Ublox Neo-M8N (The one with the built-in compass)

[Brian DeBusk]

I'm really impressed with the Ublox Neo-M8 in general.  It picks-up lots of satellites and seems to have great reception.

I put it in the dynamic airborne model, but have left to reporting rate at 1 Hz.  I didn't change any other parameters either.

What I did notice was it has a built-in EKF algorithm.  Does anyone have any experience with GPS+Compass based results?  I don't mind doing my own testing, but I wonder if anyone has already worked with an M8N specifically.
 

[Steve McLaughlin]

The M8N chip does not have a built in compass, it's added to the PCB that it sits on.  Yes, the ublox has an internal kalman filter that blend GNSS measurements with a very simple, user selectable dynamic model, but it does not use compass data as an input.

There are ublox modules made for automotive navigation that have additional INS like features but not in the M8N:

http://www.ublox.com/en/dead-reckoning.html

[Brian DeBusk]

Fair enough.  I should have specified that I'm using the integrated module from CSG:  http://www.csgshop.com/product.php?id_product=173

I'm still at a loss as to what the M8's Kalman filter could even do.  The GPS natively receives the data it needs to do a position estimate, and infers velocity and heading based on changes in location.  But where is the innovation necessary to improve the estimate of the state? 

I guess the Kalman filter could be estimating the velocity based on location updates, but I'd think that could be done much more easily (and accurately) by just estimating the second-by-second velocity and low-pass filtering the result.

[Jonathan Challinger]

GPS takes multiple measurements (pseudo- and delta- ranges from each satellite) and needs to fuse them into states (3d position and velocity) while being robust to things like multipathing. Thus, Kalman filter.

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Brian DeBusk]

The M8 ships with the Kalman filter turned off.  I haven't downloaded the APM-configured Ublox configuration, but I would think it also has the Kalman filter disabled since it isn't one of the few parameters that have to be changed to make an M8 work with the 3DR version.

Does anyone know for sure if the 3DR Ublox's internal Kalman filter is disabled?

[Marco Robustini]

M8 in action with Pixhawk... good watching!

https://www.youtube.com/watch?v=i5q1iM6_37U

- Marco

[Jonathan Challinger]

As I said, a Kalman filter is an algorithm for estimating states from measurements in a statistically optimal way. A GPS must do that, because it has multiple measurements from each satellite. It would be absolutely wonderful if we could get each of those individual measurements and fuse them in our EKF. That would mean that the sensor data from the IMU and other sensors would be used to truth each GPS satellite and reject ones that don't correlate to the motion that we're measuring with other sensors.

So, the GPS does have a kalman filter internally and there is nothing you can do about that. However, on some ublox GPSes, that kalman filter can also accept a yaw gyro and speedometer measurement in order to do dead reckoning on ground vehicles that can be assumed to be driving forwards. So, a solution to keep GPS lock in tunnels. I think this is the feature that is confusing you.

[Brian DeBusk]

The U-Center software clearly has some sort of additional Kalman filter that incorporates Gyroscopic and Temperature measurements.  It defaults to "off" on a freshly-shipped U-Blox and I am virtually certain the 3DR version doesn't enable it either.

My original question was if this may be a better level to implement the Kalman or if gyros input would somehow improve the heading measurements from the GPS.  It's been my personal experience that headings measurements from a GPS are really terrible, unless you're willing to wait several seconds for them to stabilize.

[Jonathan Challinger]

Yes, that is the ground vehicle dead reckoning.

--

[Jonathan Challinger]

That would not help for an aerial vehicle because it makes an assumption about which directions the vehicle can travel.

On Oct 18, 2014 3:44 PM, "Brian DeBusk" <brianc...@gmail.com> wrote:

--