Migrating from pybricks v2.0 to version 3.3.x on EV3

[AAATom]

Intro: I'm co-coaching my son's school team which is preparing for national WRO competition. The v2.0 works fine up to its limits, one of largest being that DriveBase has Stop.HOLD as default and no way to change it (like using then parameter in v3.x), what makes movements jerky. I assume this could be worked around with some clever code but, hey, they are just kids starting with Python and "robotics" and would be happy to see such a thing working out-of -the-box (like they found it in the docs, alas, for v3.2).

Last week I have finally found the eluding EV3 "pybricks-micropython" executable (thanks David!) and have successfully installed it (cookbook here.

It works well so far apart from some issues, which I intend to post here in this thread and, hopefully, get some constructive responses from the community 🙂

[AAATom]

Proportional line follower gone bad

The first issue and currently the deal breaking one for the team (in using v3.3.) is that line follower code that works fine in v2.0 works badly in v3.3 (build 2828).
When starting away from the line (over white) it cannot detect the line (or rather, the line edge) and wanders off to the other side of the line and on.

With some debug prints we discovered that the loop runs some 3 times faster on v3.3 (makes 3 times more passes) for the same distance the DriveBase covers.
The values read by the ColorSensor are practically the same - except that at the point where it gets over the line edge it does not result in the same behavior.

The line follower code is this:

def TrackLineDistance(robot:DriveBase, distance:int, speed:int,
                      sensor:ColorSensor, isLeftSide:bool,
                      sensorThreshold:int, proportionalGain:float):
    
    sign = -1
    if isLeftSide:
        sign = 1
    
    robot.reset
    
    while robot.distance() < distance:
        deviation = sensor.reflection() - sensorThreshold
        turn_rate = proportionalGain * deviation * sign
        robot.drive(speed, turn_rate)

The parameters pass to TrackLineDistance() are
speed = 40
sensorThreshold = 20
proportionalGain = 2

The team tried changing the speed and gain but it did not change much.
At the moment they are back to v2.0 as they don't have much time to debug this.

Here is a graph of the v2.0 data:

And graph of v3.3 data:

Notice the "hump" in v3.3 that is not there in v2.0 - it is the robot passing over the other edge of the line into white area and then into blue area where it wanders off mostly straight.

What could be the reason v3.3. is behaving so differently?

[laurensvalk]

I think this is the same as #1035.

This is why we currently don't release EV3 versions and recommend that people use 2.0 for the time being :)

But we do get quite a lot of questions from WRO teams about this, so hopefully we can pick this up at some point.

[AAATom]

Thanks! I am aware about potential problems with v3.x on EV3, of course.

Reading posts in #1035, looks like these things were appearing on and off after v3.2 - can I get that version please?

Interesting is that other driving code worked just fine. Is this something about doing robot.drive() in a loop?
Should I try adding some delays with wait() as mentioned in couple of places?

[laurensvalk]

Yes, that’s always good.