[examples] Add Computer Vision Pose Estimation and Latency Compensation Example (#4901)

This PR updates the existing differentialdriveposeestimator example to include computer vision pose estimation and latency compensation.

The example generates a simulated cameraToTarget transformation, which is then fed into ComputerVisionUtil.objectToRobotPose() to compute the robot's field-relative position exclusively from vision measurements. The vision measurements are applied through DifferentialDrivePoseEstimator.addVisionMeasurement().

The updated example constructs an AprilTagFieldLayout from JSON. This requires a deploy directory, something which isn't currently supported in wpilibjExamples and wpilibcExamples.
This commit is contained in:
CarloWoolsey
2023-01-18 20:46:05 -08:00
committed by GitHub
parent cb9b8938af
commit 371d15dec3
5 changed files with 386 additions and 35 deletions

View File

@@ -4,21 +4,42 @@
package edu.wpi.first.wpilibj.examples.differentialdriveposeestimator;
import edu.wpi.first.apriltag.AprilTagFieldLayout;
import edu.wpi.first.apriltag.AprilTagFields;
import edu.wpi.first.math.ComputerVisionUtil;
import edu.wpi.first.math.VecBuilder;
import edu.wpi.first.math.controller.PIDController;
import edu.wpi.first.math.controller.SimpleMotorFeedforward;
import edu.wpi.first.math.estimator.DifferentialDrivePoseEstimator;
import edu.wpi.first.math.geometry.Pose2d;
import edu.wpi.first.math.geometry.Pose3d;
import edu.wpi.first.math.geometry.Quaternion;
import edu.wpi.first.math.geometry.Rotation3d;
import edu.wpi.first.math.geometry.Transform3d;
import edu.wpi.first.math.geometry.Translation3d;
import edu.wpi.first.math.kinematics.ChassisSpeeds;
import edu.wpi.first.math.kinematics.DifferentialDriveKinematics;
import edu.wpi.first.math.kinematics.DifferentialDriveWheelSpeeds;
import edu.wpi.first.math.numbers.N2;
import edu.wpi.first.math.system.LinearSystem;
import edu.wpi.first.math.system.plant.DCMotor;
import edu.wpi.first.math.system.plant.LinearSystemId;
import edu.wpi.first.math.util.Units;
import edu.wpi.first.networktables.DoubleArrayEntry;
import edu.wpi.first.networktables.DoubleArrayTopic;
import edu.wpi.first.wpilibj.AnalogGyro;
import edu.wpi.first.wpilibj.Encoder;
import edu.wpi.first.wpilibj.RobotController;
import edu.wpi.first.wpilibj.Timer;
import edu.wpi.first.wpilibj.motorcontrol.MotorController;
import edu.wpi.first.wpilibj.motorcontrol.MotorControllerGroup;
import edu.wpi.first.wpilibj.motorcontrol.PWMSparkMax;
import edu.wpi.first.wpilibj.simulation.AnalogGyroSim;
import edu.wpi.first.wpilibj.simulation.DifferentialDrivetrainSim;
import edu.wpi.first.wpilibj.simulation.EncoderSim;
import edu.wpi.first.wpilibj.smartdashboard.Field2d;
import edu.wpi.first.wpilibj.smartdashboard.SmartDashboard;
import java.io.IOException;
/** Represents a differential drive style drivetrain. */
public class Drivetrain {
@@ -50,6 +71,18 @@ public class Drivetrain {
private final DifferentialDriveKinematics m_kinematics =
new DifferentialDriveKinematics(kTrackWidth);
private final Pose3d m_objectInField;
private final Transform3d m_robotToCamera =
new Transform3d(new Translation3d(1, 1, 1), new Rotation3d(0, 0, Math.PI / 2));
private final DoubleArrayEntry m_cameraToObjectEntry;
private final double[] m_defaultVal = {0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0};
private final Field2d m_fieldSim = new Field2d();
private final Field2d m_fieldApproximation = new Field2d();
/* Here we use DifferentialDrivePoseEstimator so that we can fuse odometry readings. The
numbers used below are robot specific, and should be tuned. */
private final DifferentialDrivePoseEstimator m_poseEstimator =
@@ -65,11 +98,21 @@ public class Drivetrain {
// Gains are for example purposes only - must be determined for your own robot!
private final SimpleMotorFeedforward m_feedforward = new SimpleMotorFeedforward(1, 3);
// Simulation classes
private final AnalogGyroSim m_gyroSim = new AnalogGyroSim(m_gyro);
private final EncoderSim m_leftEncoderSim = new EncoderSim(m_leftEncoder);
private final EncoderSim m_rightEncoderSim = new EncoderSim(m_rightEncoder);
private final LinearSystem<N2, N2, N2> m_drivetrainSystem =
LinearSystemId.identifyDrivetrainSystem(1.98, 0.2, 1.5, 0.3);
private final DifferentialDrivetrainSim m_drivetrainSimulator =
new DifferentialDrivetrainSim(
m_drivetrainSystem, DCMotor.getCIM(2), 8, kTrackWidth, kWheelRadius, null);
/**
* Constructs a differential drive object. Sets the encoder distance per pulse and resets the
* gyro.
*/
public Drivetrain() {
public Drivetrain(DoubleArrayTopic cameraToObjectTopic) {
m_gyro.reset();
// We need to invert one side of the drivetrain so that positive voltages
@@ -85,6 +128,21 @@ public class Drivetrain {
m_leftEncoder.reset();
m_rightEncoder.reset();
m_cameraToObjectEntry = cameraToObjectTopic.getEntry(m_defaultVal);
try {
m_objectInField =
AprilTagFieldLayout.loadFromResource(AprilTagFields.k2022RapidReact.m_resourceFile)
.getTagPose(0)
.get();
} catch (IOException e) {
e.printStackTrace();
throw new RuntimeException();
