Find the Game Piece with the Motor Current

In tutorial 3 you stopped the indexer with one number. It worked, because everything in that tutorial was perfect.

Nothing on a robot is perfect.

What is wrong with one number

Put the same code on the real robot and it fails for five separate reasons.

The motor starts. A motor that begins to turn pulls a huge current for a moment, because it is not turning yet and nothing limits it except resistance. This is called inrush current, and it can be 70 A on a motor that never pulls more than 30 A while it works. Any rule with a number below 70 fires the moment the indexer starts.

Every robot idles differently. A cold indexer, packed with game pieces, with a tight chain, pulls 30 A with nothing in the path. The same design, run in, on a weak battery, pulls 5 A. Your number cannot be right for both.

The battery matters. A motor on a tired battery gets less voltage, so it pulls less current for the same work. The same game piece reads a fifth lower.

The signal is noisy. The reading moves by an amp or two every sample, and the battery sags slowly through a match. Any line gets crossed sooner or later by noise alone.

Not every touch is a game piece. A piece can brush the intake and be knocked away. The current goes up and comes back down, and the robot must not stop for it.

The idea you need

You cannot know the idle current in advance. So do not try to. Measure it on the robot, at the start of every run, and look for a rise above what you measured. That is the whole idea, and it is the answer to a great many problems on a robot, not only this one.

Two details make it work:

Measure it after the motor has settled, or the inrush ends up in your measurement.
Use the middle value of your samples, not the average. One strange sample moves an average. It cannot move the middle value. This is a median, and it is worth knowing.

The task

You did this before with one number. On a real robot one number is not enough.

The indexer motor tells you when a game piece arrives, because a motor that works harder pulls more current. The current is not a clean signal:

The motor starts. For the first fifth of a second it pulls about 70 A whatever is happening. This is inrush current.
Every robot idles differently. A cold, full, stiff indexer pulls 30 A with nothing in it. A loose one on a weak battery pulls 5 A. Your code must work on both, so it cannot contain the number 20.
The battery matters. The same mechanism on a tired battery pulls a fifth less current for the same work.
The signal is noisy, and the battery sags slowly through the match.
A game piece can brush past without going in.

update(amps, dt) runs every 20 ms. Return True once, at the moment the indexer must stop. Return False the rest of the time.

The game piece is inside the robot between 18 cm and 5 cm. Stop too early and it falls back out. Stop too late and it is against the Jamomatic.

There are seven runs. Get all seven.

You can call play(run) and read RUNS from your own code to see what the seven runs give you.

Your code

class JamDetector:
    """Decide when the game piece is inside the robot.

    update() is called every 20 ms with the current the indexer motor pulls.
    Return True once, at the moment the indexer must stop.
    """

    def __init__(self):
        pass

    def update(self, amps, dt):
        # The obvious answer. It is also wrong on most of the runs.
        return amps > 20.0

The code operates in your browser. The page sends nothing to a server and saves your work on this computer. Press Ctrl+Enter to run it.

When you are stuck

Print things. Your code can call play(run) and read RUNS. Run one case by hand and look at what it gives you:

for i in range(0, 450, 25):
    print(i * DT, _amps(RUNS[0], i, _at(RUNS[0]["path"], i * DT)))

Why the obvious answers fail

tools/validate-jam.py runs whole families of wrong answers against the seven runs, so these are measured, not opinions:

Answer	Result
The best possible fixed threshold, already ignoring the inrush	5 of 7
An average baseline over the first 0.4 s	5 of 7
A measured baseline, but firing on the first sample above it	0 of 7
A measured baseline, a margin of 4 A to 10 A, held for 0.16 s	7 of 7

The margin has a band that works, from 4 A to 10 A. Below it the sagging battery trips the detector. Above it the weak motor never reaches the line. You are not meant to guess the number. You are meant to find the edges of the band, the same way you tuned a controller.

Shoot while you are moving