No usable results when using insEKF() instead of imufilter()
1 次查看(过去 30 天)
显示 更早的评论
Part of what I want to do is fuse gyroscope and accelerometer data to get an orientation estimate. When I fuse the data using the imufilter() I get results looking very similar to what I am expecting, without majorly tweaking the noise settings and so on.
I then proceeded to try to fuse it using the insEKF() because I eventually also need to fuse GPS measurements. For the first part I only added the gyroscope and the accelerometer to see the orientation estimate. I coded it very similar to the example in the documentation but for some reason the estimated orientation just start spinning and jumping around "uncontrollably".
Heres part of how Ive implemented it:
I also tried tuning it using the data I got from the other filter but that not seems like the issue. Also fusing the data sequentially using fuse() didnt change anything.
...
gyrofix=[gyro(:,2) gyro(:,3) gyro(:,1)]; %data already in workspace
acclfix=[-accl(:,2) -accl(:,3) -accl(:,1)];
option = insOptions(ReferenceFrame="ENU");
insAccel=insAccelerometer;
insGyro=insGyroscope;
tt_meas=timetable(acclfix,gyrofix,'RowTimes', seconds(0:0.01:(size(acclfix,1)/100)-0.01))
tt_meas.Properties.VariableNames={'Accelerometer' 'Gyroscope'}
tt_meas.Accelerometer(:,:)=0;
filt = insEKF(insAccel, insGyro, option);
initOrient=quaternion(1, 0, 0, 0);
stateparts(filt,"Orientation",compact(initOrient));
statecovparts(filt,"Orientation",0.01);
mnoise=tunernoise(filt);
mnoise.AccelerometerNoise=0.0002;
mnoise.GyroscopeNoise=9e-5;
untunedEst = estimateStates(filt,tt_meas,mnoise);
8 个评论
Paul
2024-7-24
Hi Christian,
Have you made any progress on this problem, with either imufilter or insEKF, with the data posted in sensor_measurments.mat?
回答(1 个)
Brian Fanous
2024-7-16
Christian, do you have a ground truth recorded? Meaning, do you have logged data of what you want the filter output to look like? If so, you can try the tune() method on the insEKF to optimize the noise parameters. (You'll still have to handle initializing the filter states correctly).
If you are happy with the results from the imufilter, you could use that as your ground truth orientation. Be aware though that you'll be tuning the insEKF to also match the noise in the output of the imufilter in that case.
另请参阅
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!