homemade autopilot

[gonzo]

I used Sperry gyrocompasses for a long time working as a navigator. They were the size of a refrigerator and very accurate.

[Frosty]

Im all for making stuff, be it fun or necessity.

But making an auto pilot is a difficult job, its not like they are expensive.

If you got the control box off something like the Navico wheel pilot you could make the rest. The control box houses everything.

A rudder feed back is not necessary.

[gonzo]

Some boats will steer wildly without a rudder feedback sensor.

[Frosty]

Granted but the Navico wheel pilot does'nt use one.

Its the most simple auto pilot which is why I suggest getting an old control box off one.

I have heard of people using this control box to track a satellite dish on boats at sea. !!!!!! might work,--I dont know if it does or not.

[CDK]

Originally Posted by gonzo

I used Sperry gyrocompasses for a long time working as a navigator. They were the size of a refrigerator and very accurate.

Those were the days....
The Honeywell fluxgate sensor Fanie pointed to measures 4x4 millimeter and has 1-2 degrees accuracy.

With a microprocessor doing the math it should not be necessary to use rudder sensors if the program is able to learn from previous errors and stores the results. Something that was impossible with analog technology.
Long ago I designed a central heating clock/thermostat for a client. It kept track of the rate at which the temperature rose and fell; approx. a week after installation the thermostat was able to provide the set temperature within 0.2 degrees, even with very slowly responding floor heating systems. Only a sudden change in outside temperature could cause a somewhat larger error on the next set point.

A similar system for an autopilot would over-correct during the first few minutes only and perhaps again briefly when reaching the lee of an island.

[gonzo]

I've been toying with the idea of software that will steer correctly in following seas. A good helmsman will start correcting before the boat changes course. The correction starts when the boats gets the bow down. I wonder if a gyro could be used for that.

[catsketcher]

Frosty -

I guess I am looking for top of the line performance with gyros and a bigger motor without the expense of a below decks autopilot. I know that you can get an okay wheel autopilot for about $1800 Aus but if you want accelerometers and bigger motors mounted under deck you can pay double really quickly. My friend with a similar sized multi to mine can't use his $1800 autopilot in nice sailing weather downwind - it is too slow for a multi.

I don't care about current draw. I would much rather have a nice chunky motor pulling amps and have my tow generator off the back feeding them back in whilst I watch the boat get steered. I have spent many hours steering my boat downhill because the pilot can't handle the conditions quickly enough. I know to trim the boat well and I do but response time and grunt are needed in spades. If I buy a system suited to my cat I can quickly spend $4000.

So I would love a fast motor and a trainable software that I can set to my boat. I will have a laptop anyway so would like to program a PIC or some other type of chip like they have in radio controlled models.

That's my wish list anyway

Phil

[gonzo]

With a powerboat, a standard automotive type hydraulic powersteering is a good choice. You can drive the system with a small fast electric motor

[CDK]

Originally Posted by gonzo

I've been toying with the idea of software that will steer correctly in following seas. A good helmsman will start correcting before the boat changes course. The correction starts when the boats gets the bow down. I wonder if a gyro could be used for that.

An inclinometer, being a drop of mercury in a slightly bent glass tube could detect changes in the longitudinal axis. But as long as the direction doesn't change the autopilot cannot take action because it doesn't know in which direction it must steer. With the following seas coming from port or stbd it could anticipate, but from almost dead astern it would make mistakes.

Your 'good helmsman' also has visual information and sees the big one approaching. When blindfolded he will perform less good. I once entered a harbor with lots of obstacles ahead and a friend of mine threw an air mattress over the windscreen, saying "you have radar, you don't need this". I managed, very very slowly, but it made a hell of a difference.

[gonzo]

I think it could make an improvement. Also, it could use winspeed to anticipate the boat heeling over and rounding up when going upwind.

[yipster]

good to stand corrected and around '92 a fluxgate it was..

