UAV toolbox support package for PX4 : local position estimation problem
6 次查看(过去 30 天)
显示 更早的评论
Hello,
I am trying to estimate my vehicle's local position using EKF algorithm in PX4 software.
To evaluate the accuracy of the estimator, I put my drone stationary and estimate my local position.
I used vehicle_local_position.msg uorb message to read my vehicle local position.
When I read my vehicle local position, it exactly follows the data of gps as you can see beneath
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/943414/image.jpeg)
First graph shows the local position estimation of x, and second graph shows the raw latitude data of gps. Third graph shows the covariance as time passes. As you can see, the covariance converges well as time passes but vehicle local position estimation follows gps exactly.
This is strange because EKF fuses IMU and gps, therefore local position estimate and gps data should be somewhat different.
I speculated this is because I put the noise of gps(covariance of gps noise) too low.
Therefore, I changed px4 parameter gps_noise_p from 0.5 to 10m, which increases the covariance of gps noise and therefore not believe gps data much, and expected not to follow the gps data.
However, the data still followed the gps data as you can see beneath.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/943419/image.jpeg)
As you can see, it still follows the gps data exactly though the covariance increases from 0.2 to 1.
Can you help me why vehicle local position follows exactly gps data? Is it because drone is stationary and mean value of IMU is 0 so follow the gps data exactly?
0 个评论
采纳的回答
Ankur Bose
2022-4-1
We recommend looking into the PX4 EKF2 estimator code to evaluate why you are seeing the tight coupling between GPS and estimated value. The estimation and fusion depends on a lot of parameters and PX4 has a way of picking/rejecting sensor values it does not find suitable or appropriate
0 个评论
更多回答(0 个)
另请参阅
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!