Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/spectrum3847/2026-Spectrum/llms.txt

Use this file to discover all available pages before exploring further.

The REBUILT robot uses two complementary vision approaches: Limelight 4 cameras that report AprilTag-based pose estimates via the MegaTag protocol, and a PhotonVision-based simulation layer (VisionSystem.java) that validates camera logic before hardware is available. Both feed pose corrections into WPILib’s swerve pose estimator so the robot knows where it is on the field at all times.

​
Hardware

The robot mounts Limelight 4 cameras at fixed positions defined in the robot CAD. Each Limelight has its own IP address and pipeline configuration managed through the Limelight web interface. The physical offset of each camera from the robot center — translation and rotation — must be entered in the Limelight web interface or set in code to produce accurate field-relative pose estimates.

​
MegaTag1 vs MegaTag2

Limelight exposes two MegaTag pose formats:
FormatMethodOutput typeNotes
MegaTag1getMegaTag1_Pose3d()Pose3dRaw 3D pose; rotation can be ambiguous with fewer tags visible
MegaTag2getMegaTag2_Pose2d()Pose2dUses the Limelight 4’s onboard gyro for rotation; preferred for odometry fusion
MegaTag2 is more reliable for continuous odometry correction because the LL4’s internal gyro removes heading ambiguity that affects single-tag estimates.

​
Adding vision measurements

Vision pose corrections are fed into the swerve pose estimator inside a subsystem periodic() method using the WPILib API: 
// Overload used in this codebase — timestamp is required, std devs use estimator defaults
swerve.addVisionMeasurement(pose, timestampSeconds);

// Full overload with explicit confidence tuning
swerve.addVisionMeasurement(
    pose,                        // Pose2d from getMegaTag2_Pose2d()
    timestampSeconds,            // Capture timestamp, not current time
    visionMeasurementStdDevs     // Matrix<N3,N1>: x, y, theta confidence
);
The timestamp must reflect when the image was captured, not when the code runs. Limelight provides a latency value you subtract from the current FPGA timestamp. 
// Pattern from VisionSystem — getSimPose supplier is the swerve pose getter
public VisionSystem(Pose2dSupplier getSimPose) {
    this.getSimPose = getSimPose;
    // ...
}

@Override
public void simulationPeriodic() {
    visionSim.update(getSimPose.getPose2d());
}

​
AprilTag field layout

The simulation and localization code both reference the 2026 REBUILT field layout loaded by name from the WPILib resource bundle: 
AprilTagFieldLayout tagLayout =
        AprilTagFieldLayout.loadField(AprilTagFields.k2026RebuiltWelded);
visionSim.addAprilTags(tagLayout);
Using k2026RebuiltWelded ensures the tag positions in simulation and on the real field match exactly.

​
PhotonVision camera simulation

VisionSystem.java creates a VisionSystemSim instance and registers simulated cameras with calibrated noise properties, allowing you to run vision logic in simulation without physical hardware: 
SimCameraProperties props = new SimCameraProperties();
props.setCalibError(0.25, 0.08);   // mean error, std dev (pixels)
props.setFPS(20.0);
props.setAvgLatencyMs(35.0);
props.setLatencyStdDevMs(5.0);
Front and back camera transforms are pre-defined. Uncomment the PhotonCameraSim lines in VisionSystem.java to activate simulated cameras when testing vision pipelines off-robot.

​
Pose alignment types

Coarse alignment using the detected area and position of an AprilTag in the camera frame. Useful for driving directly toward a known target without a full pose estimate.
Full field-relative pose correction from the Limelight MegaTag2 estimate. This is the primary mechanism used in VisionSystem and feeds addVisionMeasurement on the swerve subsystem.
A future enhancement that would use QuestNav’s advanced pose estimation for robust localization when AprilTag visibility is sparse. Not implemented in the current codebase.

Swerve and odometry

How the swerve subsystem maintains and exposes the robot pose estimator.

Simulation

How PhotonVision camera simulation integrates with the broader RobotSim setup.

Build docs developers (and LLMs) love

Get started for freeTalk to us