Skip to main content
PoseEstimate is the primary result type for robot localization. It bundles a WPILib Pose2d with the timestamp, latency, and per-tag details needed to feed WPILib’s SwerveDrivePoseEstimator or DifferentialDrivePoseEstimator.

​
Retrieving a PoseEstimate

Four static methods on LimelightHelpers return a PoseEstimate:
MethodAllianceAlgorithm
getBotPoseEstimate_wpiBlue(name)BlueMegaTag1
getBotPoseEstimate_wpiRed(name)RedMegaTag1
getBotPoseEstimate_wpiBlue_MegaTag2(name)BlueMegaTag2
getBotPoseEstimate_wpiRed_MegaTag2(name)RedMegaTag2
All four read directly from NetworkTables and do not involve JSON parsing. Use the WPILib Blue coordinate system (_wpiBlue variants) for new code — the Red-origin variants exist for compatibility.
MegaTag2 requires you to call LimelightHelpers.SetRobotOrientation(name, yawDegrees, ...) every loop before calling getBotPoseEstimate_wpiBlue_MegaTag2. Without a current yaw feed, MegaTag2 pose accuracy degrades significantly.

​
PoseEstimate fields

​
pose
Pose2d
required
WPILib Pose2d of the robot in the selected alliance coordinate system. Constructed from the [x, y, yaw] components of the raw NT array; z, roll, and pitch are discarded.
​
timestampSeconds
double
required
Capture timestamp in seconds, already adjusted for latency. Pass this directly to addVisionMeasurement.
Computed as (ntTimestamp / 1_000_000.0) - (latency / 1000.0) where ntTimestamp is the NetworkTables server-side microsecond timestamp captured atomically with the pose array, and latency is the pipeline latency in milliseconds (index 6 of the raw array).
​
latency
double
required
Pipeline processing latency in milliseconds. This is the tl value embedded at index 6 of the raw botpose NT array.
​
tagCount
int
required
Number of AprilTags used to compute this pose estimate. A value of 0 means no valid pose was obtained.
​
tagSpan
double
required
Distance in meters between the two outermost tags that contributed to this estimate. Larger span values indicate a more geometrically constrained — and therefore more reliable — pose.
​
avgTagDist
double
required
Average distance in meters from the camera to all contributing tags. Shorter distances produce more accurate estimates.
​
avgTagArea
double
required
Average area of contributing tags as a percentage of the image (0–100). Higher values indicate closer or larger tag appearances.
​
rawFiducials
RawFiducial[]
required
Per-tag detail array. Contains one RawFiducial entry for each tag that contributed to the pose. Empty (length == 0) if the NT array length did not match the expected layout — treat this as an invalid estimate. See RawFiducial fields below.
​
isMegaTag2
boolean
required
true if this estimate was produced by the MegaTag2 algorithm (botpose_orb_wpiblue / botpose_orb_wpired), false for MegaTag1.

​
RawFiducial fields

RawFiducial is embedded inside PoseEstimate.rawFiducials. Each entry maps to 7 consecutive values in the raw botpose NT array starting at index 11.
​
id
int
required
AprilTag ID.
​
txnc
double
required
Horizontal offset from the camera’s principal point to the tag center, in degrees. Positive values are to the right.
​
tync
double
required
Vertical offset from the camera’s principal point to the tag center, in degrees. Positive values are upward.
​
ta
double
required
Tag area as a percentage of the image (0–100).
​
distToCamera
double
required
3D distance from the camera lens to the tag center, in meters.
​
distToRobot
double
required
3D distance from the robot origin to the tag center, in meters.
​
ambiguity
double
required
Pose ambiguity ratio for this tag. Values close to 0 are unambiguous; values approaching 1 indicate the solver found two near-equally-plausible poses.
If rawFiducials[i].ambiguity > 0.1, the individual-tag pose is likely unreliable. When tagCount == 1, consider rejecting the entire PoseEstimate if its sole tag has high ambiguity.

​
Helper methods

​
validPoseEstimate

public static Boolean validPoseEstimate(PoseEstimate pose)
Returns true if pose is non-null, pose.rawFiducials is non-null, and pose.rawFiducials.length > 0. Use this as a quick guard before passing a pose to the estimator.

​
printPoseEstimate

public static void printPoseEstimate(PoseEstimate pose)
Prints all PoseEstimate fields and the full rawFiducials breakdown to standard output. Useful during debugging. Prints "No PoseEstimate available." if pose is null.

​
Code example

// In your robot's periodic method:

// 1. Keep the gyro yaw fed to MegaTag2 every loop.
LimelightHelpers.SetRobotOrientation(
    "limelight",
    m_gyro.getYaw(), 0,
    0, 0,
    0, 0
);

// 2. Fetch the estimate.
LimelightHelpers.PoseEstimate estimate =
    LimelightHelpers.getBotPoseEstimate_wpiBlue_MegaTag2("limelight");

// 3. Guard against invalid data.
if (!LimelightHelpers.validPoseEstimate(estimate)) {
    return;
}

// 4. Optional: reject high-ambiguity single-tag estimates.
if (estimate.tagCount == 1 && estimate.rawFiducials.length == 1) {
    if (estimate.rawFiducials[0].ambiguity > 0.1) {
        return;
    }
}

// 5. Optional: reject estimates that are far away or have poor geometry.
if (estimate.avgTagDist > 4.0 || estimate.tagCount < 1) {
    return;
}

// 6. Feed the estimator.
m_poseEstimator.addVisionMeasurement(
    estimate.pose,
    estimate.timestampSeconds
);

Build docs developers (and LLMs) love

Get started for freeTalk to us