Quaternion Drift Problem
Hi Folks,
I'm having a problem with the set of MMC sensors I have drifting when used at higher speeds. We are using 3 sensors on each leg to get joint angle information and animate an avatar's legs that follow the user's own legs as closely as possible. Here is a video of it working as intended:
When I move to the treadmill in the lab and start walking at higher speeds, we encounter a problem with drift of the quaternions as is demonstrated in the other video below:
As soon as I step onto the sides of the treadmill and remain stationary, the correct orientation is quickly returned to, but it is a problem for us that we can't track the motion of legs that are walking quickly (ideally we would even like to be able to capture jogging accurately). I have tried using the NDOF mode on the sensors instead of IMU_PLUS but the problem persisted. Any suggestions while I troubleshoot?
Thanks
Comments
Quantify "higher speeds".
What are your sensor fusion settings?
I would have to do further testing to be able to quantify this. I just noticed that as I increased the speed on the treadmill the drift got quite bad as you can see in the second video.
My sensor fusion is set up as follows (note that I tried switching to NDOF but it didn't help so I changed back):
sensorFusion1.configure()
.mode(SensorFusionBosch.Mode.IMU_PLUS)
.accRange(SensorFusionBosch.AccRange.AR_2G)
.gyroRange(SensorFusionBosch.GyroRange.GR_250DPS)
.commit();
I'm going to try to up the Gyro speed and see if that helps.
Also don't forget to calibrate: https://mbientlab.com/tutorials/Calibration.html
Increasing the Acc/Gyro ranges did the trick. Not sure why I didn't think to try this before posting. If anyone else is having a similar issue make sure your ranges are set to be large enough for your application. I actually ended up using this site (https://musculoskeletalkey.com/biomechanics-2/) to find the expected angular velocities of joint angles for human gait and went from there. The application is working very well now.