8". Through program I plan on testing a number of sensor configurations. At this time I plan on testing the following configurations: one sensor finding the edge of the line. the two center sensors, using the logic of, if both over the line go straight, if one sensor over the line turn away from line. three sensors, using the logic of, if center sensor go straight, if outside sensor sees line turn away from line. four sensors, using a logic based on turning faster depending on how far the line is from the center of the sensors. These four sensor configurations, time the number of base configurations, gives a total of 48 configurations.
The Course To make for consistent testing I set up a track just for these tests. The track was made to match the conditions used by ChiBots in their line following contests. The bot will travel starting at the bottom right moving to the left. The first wide turn has a radius of 14", the two sharper turns have a radius of 6" to match the ChiBots course. The bot is programmed to turn on the BLM (see below) at the first white cross line and to turn off the BLM and motors at the second white cross line. This gives a total course length of 380cm or 150 inches. The fixed length makes for easy speed calculations.
BLM (Blinky Light Module) The heart of the whole testing and recording system is the BLM. The BLM has five LEDs. Three are on all the time the BLM is powered, the white one in the center marks the center of the sensor array. The two red ones are on long wires and mount close to the wheels on each base to mark the wheel positions. The "Blinky" part is the two green LEDs. I wanted to make them flash at a set rate. I remembered I have a few leftover logic probes from Peter Anderson's and his Home Brew Basic Stamps. These logic probe chips also have two outputs, one at 1 cycle per second the other at 10 cycles per second. These flashing LEDs make it easy to see on the pictures how long it took to complete a course. The BLM is hooked to one of the Sumo11 boards output to allow it to be turned on and off as needed.
Testing The testing was done by placing a digital camera over the test track. The camera was set to take a 15 second exposure. As the exposure was started the robot was started. As the bot crossed the first white cross line the BLM turns on. The robot, hopefully, continues around the track until it reaches the second white cross line at which time the motors and BLM turn off. I will be testing each of the bases with each of the sensor styles, 48 total tests. As the robot is designed to run at any speed up to about 1 meter per second I plan on testing each configuration at slow speeds and then work the speed up to a maximum for that configuration This could theoretically make for hundreds of test. My hope is to find trends to support or dispute some of the seat-of-the-pants type designs we have now.