autopilot route plugin

[vladi58]

Originally Posted by rgleason

Sean has updated his Pypilot Autopilot Route 0.3.0 plugin. It compiles and opens in Opencpn, but I have not tested it.

Hi, just come along this


downloaded and installed plugin
however can not get it to work
documentation is very basic
i have raymarine s3 course computer with st6002 head unit
all other raymarine instruments connected over seatalk
have as well one multiplexer
this one here :


ShipModul Marine Electronics

ShipModul Maritieme Elektronica


over this one i have Emtrack AIS connected
and of course
PC with installed opencpn 4.88
all things working well
however i would like to use rm autopilot plugin as well



does anone know how to get raymarine autopilot plugin working ?

settings, any other recomandations ?



or how to get in touch with raymarine autopilot plugin developer Bernd Cirotzki ?


thank you
kind regards
vladi

[Douwe Fokkema]

Following message was originally published on "OpenCPN to Seatalk" thread. Fits better here.


Rick,


I also had a 4000 before on a previous boat. It had the same "track" behavior as the Evolution, difficult to start, steers wrong way, then exactly on track. But you often lack the space to allow is its wrong initial behavior. The EV1 based pilot is much better than the 4000, steers excellent on waves, really anticipating through its acceleration sensors. I feel the tracking behavior should be cured with our own software, like Sean's autopilot_route.


So I adapted the autopilot_route to handle the Evolution pilot. Sending heading commands through an Actisense NGT-1 (NOT the NGW-1) directly to the pilot. I earlier decoded these proprietary PGN's. Here I used the route position bearing mode. This worked quite well with short routes. With long routes (> 100 WPt's) Seans software can't handle it. It calculates the distance to every segment in the route every second, killing the system. Further I like the route handling of OpenCPN better than Sean's route handling. OpenCPN just focuses on the current segment and provides steering info for that. And I really like the functions Zero XTE and Activate Next WP. OpenCPN has it standard. But Seans handling at waypoint turns is better.
I now made a version that is completely based on the OpenCPN route handling. From the XTE as provided by OpenCPN it derives a next position on the route to steer to (idea of Sean). I might have to add an integration term for correction of cross currents. I have a first prototype (just) ready, sea trial tomorrow.

Douwe



Further remark: With OpenCPN routes, the user decides which WP is the active one. In Autopilot route it is the software deciding which WP is the next.

[Sirmikeyp]

What would you guys recommend for hardware that allows NMEA or seatalk communication between a ST4000 and a raspberry pi (openplotter)?

[seandepagnier]

Originally Posted by Douwe Fokkema

Following message was originally published on "OpenCPN to Seatalk" thread. Fits better here.


Rick,


I also had a 4000 before on a previous boat. It had the same "track" behavior as the Evolution, difficult to start, steers wrong way, then exactly on track. But you often lack the space to allow is its wrong initial behavior. The EV1 based pilot is much better than the 4000, steers excellent on waves, really anticipating through its acceleration sensors. I feel the tracking behavior should be cured with our own software, like Sean's autopilot_route.


So I adapted the autopilot_route to handle the Evolution pilot. Sending heading commands through an Actisense NGT-1 (NOT the NGW-1) directly to the pilot. I earlier decoded these proprietary PGN's. Here I used the route position bearing mode. This worked quite well with short routes. With long routes (> 100 WPt's) Seans software can't handle it. It calculates the distance to every segment in the route every second, killing the system. Further I like the route handling of OpenCPN better than Sean's route handling. OpenCPN

I did not do large routes, but some simple optimization would solve this.

just focuses on the current segment and provides steering info for that. And I really like the functions Zero XTE and Activate Next WP. OpenCPN has it standard. But Seans handling at waypoint turns is better.

A zero xte is easy to implement in the plugin.

I now made a version that is completely based on the OpenCPN route handling. From the XTE as provided by OpenCPN it derives a next position on the route to steer to (idea of Sean). I might have to add an integration term for correction of cross currents. I have a first prototype (just) ready, sea trial tomorrow.

Not sure how you did it, but the plugin supports several modes of steering. It would be a great addition if you just added a mode for your particular algorithm. This makes comparisons easy. The standard xte mode is meant to basically be the same as what is in opencpn by default, with a few improvements.

I also want to try to command the autopilot in compass mode, where opencpn calculates the compass course to follow the route rather than using the autopilot in gps mode. This could give interesting results especially depending on currents. As well it is possible to command the autopilot in wind mode (handle gusts better perhaps) following the route as additional improvements, and eventually even computing tacks/jibes in a boundary.

Further remark: With OpenCPN routes, the user decides which WP is the active one. In Autopilot route it is the software deciding which WP is the next.

