Hello,
I am doing some testing with the beta 5 and altitude, using gps as the primary source for alt and also using gps for yaw, so that forces me to use EKF3.
I don’t think anybody has done comparisons yet of GPS-for-yaw vs GSF but I would expect GPS-for-yaw to do a better value as long as the GPSs are reporting good positions.
So here is a graph of what I think you’re looking at. So this is the GPS alt vs the EKF alt (I’ve graphed POS.Alt because this version of the EKF alt is expressed as the alt-above-sea-level).
So you’re seeing oscillations of about 12cm or so that we don’t see in the GPS. I’ll ask @priseborough if he has any idea. The altitude should be simiply coming from the accelerometers + GPS.
Gotcha, I’m really down to test GSF feature in this week, i need just one help, I get it that GSF feature will automatically takes over compass for yaw measurement when compass becomes unhealthy or start getting high compass variance (XKF4.SM), but I don’t know how high compass variance we are talking about before GSF comes into picture ? Is it when compass variance crosses FS_EKF_THRESH value ?
If yes, then I like to fiddle with that condition, I want GSF takes over compass much earlier like when compass variance will be more than 0.3 for more than 3 seconds, can you please help me where in AP_NavEKF or AP_NavEKF3 this condition for switching to GSF is getting checked ?
Re GSF taking over more quickly, I actually don’t know where the code is that it uses to make the decision but I think it requires 3 of the 5 mini EKFs to agree on the heading within some threshold but then it also has a hardcoded minimum time I think of a couple of seconds. Those thresholds will all be within AP_NavEKF3 somewhere but I don’t know where. The FS_EKF_THRESH is only used for the EKF failsafe which is really more “vehicle” code so this particular parameter is not involved.