tf API: Axes/order of rotation

Question

Hello ROS users!

I'm working with tf and I've a few questions about the API.

I am defining frames such that Yaw is applied around Z, followed by Pitch around the new Y and then Roll around new X.

Consider a transform from base_link --> camera_frame. Assuming I only have constant rotation between the frames (zero translation), here's how I create the transform:

//Fixed transform values
double roll, pitch, yaw;
yaw = DEGREES_TO_RADIANS(90); // 90 degrees around Z 
pitch = DEGREES_TO_RADIANS(0); // 0 degrees around new Y
roll = DEGREES_TO_RADIANS(-90); // -90 degrees around new X

//Create Quaternion from Euler angles
tf::Quaternion quat;
quat.setRPY(roll, pitch, yaw);
//quat.setEuler(pitch, roll, -yaw); // 

//Set rotations
geometry_msgs::TransformStamped bLink_to_cam;

bLink_to_cam.transform.rotation.x = quat.x();
bLink_to_cam.transform.rotation.y = quat.y();
bLink_to_cam.transform.rotation.z = quat.z();
bLink_to_cam.transform.rotation.w = quat.w();

//Set translations
bLink_to_cam.transform.translation.x = 0;
bLink_to_cam.transform.translation.y = 0;
bLink_to_cam.transform.translation.z = 0;

In another node, I have a tf::TransformListener object which receives this transform. Here I print out the values :

void printEulerAngles(tf::StampedTransform& transform){
    double roll, pitch, yaw;
    tf::Matrix3x3 m(transform.getRotation());

    m.getRPY(roll, pitch, yaw);
    //m.getEulerZYX(yaw, pitch, roll);
    ROS_INFO("\nRoll    : %.3f\n"
            "Pitch  : %.3f\n"
            "Yaw    : %.3f\n",
            RADIANS_TO_DEGREES(roll), RADIANS_TO_DEGREES(pitch), RADIANS_TO_DEGREES(yaw)  );

}

I expected to see the transform as defined above. However, I see Roll, Pitch, Yaw (90,90, 0). I did some research and found that setRPY() and getRPY() follow this convention:

• (roll around X, pitch around Y and yaw around z) around fixed axes.

Here are my questions:

1. How do I define rotations when I need to rotate (w.r.t to base_link axes) in the following convention?

• (yaw around Z, then pitch around new Y, then roll around new X)

2. Once I get the above rotation from tf, how do I modify the printEulerAngles() method to display them in the same convention as above?

I'd really appreciate any assistance in understanding tf. Thanks!

---

Originally posted by Venkat Ganesh on ROS Answers with karma: 57 on 2016-11-10

Post score: 1

Answer 1

I think you are looking for the setEulerYPR and getEulerYPR functions of tf::Matrix3x3. As far as I can tell, these follow the convention you are describing.

To generate the quaternion:

//Fixed transform values
double roll, pitch, yaw;
yaw = DEGREES_TO_RADIANS(90); // 90 degrees around Z 
pitch = DEGREES_TO_RADIANS(0); // 0 degrees around new Y
roll = DEGREES_TO_RADIANS(-90); // -90 degrees around new X

//Create Matrix3x3 from Euler Angles
tf::Matrix3x3 m_rot;
m_rot.setEulerYPR(yaw, pitch, roll);

// Convert into quaternion
tf::Quaternion quat;
m_rot.getRotation(quat);

Your printEuler would then be

void printEulerAngles(tf::StampedTransform& transform){
    double roll, pitch, yaw;
    tf::Matrix3x3 m(transform.getRotation());
    int solution_number = 1;

    m.getEulerYPR(yaw, pitch, roll, solution_number);
    ROS_INFO("\nRoll    : %.3f\n"
            "Pitch  : %.3f\n"
            "Yaw    : %.3f\n",
            RADIANS_TO_DEGREES(roll), RADIANS_TO_DEGREES(pitch), RADIANS_TO_DEGREES(yaw)  );
}

However, like I wrote in your previous question, there can be multiple representations of an identical rotation, so your angles might not be the same. You can play around witht the solution_number argument of getEulerYPR() but I don't know if that will help.

Good luck.

---

Originally posted by rbbg with karma: 1823 on 2016-11-14

This answer was ACCEPTED on the original site

Post score: 4

---

Original comments

Comment by Venkat Ganesh on 2016-11-14:
Hello, thanks for your suggestion. Unfortunately, I still don't see the correct angle. In tf_broadcaster_node, I have the same input: Y,P,R = (90,0,-90). However, in tf_listener_node, I see (0, 90, 90). I don't think it's an equivalent rotation.