There is an option for this in the plugin if you want to be prompted or not.

[Douwe Fokkema]

Hi Sean,

I am often sailing in confined water, narrow channels, shallow tidal waters, sandbanks. Here I need to follow routes precisely, independent of cross currents.

So I am not interested in a choice of steering algorithms, I just need one that will keep me on track in all circumstances. On top of that it should steer my Raymarine Evolution pilot directly and be controlled through the P70 control head at the steering position.

So first the steering algorithm. As I want to stay on the track, I need to minimize my XTE. So the steering has to be based on XTE. But you can not steer an XTE, you need to derive a heading from it. Base heading is the Bearing to Waypoint (BTW). On this a correction is applied. This correction is derived from the XTE (off track) to a point on the track some distance ahead. This functions like your waypoint position bearing method. Result is a course correction that is applied to the BTW. Steering the resulting heading you will get towards the track. But in case of cross currents it is not sufficient to base the heading correction on purely the XTE. We need to base the correction on a PID function of the XTE. So we integrate the XTE over time, giving IXTE. To improve stability (dampen oscillations) we need to add the derivative of the XTE: DXTE. So the function on which we base the course correction is
c1 * XTE + c2 * IXTE + c3 * DXTE. Here c1, c2 and c3 are tuning parameters that should be set to optimize the system. I have implemented this in https://github.com/douwefokkema/AutoTrackRaymarine_pi see specifically the Compute() function, including the best values for c1, c2, c3 that I managed to select.

With this method the boat will stay within 2 meters of the track, with some fluctuations dependent on the stability of the GPS. In sharp turns it may swing off track some meters more, independent of cross currents. I also tried (artificial) compass errors of up to 20 degrees. After some 2 minutes of stabilizing, the ship is on track again.

As the system is functioning on the route information from OpenCPN alone (no own calculations) we need a better precision in the XTE from O. Currently it is 3 decimals in nautical miles, not enough. So I had to modify O, sending 5 decimals now.

The system can be controlled from the Raymarine P70 control head. Pressing Auto again when the pilot is in Auto mode, the plugin switches to Track mode and takes control by sending headings to the pilot every second. It will follow complex routes for many miles and hundreds of waypoints within meters from the active route in O. Pressing a + or - 1 or 10 button, will make the pilot switch back to Auto. Standby button will of course set system to manual steering again.
When the Tracking function is started the plugin will do a ZeroXTE in OpenCPN, so it will steer straight to the next waypoint from the current position. This required a change in the OpenCPN plugin API. For the changes in OpenCPN I will submit a PR.
To communicate to the Raymarine Raytalk ng (= NMEA2000 with proprietary PGNs) a Actisense NGT-1 is required.

Douwe Fokkema

[partyvi]

Can somebody tell me how to ‘pair’ autopilot route pi with pypilot?
Pypilot works wall itself. Autopilot route works itself and shows right overlay but do not direct pypilot in any gps or compass mode.
Looks like I have to find how to direct nmea output from autopilot to signalk.

Could somebody share how you use autopilot with pypilot or pypilot motor controller.

[seandepagnier]

Originally Posted by partyvi

Can somebody tell me how to ‘pair’ autopilot route pi with pypilot?
Pypilot works well itself. Autopilot route works itself and shows right overlay but do not direct pypilot in any gps or compass mode.
Looks like I have to find how to direct nmea output from autopilot to signalk.

Could somebody share how you use autopilot with pypilot or pypilot motor controller.

The autopilot route plugin is not specific to pypilot or any other autopilot. It replaces the nmea APB output built into opencpn with many more options for computing how this sentence is generated. It does output a continuously changing bearing in some modes which is no problem for pypilot, but some autopilots might not allow it.

To make pypilot follow routes from either opencpn without the autopilot route plugin or with it is the same. You need to make tcp connection from opencpn to pypilot port 20220 with "output on this port" checked.

The latest version of the pypilot plugin can automatically send nmea sentences to port 20220 on whatever host it connects to pypilot for signalk which makes it easier to change the autopilot host (if you have more than one for example) without needing to add or change the connection in opencpn.

[rgleason]

I've updated a tried to clarify this page, based on recent comments, but it needs some screenshots and more useful description regarding operation.

Autopilot Control Discussion
http://www.cruisersforum.com/forums/...ol-196347.html

Also some other useful threads and developments with the more expensive Raymarine solution:
Fast heading output from RM Ev1 Douwe writes "So now I made a little interface program that reads the PGN 127250 from the EV1 on the NMEA2000 network through an Actisense NGT-1 and outputs a HDM NMEA183 message on another port. Works fine. That is actually what an Actisense NGW-1 should do, but doesn't."
AutoTrackRaymarine_pi Recent PR on OpenCPN by Douwe
Increase precision on XTE. Add ZeroXTE() to pluginmanager #1470
I was not able to find Douwe's thread on CF



Perhaps this effort will enable an improvement to Autopilot and PyPilot too?

