12vdc Marine Computers

[rgleason]

Matt thanks for yoyr thorough blog on installing raspberry pi and cubietruck opencpn. Also your arduino gps autopilot ais and sensors project looks interesting. The painting countertops project was a surprise. How is it holding up?

Sent from my SCH-I545 using Cruisers Sailing Forum mobile app

[tonyappuk]

Mattkab
I have been playing with the idea of using an Arduino to control an autopilot from Open CPN. Rather than use a compass to provide the heading info I thought to use trackangle from RMC which can be made available from Open CPN as well as RMB. You say there is insufficient resolution from GPS which may be true but could you say why, please?
Tony

[seandepagnier]

Originally Posted by tonyappuk

Mattkab
I have been playing with the idea of using an Arduino to control an autopilot from Open CPN. Rather than use a compass to provide the heading info I thought to use trackangle from RMC which can be made available from Open CPN as well as RMB. You say there is insufficient resolution from GPS which may be true but could you say why, please?
Tony

I build a working autopilot using a raspberry pi 2 years ago. I used magnetic sensors and gyroscopes.


You might find the gps can work if the sea is flat, but with waves, the boat can get knocked around. This doesn't register as gps heading until 3-4 seconds after the sails are luffing. The gps just can't even come close to handling this.

[mattkab]

Originally Posted by rgleason

Matt thanks for yoyr thorough blog on installing raspberry pi and cubietruck opencpn. Also your arduino gps autopilot ais and sensors project looks interesting.

Just giving back to the community -- these projects are just a matter of putting the pieces together; pieces that other people have done a fantastic job building before me.

Originally Posted by rgleason

The painting countertops project was a surprise. How is it holding up?

The countertops are fantastic - look the same today as they did the day they dried. I actually want to do it in my house as well. It's an epoxy coating from Rustoleum for those that didn't read the post.

[mattkab]

Originally Posted by tonyappuk

Mattkab
I have been playing with the idea of using an Arduino to control an autopilot from Open CPN. Rather than use a compass to provide the heading info I thought to use trackangle from RMC which can be made available from Open CPN as well as RMB. You say there is insufficient resolution from GPS which may be true but could you say why, please?
Tony

The why is mostly from reading and talking to the quad-copter guys that spend a lot of time and effort getting autopilot technology on these platforms stable. The RMB messages are great for setting the destination/waypoint -- but for better tracking you need something more fine-grained, thus the compass.

Now... this is all theoretical, as I have yet to attempt to actually build the autopilot portion. And it's not a horrible idea at all to start with just the GPS messages (maybe run through a smoothing algorithm to account for some of the issues) and see what happens. At trawler and sailboat speeds (and sizes) it may be sufficient.

Originally Posted by boat_alexandra

I build a working autopilot using a raspberry pi 2 years ago. I used magnetic sensors and gyroscopes.

You might find the gps can work if the sea is flat, but with waves, the boat can get knocked around. This doesn't register as gps heading until 3-4 seconds after the sails are luffing. The gps just can't even come close to handling this.

Brilliant. Do you have a writeup of what you did? I didn't even think about the effect of swells.

I also note that just sitting tied to the dock, the GPS thinks the boat changing heading, and is even moving at up to .3 kts... now, there may once again be a way to filter the information to get higher quality signals. Probably worth looking into in fact.

[seandepagnier]

Originally Posted by mattkab

I also note that just sitting tied to the dock, the GPS thinks the boat changing heading, and is even moving at up to .3 kts... now, there may once again be a way to filter the information to get higher quality signals. Probably worth looking into in fact.

When your gps speed is low, the heading is not reliable. Further, gps heading is the direction you are moving, not the direction the boat is facing, and also knows nothing about how the apparent wind changes direction as you reach the crest of a wave.

Instead consider combining gps (for conditions with little wind and waves) with a wind direction sensor. At least it can react fast enough, and also deal with subtle wind shifts, but there are limitations (besides not working when there are swells with little wind) For example:

Further, depending on where the wind direction sensor is mounted, and how it is constructed, the boat motion itself may cause the paddle to shake and it can also measure an apparent wind that is not what affects the sails appropriately. Consider the wind sensor at the back of the boat when yawing, or top of the mast when pitching/rolling.

