 |
| 
|
 |
Kerwin's Line Following
Robot Page |
| My line follower bot is a pretty basic design
using a differential drive system. I built with Futaba S-148
servo motors mounted to the bottom of the plexiglass. I bolted
the hub of the Dave Brown Lite Flite foam wheels to the control
horn of the servos. The sensor system consists of an array
of 3 matched IR transmit/receive pairs mounted on a circuit
board that can be raised or lowered to fine tune the sensitivity.
The omni-directional wheel on the front is not really necessary,
but it looks cool. I used it because I had it laying around
and added color. I made a mounting bracket for it using a
PC's PCI blank port cover I pulled out of the trash. Nice
soft metal in those things and easy to bend.
The circuit board mounted on top is one of Dennis Clark's
AVR based robot boards. The Denver Area Robotics Club bought
a bunch of these from him en masse, but any controller would
work of course. |
 |
The IR sensor array consists of 3 IR transmit/receive
pairs that are matched for performance. I got these from Electronic
Goldmine, part number G2540. But any similar part would
work. I painted the circuit board black and placed some black
felt between each pair to cut down light being received from
the side rather than as a reflection. The bolts used to adjust
height are on each side. |
| A rough block diagram is shown here. The diode
is the IR transmitter and the receiver is the phototransistor.
I only show one here for clarity. The microcontroller takes
input from sensor array, and drives the servo motors in response. |
 |
| The output of the phototransistor looks like this. The
plot here shows the voltage response over time as the sensor
travels from black to white (as the robot loses the line).
The output voltage drops from about 3.5V on average to about
0.2V in 18 ms.
The Atmel Botboard I use on this project has 2 A/D channels
via an A/D chip. The data from this A/D chip is read by the
Atmel 90S2313 by a serial interface. I use the 2 A/D channels
for the left and right sensor. The center sensor I run to
the 2313's onboard comparator. |
 |
| I retrieve the data from the A/D channels as a 12 bit digital
value that represents the Voltage. 2^12 is 4095, so if I had
5V input, the A/D would report a value of 4094 (decimal). 0V
would be a value of 0. 2.5 V would be halfway at 2047. I read
the value, and then compare it to a threshold value of about
2050 or so (a little over 2.5V). If the output of either of
the A/D channels goes below 2050 then I call that a "white"
value and I call that a 0 output. Above that threshold is black,
which is a 1. |
| The center sensor goes to the 2313's onboard
comparator. I set the compare or reference voltage with a
divider circuit to get 2.5V as a threshold. Then with software
I invert the output of the comparator so that a Voltage above
2.5V is an output of 0 (white). Below the ref is a 1 (black).
I take a sample every 20 ms or so since that is the fastest
the data can change, and make a decision based on a state
table of the possible states of the three sensors. This table
is shown to the right. |
| L |
C |
R |
Meaning |
Action |
| 0 |
0 |
0 |
Lost |
Decide |
| 0 |
0 |
1 |
Rt is on line |
Right turn |
| 0 |
1 |
0 |
Center on line |
Go forward |
| 0 |
1 |
1 |
Rt/Cen on line |
Right turn |
| 1 |
0 |
0 |
Left on line |
Left turn |
| 1 |
0 |
1 |
Confused |
Back up,try again |
| 1 |
1 |
0 |
Lft/Cen on line |
Turn Left |
| 1 |
1 |
1 |
Confused |
Back up,try again |
|
 |
This photo shows a closeup of the sensors in action. IR light
is invisible to the human eye, but not to most video cameras.
Any CCD camera can "see" IR light. A CCD camera is
a very handy tool to have when debugging and testing an IR circuit.
Since you can't see it, you don't know if its on. If you use
IR LED's, you should use a CCD if you have one to be sure they
are working before you waste a lot of time chasing some other
problem that isn't there after all. |
|
|
