[Altainia]

Refs #99.

Updates examples/maxswerve to REV's current MAXSwerve Java template (REVrobotics/MAXSwerve-Java-Template @ b614589, "Update for 2026"), written for the 2027 API on main. The old example didn't import under 2027 (wpimath.controller is gone).

It also takes maxswerve off the ignore list in examples/run_examples.py, so CI now runs its tests like the other examples. Because the example uses commands2, tests/requirements.txt now includes robotpy-commands-v2>=2027.0.0a7,<2028. The 2027 version has to be named, or pip installs the stable 2026 release; the requirement will need updating along with the wpilib version.

What follows the template

• configs.py (new) mirrors Configs.java:
  • the drive velocity feedforward is nominalVoltage / free speed, i.e. kV in volts. The old 1 / free speed is maxswerve example: crashes in robotInit on 2026.0.4, and its drive feedforward is 12x too weak under REVLib 2026 #99's 12x-too-weak feedforward;
  • the turning encoder uses the Through Bore Encoder V2 preset;
  • feedback sensors use rev.FeedbackSensor.
• The module, drive subsystem, robot container (auto trajectory, X-formation binding) and constants follow the template.
• swerveutils.py and the slew-rate limiting are gone, as they are in the template.

Where it differs, because of 2027 API changes

1. No encoder conversion factors (removed from REVLib 2027). The SPARKs run in native units: RPM and rotations for the drive encoder, rotations for the absolute encoder. The module converts to meters and radians itself. The template's gains are kept in its own units and scaled when configured:

   • drive P is 0.04 × (m/s per RPM);
   • drive kV is 12/5676 V/RPM;
   • turning P is 2π.

   Position wrapping is enabled over one rotation.

2. No SwerveControllerCommand or HolonomicDriveController in WPILib 2027. swervecontrollercommand.py is a small equivalent with the same behaviour:

   • it follows a trajectory from DrivetrainSplineTrajectoryGenerator;
   • x/y PID corrections are added to the sampled field-relative velocities;
   • a profiled theta controller holds the final heading.

3. Gyro: the ADIS16470 class is gone, so the example uses wpilib.OnboardIMU (FLAT mounting).

4. Buttons: 2027 renumbered the Xbox buttons. The template's set-X binding (raw button 6) is now RIGHT_BUMPER.

5. TimedRobot changes: initialization is in __init__, and test mode is now utility mode. Usage reporting is hal.report_usage("RobotDrive", "MAXSwerve").

Testing

All runs used RobotPy 2027.0.0a7 and robotpy-rev 2027.0.0a7.post1 on Linux.

• robotpy test --builtin in examples/maxswerve gives 11 passed and 1 skipped (test_opmode, which every example skips). The new tests cover:
  • heading units;
  • the drive kV (a guard against maxswerve example: crashes in robotInit on 2026.0.4, and its drive feedforward is 12x too weak under REVLib 2026 #99);
  • the turning encoder config;
  • module setpoints in native units, with the chassis offset;
  • state and position conversion;
  • the X-formation binding;
  • autonomous running to completion.
• examples/run_examples.py passes for all 12 examples, and examples/check_header.py and black --check are clean.
• In simulation, I drove each SPARK through SparkMaxSim.iterate with a simple motor model (not included in this PR):
  • One module at 2 m/s: it settles at 2.03 m/s. With the old feedforward it reached 0.47 m/s.
  • The autonomous S-curve: it ends about 3 cm from (3, 0) with a near-ideal plant. With a 50 kg robot it ends 0.24 m past the endpoint, which is momentum against the template's example gains. REV's Java template overshoots by about the same amount in stingr-sim, a MuJoCo-based simulator.

Merging this won't fix the 2026.0.4 release that #99 was reported against. The fix for that release is in the issue.

[virtuald]

Thanks!