Ideally you would have wind direction sensors in several locations on the boat, and also sensors which can measure the vertical component of wind, but this is probably just not very practical.

You can use very clever filtering to probably mostly sort this type of problem out to reduce oscillations (most people who are familiar with mechanical wind vanes know about his) but still doesn't completely solve the problem.

"Bearing off at the crest prevents the stern from falling into the trough." It is difficult to determine where the crest of the wave is with only a wind direction sensor and gps, but not very difficult with accelerometers and gyroscopes. This adds considerable safety factor, and also allows fine-tuning between a comfort/speed trade-off.


Inertial sensors are cheaper than gps and can sample at a few hundred hz easily. What is the problem with using them?

[mattkab]

Originally Posted by boat_alexandra

When your gps speed is low, the heading is not reliable. Further, gps heading is the direction you are moving, not the direction the boat is facing, and also knows nothing about how the apparent wind changes direction as you reach the crest of a wave.

That's a good point about the movement vs direction... I need more coffee.

Originally Posted by boat_alexandra

Inertial sensors are cheaper than gps and can sample at a few hundred hz easily. What is the problem with using them?

Not a problem. I was just responding to a question regarding why I didn't think GPS alone was good enough to run an autopilot. As I was responding I was thinking that in some cases a filtered GPS signal may suffice -- enough that it's probably worth a test to see. This is for fun and learning as much as it is anything else.

In the end, in my autopilot project for my trawler, I plan to use GPS with a tilt-compensated magnetic compass, and have all the hardware to do so.

[tonyappuk]

I seem to have provoked lots of interesting comments which I did not really anticipate! Thanks to everyone. In my dotage I confine my sailing mostly to the Solent and a lot of it is motorsailing. The development of an autopilot accessory is a hobby plaything - you hardly need a sophisticated autopilot in the Solent but it would be useful for my sort of single handed sailing. In the interest of cheapness I shall use the windvane input on my antique Autohelm which is another reason to use an Arduino.
Tony

[seandepagnier]

Originally Posted by tonyappuk

I seem to have provoked lots of interesting comments which I did not really anticipate! Thanks to everyone. In my dotage I confine my sailing mostly to the Solent and a lot of it is motorsailing. The development of an autopilot accessory is a hobby plaything - you hardly need a sophisticated autopilot in the Solent but it would be useful for my sort of single handed sailing. In the interest of cheapness I shall use the windvane input on my antique Autohelm which is another reason to use an Arduino.
Tony

I believe gps will work fine provided:

1) minimal wave motion
2) you are not relying on erratic wind for propulsion (always enough wind in the right direction, or using a motor of some kind)

You are using the stm32 based arduino?

[jonasaberg]

Originally Posted by boat_alexandra

I build a working autopilot using a raspberry pi 2 years ago. I used magnetic sensors and gyroscopes.

Sean,
Have you posted details about this project somewhere?

/Jonas

[seandepagnier]

It is still incomplete and I got sidetracked (working on opencpn) there is some source code here:

https://github.com/seandepagnier/rpi_autopilot

[rgleason]

Jonas look on Sean's github very interesting projects.

Sent from my SCH-I545 using Cruisers Sailing Forum mobile app

[tonyappuk]

BoatAlexandra
It is a Duemilanova with my own relay card mounted above. I really must do a bit more coding but really I have done very little with it. Nice to see all the interest though.
Tony

[seandepagnier]

Originally Posted by tonyappuk

BoatAlexandra
It is a Duemilanova with my own relay card mounted above. I really must do a bit more coding but really I have done very little with it. Nice to see all the interest though.
Tony

Oh... isn't the duemilanova obsolete?

I have used relays also, but I was concerned that they wear out after 100,000 cycles...

Another problem with relays is they cannot do a variable speed, so therefore they just aren't as efficient for power consumption. Also, obviously cannot drive brushless motors (which is what I was using) which are also even more efficient and last a long time.

[tonyappuk]

It probaly is obsolete but does everything I need. It can drive the reed relays directly with suitable overswing diodes and the input is to the old Autohelm windvane input. The Arduino parses the two NMEA strings RMB and RMC and determines the error in course and the probable amount of tiller movement needed. This way it uses the Autohelm drive to the tiller which saves building anything extra. At least that's the theory!
Tony