}
SmartDashboard.putData("Field", m_fieldSim);
SmartDashboard.putData("FieldEstimation", m_fieldApproximation);
}
/**
@@ -115,16 +173,109 @@ public class Drivetrain {
setSpeeds(wheelSpeeds);
}
/**
* Computes and publishes to a network tables topic the transformation from the camera's pose to
* the object's pose. This function exists solely for the purposes of simulation, and this would
* normally be handled by computer vision.
*
* <p>The object could be a target or a fiducial marker.
*
* @param objectInField The object's field-relative position.
* @param robotToCamera The transformation from the robot's pose to the camera's pose.
* @param cameraToObjectEntry The networktables entry publishing and querying example computer
* vision measurements.
*/
public void publishCameraToObject(
Pose3d objectInField,
Transform3d robotToCamera,
DoubleArrayEntry cameraToObjectEntry,
DifferentialDrivetrainSim drivetrainSimulator) {
Pose3d robotInField = new Pose3d(drivetrainSimulator.getPose());
Pose3d cameraInField = robotInField.plus(robotToCamera);
Transform3d cameraToObject = new Transform3d(cameraInField, objectInField);
// Publishes double array with Translation3D elements {x, y, z} and Rotation3D elements {w, x,
// y, z} which describe
// the cameraToObject transformation.
double[] val = {
cameraToObject.getX(),
cameraToObject.getY(),
cameraToObject.getZ(),
cameraToObject.getRotation().getQuaternion().getW(),
cameraToObject.getRotation().getQuaternion().getX(),
cameraToObject.getRotation().getQuaternion().getY(),
cameraToObject.getRotation().getQuaternion().getZ()
};
cameraToObjectEntry.set(val);
}
/**
* Queries the camera-to-object transformation from networktables to compute the robot's
* field-relative pose from vision measurements.
*
* <p>The object could be a target or a fiducial marker.
*
* @param objectInField The object's field-relative pose.
* @param robotToCamera The transformation from the robot's pose to the camera's pose.
* @param cameraToObjectEntry The networktables entry publishing and querying example computer
* vision measurements.
*/
public Pose3d objectToRobotPose(
Pose3d objectInField, Transform3d robotToCamera, DoubleArrayEntry cameraToObjectEntry) {
double[] val = cameraToObjectEntry.get();
// Reconstruct cameraToObject Transform3D from networktables.
Translation3d translation = new Translation3d(val[0], val[1], val[2]);
Rotation3d rotation = new Rotation3d(new Quaternion(val[3], val[4], val[5], val[6]));
Transform3d cameraToObject = new Transform3d(translation, rotation);
return ComputerVisionUtil.objectToRobotPose(objectInField, cameraToObject, robotToCamera);
}
/** Updates the field-relative position. */
public void updateOdometry() {
m_poseEstimator.update(
m_gyro.getRotation2d(), m_leftEncoder.getDistance(), m_rightEncoder.getDistance());
// Also apply vision measurements. We use 0.3 seconds in the past as an example -- on
// a real robot, this must be calculated based either on latency or timestamps.
m_poseEstimator.addVisionMeasurement(
ExampleGlobalMeasurementSensor.getEstimatedGlobalPose(
m_poseEstimator.getEstimatedPosition()),
Timer.getFPGATimestamp() - 0.3);
// Publish cameraToObject transformation to networktables --this would normally be handled by
// the
// computer vision solution.
publishCameraToObject(
m_objectInField, m_robotToCamera, m_cameraToObjectEntry, m_drivetrainSimulator);
// Compute the robot's field-relative position exclusively from vision measurements.
Pose3d visionMeasurement3d =
objectToRobotPose(m_objectInField, m_robotToCamera, m_cameraToObjectEntry);
// Convert robot pose from Pose3d to Pose2d needed to apply vision measurements.
Pose2d visionMeasurement2d = visionMeasurement3d.toPose2d();
// Apply vision measurements. For simulation purposes only, we don't input a latency delay -- on
// a real robot, this must be calculated based either on known latency or timestamps.
m_poseEstimator.addVisionMeasurement(visionMeasurement2d, Timer.getFPGATimestamp());
}
/** This function is called periodically during simulation. */
public void simulationPeriodic() {
// To update our simulation, we set motor voltage inputs, update the
// simulation, and write the simulated positions and velocities to our
// simulated encoder and gyro.
m_drivetrainSimulator.setInputs(
m_leftGroup.get() * RobotController.getInputVoltage(),
m_rightGroup.get() * RobotController.getInputVoltage());
m_drivetrainSimulator.update(0.02);
m_leftEncoderSim.setDistance(m_drivetrainSimulator.getLeftPositionMeters());
m_leftEncoderSim.setRate(m_drivetrainSimulator.getLeftVelocityMetersPerSecond());
m_rightEncoderSim.setDistance(m_drivetrainSimulator.getRightPositionMeters());
m_rightEncoderSim.setRate(m_drivetrainSimulator.getRightVelocityMetersPerSecond());
m_gyroSim.setAngle(-m_drivetrainSimulator.getHeading().getDegrees());
}
/** This function is called periodically, no matter the mode. */
public void periodic() {
updateOdometry();
m_fieldSim.setRobotPose(m_drivetrainSimulator.getPose());
m_fieldApproximation.setRobotPose(m_poseEstimator.getEstimatedPosition());
}
}

