morgil / madgwick_py Goto Github PK

I am trying to learn motion processing and have managed to write myself a driver for the MPU9250 9DOF sensor and now I need to implement Madgwick's filter to get accurate orientation. I see your MadgwickAHRS class with the update method but don't understand how the method returns the result.

Using Python 3.9 and the Quaternion object I'm getting a KeyError. self.dict returns an object, not an array.

I'm mostly writing this to lets someone else feel that they are not alone.

Thanks

Hi,
Looks like there is a coding error with Euler123 output, setting different initiation and outputing without updating AHRS gives invalid data

refq = Quaternion(1,0,0,0) # look front
Euler:  (0.0, 0.0, 0.0)
Eu123:  (0.0, 0.0, 0.0)
Quat:   [1 0 0 0]

refq = Quaternion.from_angle_axis(np.pi/2,0,0,1) # look front, sideways
Euler:  (0, 1.5707963267948966, 0.0)
Eu123:  (-0.0, 0.0, 1.5707963267948963)
Quat:   [ 0.70710678  0.          0.          0.70710678]

refq = Quaternion.from_angle_axis(np.pi/2,1,0,0) # look up
Euler:  (1.5707963267948963, 0.0, 0.0)
Eu123:  (1.5707963267948963, 1.5707963267948966, -0.0)
Quat:   [ 0.70710678  0.70710678  0.          0.        ]

refq = Quaternion.from_angle_axis(np.pi/2,0,1,0) # look right
Euler:  (0.0, 0.0, 1.5707963267948963)
Eu123:  (-0.0, 0.0, -0.0)
Quat:   [ 0.70710678  0.          0.70710678  0.        ]

Also discussed here:
RiftCat/vridge-api#17

how to fix it?

I've been trying to implement the AHRS with the sensors from the PlayStationVR headset, and getting confused about which axis should be which.

From reading the sensor values, as arrays, I have:

Structure of the PSVR sensor:
Gyros (when wearing headset normally).
1 - yaw, +ve right hand forward
2 - pitch, +ve look up
3 - roll, +ve right hand down
Accelerometers, largest when 'x' is highest point
1 - top+, bottom-
2 - right+, left-
3 - front+, back-

And on the other end, when the rendering the scene from the output of Madgewick I know that the following are true (tested by setting during init and not updating the AHRS):

Quaternion that points directly forward
refq = Quaternion((1,0,0,0))

Quaternion pointing/pitched directly up
refq = Quaternion.from_angle_axis(np.pi/2,1,0,0)

No matter how I've tried swapping various axis I have not yet reached a working system, it either 'moves' incorrectly or the Quaternion ends up pointing in the wrong direction....

Can someone/any provide a reference for how the axis should be arranged?

https://arxiv.org/pdf/1704.06053.pdf seems to suggest that it is.
This would be an awesome feature!

What's the right measurement units for accelerometer and magnetometer data? It's mentioned that gyroscope data should be in radians per seconds. How about others?

An example would be great. Thanks for the repo.