[partyvi]

Originally Posted by boat_alexandra

The autopilot route plugin is not specific to pypilot or any other autopilot. It replaces the nmea APB output built into opencpn with many more options for computing how this sentence is generated. It does output a continuously changing bearing in some modes which is no problem for pypilot, but some autopilots might not allow it.

To make pypilot follow routes from either opencpn without the autopilot route plugin or with it is the same. You need to make tcp connection from opencpn to pypilot port 20220 with "output on this port" checked.

The latest version of the pypilot plugin can automatically send nmea sentences to port 20220 on whatever host it connects to pypilot for signalk which makes it easier to change the autopilot host (if you have more than one for example) without needing to add or change the connection in opencpn.

Thanks, works for me. I forgot to put “output on this port”

[rgleason]

It does “output on this port” a continuously changing bearing in some modes which is no problem for pypilot, but some autopilots might not allow it.

Is this what you meant?


Douwe's post is below in Post #34

[seandepagnier]

I didn't try other autopilots with the route plugin, but apparently some autopilots require confirmation on bearing change. So the apb must give the same bearing always and only alter xte.

The autopilot route plugin outputs a continuously changing bearing in some modes always with xte of 0 (because xte is not used) so the autopilot must accept this and not require confirmation for it to work.

[rgleason]

Originally Posted by partyvi

Can somebody tell me how to ‘pair’ autopilot route pi with pypilot?
Pypilot works well itself. Autopilot route works itself and shows right overlay but does not direct pypilot in any gps or compass mode.
Looks like I have to find how to direct nmea "output on this port" from autopilot to signalk.

Could somebody share how you use autopilot with pypilot or pypilot motor controller.

Is this what you intend "output on this port"? It still is not a clear description.

[rgleason]

Wiki revised a little
https://opencpn.org/wiki/dokuwiki/do...i_with_pypilot

[Douwe Fokkema]

Originally Posted by rgleason

I've updated a tried to clarify this page, based on recent comments, but it needs some screenshots and more useful description regarding operation.

Autopilot Control Discussion
http://www.cruisersforum.com/forums/...ol-196347.html

Also some other useful threads and developments with the more expensive Raymarine solution:
Fast heading output from RM Ev1 Douwe writes "So now I made a little interface program that reads the PGN 127250 from the EV1 on the NMEA2000 network through an Actisense NGT-1 and outputs a HDM NMEA183 message on another port. Works fine. That is actually what an Actisense NGW-1 should do, but doesn't."
AutoTrackRaymarine_pi Recent PR on OpenCPN by Douwe
Increase precision on XTE. Add ZeroXTE() to pluginmanager #1470
I was not able to find Douwe's thread on CF

Perhaps this effort will enable an improvement to Autopilot and PyPilot too?

My new thread on AutoTrackRaymarine is on:AutoTrackRaymarine_pi: Remote control and route following for Evolution Pilots - Cruisers & Sailing Forums

[rgleason]

I've updated Sean's Autopilot_Route_pi for the new Plugin Manager and has all environments build except Xenial. The tarball files are available on Cloudsmith in the Opencpn organization plugin repository under autopilot-route-alpha for those who would like to test this plugin.

Keep in mind that this plugin does not have a normal toolbox icon or menu, it's settings are controlled from the preferences menu.

From the manual
"The autopilot route plugin is a completely separate plugin to be used with any autopilot. It allows you to use different algorithms and closely tune them for route following logic. It has nothing specific to pypilot. This plugin allows the user to select different algorithms and closely tune them for route following logic. The autopilot route plugin replaces the nmea APB output built into opencpn with many more options for computing how this sentence is generated. It does output a continuously changing bearing in some modes which is no problem for pypilot, but some autopilots might not allow it."

Modes:
1. Route Position Bearing
2. Standard Cross Track XTE -XTE multiplier
3. Waypoint Bearing -by Distance or by Time

Options
1. Confirm Bearing change on Waypoint Arrival.
2 .Intercept Route on Current Course
and several other features.

TESTING

I am trying to figure out how to test this plugin from the desktop and have not figured out a good strategy. If anyone has some good suggestions I would appreciate hearing them. I started testing by selecting "Navigat to here" but there is no motive action to the boat and using a vdr file with a route would be conflicting.

Alternatively, perhaps there are users who can test this plugin in real life with Opencpn guiding their autopilot! Please make your reports here, we want to know if the plugin works properly and can be added to the master catalog.