The order of matrix multiplication is different in ign-math4 and ign-math6

[erdisayar]

Hello,
I tested matrix multiplication operator in both math4 and math6 library and I obtained 2 different results.

The the first test is in Gazebo9(Ubuntu 18.04) using math4 library. Here is my result.

I think it is wrong because it is not the calculation of MatrixA x MatrixB instead MatrixB x Matrix A. However the intention is MatrixA x Matrix B

If I test the same code in Gazebo11(Ubuntu 20) using math6 library. Here is the result

The result is correct.

summary from my investigation is(maybe I am wrong)

If I want to calculate the matrix multiplication of AB,

• math4 library calculates BA instead of AB
• math6 library calculates AB

Here is the code in the picture.

std::cout << "Matrix A " << std::endl;
WORLD_POSE MatrixA = WORLD_POSE(1, 0, 0, 0, 0, 0);
std::cout << "Matrix A Pos:" << MatrixA.Pos() << std::endl;
std::cout << "Matrix A Rot:" << MatrixA.Rot() << std::endl;
std::cout << "Matrix B" << std::endl;
WORLD_POSE MatrixB = WORLD_POSE(0, 0, 0, 0, 0, 1);
std::cout << "Matrix B Pos:" << MatrixB.Pos() << std::endl;
std::cout << "Matrix B Rot:" << MatrixB.Rot() << std::endl;
std::cout << "----------" << std::endl;
std::cout << "Matrix C = A * B " << std::endl;

WORLD_POSE MatrixC = MatrixA * MatrixB;
std::cout << "Matrix C Pos:" << MatrixC.Pos() << std::endl;
std::cout << "Matrix C Pos:" << MatrixC.Rot() << std::endl;
std::cout << "----------" << std::endl;

[erdisayar]

Here is the image for comparison

[chapulina]

I wrote a quick test to check Matrix4 multiplication in #191 and I couldn't see any inconsistencies between ign-math4 and ign-math6.

Note that your example is using ignition::math::Pose3d, not ignition::math::Matrix4d. The behaviour of the * operator for Pose3 has been changed (fixed) from ign-math4 to ign-math5, as documented on the migration guide. I believe that's the underlying reason for what you're seeing.

[erdisayar]

Thanks for enlightening the issue. What kind of behavior did you implemented before (for ign-math4) ? and what was the its logic?

As you see from my calculations for ign-math4 , the behavior I saw is basically same as reverse order matrix multiplication for ignition::math::Pose3d.

In other words, What was the use case before?

[scpeters]

I don't know why it was this way. You can read more of the history in #60 and bitbucket PR 301. I'm closing since this is not a bug, but feel free to continue the discussion if you like