search

ethz-asl / kalibr

The Kalibr visual-inertial calibration toolbox
Other
4.39k stars 1.41k forks source link

Contradiction between frames in the paper and the readings of the sensors #465

Closed mhyoosefian closed 3 years ago

mhyoosefian commented 3 years ago

Hi. Thanks for the contribution.

I am a bit confused about the various frames of the EuRoC sensor setup. In the EuRoC paper the frames are described as: X_I = -Y_C; Y_I = X_C; Z_I = Z_C (I is for the IMU0 and C is for the CAM0) which results in the following rotation matrix: R_IC =[0 -1 0; 1 0 0; 0 0 1] which is totally compatible with the calibration result provided in the yaml file of the camera. So, my question is that if the X axis of the IMU0 is upward, then the accelerometer reading in this axis (when holding the platform nearly vertical) should be about -9.8, but the actual readings (according to dataset) are about 9.8 which means the X axis is downward!

Any ideas are appreciated!

NikolausDemmel commented 3 years ago

Hm... I think if the accelerometer axis is pointing "up" and the IMU is standing still w.r.t. earth, a reading of 1g is expected, not -1g. Gravity is pulling down, but the corresponding acceleration is up. (Consider a rocket launching from earth with 1g acceleration. A person inside the rocket will feel 2g gravity pulling down.)

mhyoosefian commented 3 years ago

Hm... I think if the accelerometer axis is pointing "up" and the IMU is standing still w.r.t. earth, a reading of 1g is expected, not -1g. Gravity is pulling down, but the corresponding acceleration is up. (Consider a rocket launching from earth with 1g acceleration. A person inside the rocket will feel 2g gravity pulling down.)

Thanks for your reply. Well, consider a ball thrown upward, its acceleration is always 1g downward, so if a frame attached to the ball has say a z axis pointing up, the ball acceleration will be -1g in that frame, right? Am I missing something?

Regards

NikolausDemmel commented 3 years ago

It will be in free fall and the accelerometer will read 0. The way you arrive at -1g net acceleration w.r.t. earth is by subtracting the acceleration due to gravity, +1g, so you get 0 - 1g = -1g.

mhyoosefian commented 3 years ago

It will be in free fall and the accelerometer will read 0. The way you arrive at -1g net acceleration w.r.t. earth is by subtracting the acceleration due to gravity, +1g, so you get 0 - 1g = -1g.

So, I think one can write the equation: IMU_Reading = Object_Acceleration - Gravity in any frame. In the ball example, we can write for the z axis of the ball frame (which is upward) IMU_Reading = -1g - (-1g) = 0. And for the IMU case, if the IMU is at rest, its acceleration is zero, so, for the x axis which is upward, we can write: IMU_Reading = 0 - (-1g) = 1g. I think I confused the IMU_Reading with the Object_Acceleration.