PathRecorderFTC

Turns routes driven by drivers into Java code for a Pedro Pathing 3 autonomous with Ivy commands.

Install

Add JitPack to TeamCode/build.gradle (or to build.dependencies.gradle in the Quickstart):

repositories {
    maven { url 'https://jitpack.io' }
}

dependencies {
    implementation 'com.pedropathing:revhub:3.0.2'
    implementation 'com.pedropathing.ivy:pedro:1.1.1'
    implementation 'org.westwoodrobots:pathrecorder-ftc:0.1.0'
    implementation 'org.westwoodrobots:pathrecorder-ivy:0.1.0'
    implementation 'org.westwoodrobots:pathrecorder-panels:0.1.0'
}

panels module is optional. Add only if your team uses Panels.

The library uses compileOnly for Pedro, Ivy, Panels and the FTC SDK. Your project supplies these at runtime.

Compatibility

Dependency	Version	Source
Pedro Pathing	3.0.2	v3.0.2 (cf4be31)
Ivy	1.1.1	v1.1.1 (d35242a)
Panels field plugin	1.0.10	mymaven.bylazar.com (a68ed55)
FTC SDK	12.0.0	v3.0.2
Android Gradle Plugin	8.13.2	same as Quickstart
Java	source and target 1.8	same as Quickstart

Use Pedro 3.0.2 or later. The Quickstart v3.0.2 still uses revhub:3.0.1. In Pedro 3.0.1, linear heading moves with the curve parameter t. In 3.0.2 it moves with arc length, and the heading classifier uses arc length.

Quick start

See RecordAuto in sample/.

recorder = new PathRecorder<>(follower, new Pose(8, 72, 0), config, IvyRuntime.owned())
        .action("claw.open()", () -> gamepad1.aWasPressed(), claw::open)
        .action("claw.close()", () -> gamepad1.bWasPressed(), claw::close)
        .waypoint(() -> gamepad1.xWasPressed())
        .finish(() -> gamepad1.backWasPressed())
        .overlay(new PanelsOverlay());

// in loop()
recorder.update(gamepad1, telemetry);
telemetry.update();

• The first arg of action is the Java expr that the generator writes into the output, must be valid.
• The last arg supplies a new Command for each push, scheduled by the recorder.
• Trigger is fired on the rising edge.
• The constructor calls follower.setPose(start). update calls follower.manual(...) and follower.update(). Do not call them again in the same loop.
• IvyRuntime.owned() calls Scheduler.execute() in update. If your code already calls Scheduler.execute(), use IvyRuntime.external(). Scheduler is static, so call Scheduler.reset() in init().
• The default stick mapping is the mapping of the Pedro localization test. For field-centric drive, use .drive(DriveInput.fieldCentric(0)), or give your own DriveInput.

When the driver pushes the finish button (or the OpMode stops and calls finish()):

• Code goes to /sdcard/FIRST/pathrecorder/<timestamp>.java on the Control Hub. Use adb pull /sdcard/FIRST/pathrecorder.
• http://192.168.43.1:8080/pathrecorder will display the newest file

The output format should be imports first, statements in a method in scope of follower/subsystem, and command is scheduled, and setPose() is called.

With config.style = OutputStyle.CLASS you get a class with start() and create(Follower, ...) instead.

Configuration

All settings are public fields of RecorderConfig.

Field	Default	Meaning
positionTolerance	1.0	Maximum distance (in) from a recorded point to the fitted curve
headingTolerance	3.0	RMS heading error (deg) for a heading mode, and the turn tolerance
stopSpeed	1.5	Speed (in/s) below which the robot is stopped
minStopTime	150	Minimum stop time (ms) that splits a path
sampleInterval	0	Minimum time (ms) between samples, 0 records every loop
degrees	true	Degrees through PoseFactory.degrees(), or radians through new Pose(...)
mirrorAxis	NaN	Mirror the output with PoseFactory.mirrorX(axis), for example 72 for the other alliance
style	SNIPPET	SNIPPET or CLASS
followerName, commandName, startName, poseName, pathName	follower, auto, start, p, path	Names in the output
packageName, className, parameters, imports		Settings for CLASS output, for example parameters.add("Claw claw")
outputDirectory	/sdcard/FIRST/pathrecorder	Directory for the files
livePreview	true	Fit the route again at each stop for the overlay

Notes

• A short pause (less than minStopTime) with no turn does not split a path.
• A stop with a heading change becomes PedroCommands.hold and a wait for the heading, with a timeout of two times the recorded turn time (500 ms minimum).
• A waypoint during motion splits the path into two paths. Pedro can slow the robot at the end of each path.
• The robot waits at each stop until the actions of the stop are done. The recorded wait time is not in the output.
• A path with an action during motion is a Groups.parallel. The next step starts only when all actions of that path are done.
• One path can contain more than one Bézier curve as Paths.path(...). In Pedro 3.0.2, follower.completion() is not correct for such a path. For these paths the output uses follower.pathIndex() and follower.parametricCompletion() instead.
• The output depends on follower.update() in each loop. follower.completion() is 1.0 until the first update after follow(), the same as PedroCommands.follow.
• A path that crosses itself can place an action at the wrong crossing.
• If no heading mode is inside the tolerance, the output uses linear and a // heading error note with the measured RMS error.
• Odometry drift goes into the output. The output is only as good as your localizer.

Competition rules

During a match, the Competition Manual (R704) does not permit live data streams other than Driver Station telemetry. Do not put any recorders in a competition OpMode.

License

BSD 3-Clause. See LICENSE.