DIY Autopilot

[Tscott8201]

This was on my Facebook feed this morning. Thought you all might enjoy it.

(FYI can't watch video embedded you must follow the link to youtube)

https://www.youtube.com/watch?v=-nA6...layer_embedded

Tom

[DeepFrz]

Now that is cool.

[goboatingnow]

The video of course does not demonstrate the real design issues of autopilot control and that is dynamic response. whats shown drop box description is a very simple PID, with the differential gain taken from the horizontal rate element of the compass chip.

see https://www.dropbox.com/sh/l1l1432c9vfs4g1/7PJUMgsj0E for those of you interested in the code

The issue of course is not so much XTE tracking ( almost anything will do that ), but instaneous response to significant rate of heading change

dave

[s/v Beth]

Its going to be fun to see what autopilots we have in a few years time. As I understand Dave, the problem is to get a rapid correction on a boat that is going many different vectors at once- like a multihull skimming over the water sideways.

[DeepFrz]

I watched a video at the same link (displayed after the link video) of a new autopilot on a quad copter that seemed to "learn" about the forces on it and start improving its accuracy. Its the ArduPilot Mega 2.5. Might be a step up from the one shown and software is all open source.

[goboatingnow]

Originally Posted by s/v Beth

Its going to be fun to see what autopilots we have in a few years time. As I understand Dave, the problem is to get a rapid correction on a boat that is going many different vectors at once- like a multihull skimming over the water sideways.

I think since rate gyros have been added, future improvements will be limited to small advances. The issue t now is clearly the physical dynamics of moving the rudder quick enough to respond to big perturbations in vessel position. Thats not easily solved.


dave

[colemj]

Autopilot control of multiple thrusters and stabilizers?

Mark

[tropicalescape]

I once made an in flight adjustable prop that used an electric screw driver motor to change the pitch of the prop, I dont remember where I saw it but a fella made a self steering autopilot for his boat from one also..Electric screwdrivers can be used for lots of different applications when a little bit of "throw" is needed, I would feel more comfortable with the hydrolic setup as it tends to be more robust...

[goboatingnow]

Originally Posted by colemj

Autopilot control of multiple thrusters and stabilizers?

Mark

Already built in to upmarket commercial autopilots.
Dave

[familycruisers]

Rc heli gyros are about up to the task, they have rate and heading hold as well as integration into GPS autonomous applications. Once again the big boys are behind the tech.

[DeepFrz]

The Pixhawk autopilot is just amazing.

[Lake-Effect]

Originally Posted by goboatingnow

The issue of course is not so much XTE tracking ( almost anything will do that ), but instaneous response to significant rate of heading change

Is instantaneous heading change the grail, though? From thinking about my time hanging onto a tiller, it seems the best control comes from observing how the boat heading is oscillating around the desired direction, and how the heading relates to travelling the desired line, and the feedback was a combination of observation and pressure from the tiller, and the most efficient control was a series of well-timed nudges. What we learn to do almost instinctively. Possibly too complex to push into an Arduino.

(I did take control systems 35+ years ago but it wasn't my favourite subject)

[wellin]

I agree with the previous post. I think a fast response time to keep an exact heading, would just drain the batts. Would be nice on a calm day motoring but in waves and wind, not good. Motor would run back and forth and never stop. So i think an averaging over time and then make a small correction and "wait" to see if the correction worked would me my method. You constantly get knocked back and forth heading wise so no point in correcting when the next bounce might straighten your heading out. Maybe record headings every second, then average them over 10 seconds. Ignore any angle change under 5 degrees and only move the rudder if average heading changed more than 5 degrees. Then do a "preset" rudder deflection, then return to center. The amount of deflecton tuned as needed to not spill drinks. Then wait till the next 10 second cycle and compare the new average heading. Also record the direction of heading corrections so that if the autopilot corrects the heeading 2 or more times in the same direction then make a small shift in the neutral rudder angle so next correction is smaller than the last one so any weather helm would be tuned out in a few cycles. Relieving the autopilot from constantly correcting for the same thing.
All of the perameters for "turn rate" "delay time" and so forth could be adjusted "on the fly" with pots making setup real easy.
I love working out logic problems and programming stuff like this.

Andy

Sent from my SCH-I415 using Cruisers Sailing Forum mobile app

[colemj]

This is how the good ones work now. They don't seesaw back and forth reacting to every heading change - they learn how the boat responds and the acceleration rates and anticipate or hold off steering as needed. I am always amazed watching our AP steer in largish seas just like (or better) than I would.

Mark

[GILow]

I built an autopilot a few years ago, tiller pilot style. I found the two hardest parts were finding a linear actuator that was up to the job and expressing the "learning" aspect of the code in the very simple language available to me using the chip I had chosen (picaxe).

In the end I got the the learning aspect right using a variable resistor to control the tolerance to course deviation and another to set the response rate. Playing with these two got some very good results.

In the end it was the lack of a good (affordable)durable linear actuator that scuttled the project and I went out and bought a tiller pilot instead. But I still have the bits, particularly the ic2 compatible compass chip (CMS300 or something like it from memory?) waiting for a chance to play again. I am thinking I will use the bits in conjunction with my windvane steering when I have built it as a supplement for light air sailing. Getting around the problem of a decent linear actuator by controlling a trim tab instead of a full rudder.

Fingers crossed, it should be fun.

And in the interim, the options available to replace the picaxe component are just terrific, though I still love the simplicity of the picaxe itself.

Matt