Closed boosfelm closed 9 years ago
Thanks! I took your suggestion into this PR: https://github.com/PX4/Firmware/pull/2478
Since this is merged, closing, thanks!
great, happy to contribute. Yesterday we did 4 flights about 10-15 min each with master_release (+ our own attitude control module) in different modes, namely fw att control, fw altitude control, fw position control, loiter, mission and LPOS_Z was off by 0.2 to 1.9 m, which seems pretty good to me.
As discussed posting our tuning values for Sensor and Bias noise, which led to much better estimation results. Thanks @surberj for testing this. It has to be noted that this can be really aircraft specific so please be catious when testing. In our case the Pixhawk is embedded in a foam structure of a VTOL tailsitter design.
PE_ABIAS_PNOISE = 1.0e-5 PE_ACC_PNOISE = 0.05 PE_POSD_NOISE = 1.25 (estimator diverged at PE_POSD_NOISE = 3.0) PE_VELD_NOISE = 0.3 (estimator diverged at PE_VELD_NOISE = 0.05)
According to the data PE_POSD_NOISE = 2.0 and PE_VELD_NOISE = 0.15 look best, but we have not enough evidence. I would rather choose a trade-off between enhancing and staying near the working solution (which surely does not diverge), i.e, PE_POSD_NOISE = 1.25 PE_VELD_NOISE = 0.3
This might also help on #1902.