View File

@@ -5,12 +5,18 @@
package edu.wpi.first.wpilibj.examples.differentialdriveposeestimator;
import edu.wpi.first.math.filter.SlewRateLimiter;
import edu.wpi.first.networktables.DoubleArrayTopic;
import edu.wpi.first.networktables.NetworkTableInstance;
import edu.wpi.first.wpilibj.TimedRobot;
import edu.wpi.first.wpilibj.XboxController;
public class Robot extends TimedRobot {
private final NetworkTableInstance m_inst = NetworkTableInstance.getDefault();
private final DoubleArrayTopic m_doubleArrayTopic =
m_inst.getDoubleArrayTopic("m_doubleArrayTopic");
private final XboxController m_controller = new XboxController(0);
private final Drivetrain m_drive = new Drivetrain();
private final Drivetrain m_drive = new Drivetrain(m_doubleArrayTopic);
// Slew rate limiters to make joystick inputs more gentle; 1/3 sec from 0 to 1.
private final SlewRateLimiter m_speedLimiter = new SlewRateLimiter(3);
@@ -22,6 +28,16 @@ public class Robot extends TimedRobot {
m_drive.updateOdometry();
}
@Override
public void simulationPeriodic() {
m_drive.simulationPeriodic();
}
@Override
public void robotPeriodic() {
m_drive.periodic();
}
@Override
public void teleopPeriodic() {
// Get the x speed. We are inverting this because Xbox controllers return