Extreme line following

[bens]

Six robots enter, one robot leaves, kind of.  This is an unedited video of an unofficial event at a local robotics competition we held this past weekend, so it's rather long, but it has some interesting points.  The six participants are using the same Pololu 3Pi robot, but all of the robots were programmed independently by different people.  The robots are following the line at just under a meter per second.  The eventual "winner" is playing a Hungarian rhapsody while the third place finisher is playing one of Bach's fugues (mine was the one that finished second).

[youtube]fl0CJhPiEfY[/youtube]

I think next time I'm going to add a sharp sensor to the back of my robot so that it can slam on full reverse and take out the people behind me.

- Ben

[SixRingz]

Really impressive stuff, and inspiring to see someone holding local robotics competitions! I wish I knew some robotic interested people in my hood. Maybe I need to advertise on the University, should be at least a couple robo nerds there like me. 
Great stuff anyhow! It was an exciting race and you all we're so equal. Almost to the point that it seems motor differencies or so could have determined the winner.

[ddemarco5]

those things are taking amazingly smooth turns.

[Rebelgium]

These robots are obviously designed to turn at that exact radius, and no other.
But nevertheless, it's pretty impressive. nice work!

[bens]

These robots are obviously designed to turn at that exact radius, and no other.

Not at all.  The robots were designed with no specific turn radius in mind and can make much tighter turns.  They're following the lines using PID, and the constants were picked through trial-and-error optimization on courses with turns no sharper than a 6" radius.  To function well on sharper turns, it's possible the PID constants would have to be altered somewhat, or maybe the maximum speed would have to be decreased (on this course the robots can run at full speed without losing the line).

Now when my 3pi is running maze solver code it executes pre-defined 90 and 180 degree turns at the intersections, but in the above video the turns are not hard-coded.  The robots don't have any idea that they're on a turn or a straightaway, it's all just PID.

- Ben

[Rebelgium]

These robots are obviously designed to turn at that exact radius, and no other.

Not at all.  The robots were designed with no specific turn radius in mind and can make much tighter turns.  They're following the lines using PID, and the constants were picked through trial-and-error optimization on courses with turns no sharper than a 6" radius.  To function well on sharper turns, it's possible the PID constants would have to be altered somewhat, or maybe the maximum speed would have to be decreased (on this course the robots can run at full speed without losing the line).

Now when my 3pi is running maze solver code it executes pre-defined 90 and 180 degree turns at the intersections, but in the above video the turns are not hard-coded.  The robots don't have any idea that they're on a turn or a straightaway, it's all just PID.

- Ben

ok my bad, it looked like it...
Now let Admin say again that he's religiously against PID, and pro fuzzy logic :p

[Admin]

I'm pro PID when high speed high precision is required for simplified tasks - perfect for manufacture and line following, for example.

I'm against it for low precision tasks that require a high level of robustness - perfect for highly unpredictable/complex environments and robot designs that are very difficult to mathematically model.

My robots usually operate in the second area, but not always . . .

[Parth]

Those little buggers were fast and still maintained sight on the line. They also took the turns really well. Nice bots and programming