[Tim B]

I think the tricky part of this question is actually the hardware. There are several Autopilot codes available on the web to choose from, and in it's simplest form a steer 10 deg port/starboard pilot is not hard to do (not efficient, but not hard).

There is a project called YBW OpenPilot (which at present has nothing to do with OpenPilot) which is looking to produce a software autopilot. I have invited them to develop it as part of the OpenPilot system, but as yet have had no response.

I would suggest that you look at using OpenPilot as a base, as the most of the code is pretty portable, and it could be made to run on any embedded platform that supports QT4. Furthermore, a lot of the data parsing has already been done and tested.

Gonzo, how long are you in London? (PM me)

Hope this helps,

Tim B.

[Fanie]

I think the tricky part of this question is actually the hardware.

This is a steering mechanism I made and used. It simply worked with a joystick and relays switching the motor on and off and changed direction.

That is one motherly big wiper motor, a total overkill. It's a 24V motor I ran off 12V, and it will crush you if any body part gets in it's way.

The construction is an alu plate that bolts to the motor, on the alu plate is bolted a U shape milled engineering plastic the rack fits through. The sprocket turns and the rack moves. You can still see the ball fitting on the left of the pinnion.



Sorry CDK - I think I misunderstood what you were talking about on the fluxgate. I'm sometimes like the English say here but not all there

An inclinometer can be a simple potentiometer with a weight on it

[Guest625101138]

Originally Posted by farjoe

Which characteristics of a girocompass or a fluxgate make either essential for an autopilot?

Why isn't the track data coming out of a modern GPS sufficient? It may not be correct to compare older GPS units with modern ones. The one i like most is half the size of a computer mouse, totally water proof, very low current comsumption and can almost detect satellites from within the house.

It seems to me that with a simple microcontroller the variation due to waves can be averaged out. For me the challenge seems to be more in determining the actual angle of the rudder than any other parameter since many units I have seen do not have direct feedback of rudder angle.

of course this is all theory and practice may teach me otherwise.

The ability of the control system to cope with external disturbances such as wind and waves depends on the speed at which it gets course data. If the course information is slow then the system response has to be slow as well otherwise it will become unstable.

A GPS typically has a slower update than a gyro or Hall device. The GPS requires at least two updates to know the course as it is inferring direction from the track between points. The fastest update is usually within 1 second and a slow moving boat that is bouncing around will need some smoothing of the GPS data. So there is delay in the feedback signal that limits the overall response time. It will be OK for a boat that is not being thrown off course by fast changing external inputs.

Any system really needs linear control action rather than on-off. Linear proportional control makes a small correction for a small change in direction and larger change for larger variation. It will also improve with differential and integral action - so called PID loop control or its digital equal works well.

You can get adaptive control action as well but that is getting closer to what humans can achieve. The advantage of a human is that there is much more sensory data usually available. A human can actually anticipate how the boat will move as a wave approaches and make a course correction before the wave arrives. The human can feel the motion of the boat as well.

The drive mechanism needs to be robust as they tend to do a lot of work.

I have made a two autopilots and my view is that commercially available autopilots represent good value. If you have skill with electronics and ability to machine and fabricate the necessary drive parts then you might do something cost effective with carefully selected parts. Downfall with the first one was use of on-off control while both underestimated the amount of work they do over a period of time and wear and tear on the mechanical bits (with tiller mounted units). I really did not get beyond the prototype stage before realising it was cheaper to buy a commercial unit.

If you can find a proven design then it might be cheaper to build as you avoid wastage in the development phase. Make sure the application is similar to yours.

Rick W

[CDK]

When I made the autopilot, the boat had power steering. All I had to do to put the stern drives in the desired position was push gently against the control valve. I used two small solenoids on a sub-frame that moved with the steering arm(s) and the Bourns wire wound pot that was already there for the rudder indicator. The stroke of the solenoids was only 4 mm so they required only 0.5 Amp.

That part worked perfectly. I also installed a joystick to steer the boat, using the same solenoids, but after the fun of the accomplishment I found myself using the steering wheel again. Because it felt more natural.