Opencpn backup of multiple fixes using Kalman filter

[rgleason]

This question came from an ex Sub Captain friend who uses Opencpn on MAC now.

"I would like to rig this laptop to enable it to back up the Raymarine. The Raymarine has worked perfectly up to this point, but having a backup is important. During the crossing the GPS (all GPS systems) went off line for about 2-3 minutes during which we all held our breath. This means I should think about getting celestial navigation back on the scene. Inertial navigation systems have a kalman filter algorithm which permits entry of multiple fixes from different sources and times that essentially smooth the output DR position and improve the position accuracy. Something to think about for Opencpn?"

I know with all efforts towards the beta version, this question is a backburner, but it would be interesting to understand this better. - Any possibility? I know Sean was doing some of this in Celestial Nav, and the next step was related to translation of fixes for more accuracy.

[bcn]

Originally Posted by rgleason

This question came from an ex Sub Captain friend who uses Opencpn on MAC now.


I know with all efforts towards the beta version, this question is a backburner, but it would be interesting to understand this better. - Any possibility? I know Sean was doing some of this in Celestial Nav, and the next step was related to translation of fixes for more accuracy.

That goes for Dead-Reckoning with predictive positions where the different inputs to determine position from different sensors (GPS, accelerometers, gyros, speed-log, magnetometers) are weighted in relation with the predicted and the measured result, giving a new/next predicted position.
INS (Inertial measurement systems) are working this way. Without sensor input the system can make it's "best estimate" over several cycles. Precision will decrease with time of course.
And yes this is implemented with Kalman filters.

A simple DR might be a first step:
the course predictor has already the three parameters 1. Pos. 2. COG 3. SOG (4. Time since last fix)
At GPS outage this can drop in, with the possibility to edit the parameters manually (having a manual fix, a manual log, a compass...or a handheld GPS not connected to the system.)

Interesting enough there has not been any feature request for DR up to now (at least I've seen any).

Added now as
FS#1584 - Implement Dead Reckoning as fall back at GPS outage

[petter5]

Originally Posted by bcn

That goes for Dead-Reckoning with predictive positions where the different inputs to determine position from different sensors (GPS, accelerometers, gyros, speed-log, magnetometers) are weighted in relation with the predicted and the measured result, giving a new/next predicted position.
INS (Inertial measurement systems) are working this way. Without sensor input the system can make it's "best estimate" over several cycles. Precision will decrease with time of course.
And yes this is implemented with Kalman filters.

A simple DR might be a first step:
the course predictor has already the three parameters 1. Pos. 2. COG 3. SOG (4. Time since last fix)
At GPS outage this can drop in, with the possibility to edit the parameters manually (having a manual fix, a manual log, a compass...or a handheld GPS not connected to the system.)

Interesting enough there has not been any feature request for DR up to now (at least I've seen any).

Added now as
FS#1584 - Implement Dead Reckoning as fall back at GPS outage

Interesting. I was doing a quick search and found that there is small boards that does this in hardware: http://media.digikey.com/pdf/Data%20...RM4000L_DS.pdf that can be connected with the GPS with USB or RS232 talking NMEA0183

[bcn]

Originally Posted by petter5

Interesting. I was doing a quick search and found that there is small boards that does this in hardware: http://media.digikey.com/pdf/Data%20...RM4000L_DS.pdf that can be connected with the GPS with USB or RS232 talking NMEA0183

700$ each
Haven't seen data for output format, output data, precision and most important degradation of output over time.

[petter5]

Yes, I agree that the cost is high, on the other side you will find this functionality in a wheelmark approved GPS compass witch you will have to pay $ 3000+ for so it could be a cheaper solution to get this functionality for $700,, a software solution for this in OpenCpn would be great, but without the sensors to give input to the system you will never get a useful prediction except in situations where the boat stays on a steady course.

[bcn]

The outlined idea for the DR is really very basic and rough.

In case of of a compass reading this wouldn't be COG in reality but heading, so drift and set would be needed as well.

But to start...

[rgleason]

Some have been using this

CMPS10 - Tilt Compensated Magnetic Compass

Also I found another very inexpensive sensor and I can't find the post now. I will find it later.
Wouldn't these help with the hardware part?

[rgleason]

Well, I found Sean in the middle of it with this post, inertial system & autopilot
http://www.cruisersforum.com/forums/...ml#post1681333

Still can't find that post that I had with a very small device that seemed to do it all, something like 7 sensors.... it was not that long ago. I was amazed at how small it was.

[rgleason]

Found it. http://www.cruisersforum.com/forums/...it-135563.html

[rgleason]

Some very interesting ideas here. Don't think my friend actually was thinking of an inertial guidance system as on a sub, just a tool for taking fixes from various sources to arrive at a better position when GPS goes down, but it appears the inertial part could be a part of it.

[dacust]

Originally Posted by rgleason

Found it. http://www.cruisersforum.com/forums/...it-135563.html

I have this one. Haven't done anything with it, yet. Arduino code is available.

10 DOF Mems IMU Sensor

-dan

[bcn]

All this is ok - and the output of this sensors will be a NMEA message with position, COG, SOG and perhaps heading (= you know yaw), pitch and roll.
Or a proprietary message on serial or some other channel like i2c.

The point Rick's friend made (for me) is: and what happens if your NMEA network or main network or navigation systems goes down (and the fancy inertial system with it) but you have your bare OCPN as a backup still running. Can we have then DR?
And that would be interesting indeed.

Hubert

[PjotrC]

Originally Posted by bcn

what happens if your NMEA network or main network or navigation systems goes down (and the fancy inertial system with it) but you have your bare OCPN as a backup still running. Can we have then DR?

With manual estimate of COG/SOG, if one insists on all other automation in chart handling... a GPS simulator feeding OCPN does just DR. Not easy on a submerged vessel, but so is astro then

[bcn]

PjotrC,

great - this simplified from the simulation requirements as the COM-ports (and without Altitude) might pop-up when the GPS is off and one wants to go DR.

So how to get this in OpenCPN?

Hubert

[rgleason]

There are some interesting DR alternatives suggested above.

My friend wanted a backup when and if GPS failed, and made this statement and question

Inertial navigation systems have a kalman filter algorithm which permits entry of multiple fixes from different sources and times that essentially smooth the output DR position and improve the position accuracy. Something to think about for Opencpn?"

Bilgin's Blog | Kalman Filter for Dummies
Kalman filter - Wikipedia, the free encyclopedia
http://biorobotics.ri.cmu.edu/papers...man_basics.pdf
Greg Czerniak's Website - Kalman Filters for Undergrads 1
http://www.cl.cam.ac.uk/~rmf25/paper...n%20Filter.pdf

If he is thinking of being able to enter time + position fixes (from known position, calculated DR, celestial sights, etc) a kalman filter calculator might be useful.

The next step might be the ability to add some other estimated unknown like current, wind, waves, etc).

The first step would be totally separate from any "inertial system" information.

The problem is similar to solving multiple fixes in celestial I think.

Maybe I should ask him to clarify his question?
--Actually he is a surface sailor now, but he has a good submariner's DR drill using yards & knots.