FireBall

Expressway was a very easy line follower to build and program. Since line following was so easy, why not build a fast line follower? Extra Parts was the result. It used the same line sensors as Expressway, the same wheelbase as Expressway, the original motors out

of Excuse, a spare NiMh battery, and an old control board that was in the junk box. The problem with Extra Parts was that it would not follow a straight line. About a year of fine tuning the PID algorithm and settings, and it still would not follow a line. There was a spare LiPo battery laying around, so that went into Extra Parts in place of the NiMh and everything started to work! So, in addition to speed, torque, and traction, add low weight (or mass) to the desirable characteristics of a good line follower. After all of that fun, it seemed like the ExSpurt chassis would make a good line follower since it had the proper characteristics.

At the 2009 Peoria Bot Brawl, club member Len Sheldon was raffling off one of his Probotix FireBall V90 machines. Years ago some friends tried to get me interested in a CNC mill, but there was too much other stuff to do. Now, this new machine looked very interesting. So, I tried to buy all of the raffle tickets, but ran out of money before they ran out of raffle tickets. Unfortunately, I didn't win it. Fortunately, My videographer, Cathy, did win it. Unfortunately, I had to listen to every thing that she was going to make for the five hour drive home. Fortunately, Cathy didn't have room for it at her house so FireBall V90 is in my basement. Unfortunately, I have to ask permission to use it every time.

Cutting FireBall Chassis on FireBall V90

I had extra Maxon motors, wheels, and batteries from ExSpurt. The Pololu QTR-8RC is a much nicer sensor package that was used on Expressway and Extra Parts, so one was added to the project. A new chassis was required that was a little wider and had a front sensor mount. Late one Saturday evening, I started the chassis drawing. It turned out well, so on to download Vectric Cut 2D software and go through the tutorials. G code was generated. After several "air" parts were made, the chassis was cut out of carbon fiber, my first CNC part. The whole thing from initial drawing to final finished chassis was done in about an hour! This is the way to make robots!

Take the new chassis, four Maxon motors, and the old ExSpurt jig, add some epoxy, let cure overnight, and the base for a new line follower has been built.

I've been having good luck with the Parallax Propeller in the RoboMagellan robot, so why not use it here. There is a small package called the Spin Stamp that puts a Propeller chip on the same size package as the Basic Stamp. A new control board was made with the same H-Bridge and driver as ExSpurt, but with a Spin Stamp. The voltage regulators on the Spin Stamp are a little unusual, so it took some time to work out a reasonable voltage regulator arrangement that provided both 3.3 volts and 5.0 volts.

Of course, all of the code was downloaded from Parallax forums and the Propeller Object exchange. The Spin Stamp only uses three cogs, the PID loop, QTis sensor, and SN754410 motor driver. The only change that was made to the SN754410 object was the pin numbers. The QTis object was changed quite a bit. First, only 2,000 counts (25usec) is used to charge up the capacitor. Then the discharge lasts for 40,000 counts (0.5 msec). Finally, a Case command was added to convert the sensor input to an integer offset from center. The Case command in Spin wasn't very fast at about 250 usec, so it was run during the 40,000 counts that the capacitors were discharging.

The PID loop was extensively modified. It had the following variables:

Speed factor is a way of slowing down the top speed. This is very handy while fine tuning the PID loop and adjusting to variable conditions at contests
PID loop timing can be as fast as 1 msec. There doesn't appear to be any performance gain for 1 msec PID loop over 10 msec PID loop.
Proportional feedback factor. This is the standard PID loop factor.
Integral feedback factor, this has always been zero for my robots.
Differential feedback factor. Again, the standard PID loop factor.
Center Point, it the offset from the center of the line that the robot will run. Some contests have mostly left or right hand turns, so some adjustment of the center point might be beneficial.
Corner Speed, is the amount of the PID correction factor that is fed to the "fast" or outside motor in a turn. The purpose is to slow down when in a turn.

That seems to be enough tuning things to adjust for anyone.

FireBall's first contest was the 2010 Bot Brawl in Peoria. Practice was black 3M electrical tape on the white kitchen floor. Please watch the videos here. The first run was at a speed factor of 60%, which was well within FireBall's capacity. The second run was at a speed factor of 70% which is about at the limit experienced on the kitchen floor. The third run was 75% and slower due to lack of smoothness in the corners. It appears that FireBall is about 25% faster than the Pololu 3pi, but only slightly faster than Mike Dvorsky's Low Down. FireBall was lucky!

The next contest was ChiBot's Fall 2010 at the Ihobby Expo. The racing line following contest is unusual in that they have a white line on a black surface. No problem? Not quite. It seems that one or two sensors detect a black line on a white surface, but three or four sensors might detect a white line on a black background. The white must "bleed" over a lot! So, the Case command was a lot longer. For practice, a sheet of plywood was painted black and white 3M electical tape marked the course. Lots of runs were made on this surface fine tuning the variables. Due to the wider appearing line, fastest speed was lower than for the Peoria contest. FireBall won again against international competition.