Open Source Autopilot

[seandepagnier]

I have been developing an open source autopilot program to use gps, magnetometer and wind sensors to steer the boat. Currently initial tests in wellington harbor using tilt-compensated magnetometer prove successful but extensive testing is needed. Total hardware cost is ~$120 but it has about 2000lbs of pull which I'm afraid might break something (I overengineered it for my boat) It uses 1 amp without load and up to 10 when it is pulling hard (and can move hard over in just over 1 second) This is a brushless rc outrunner ($25 on hobbyking.com) driving a 120:1 planetary gearbox from a truck winch. The motor that came with the winch had too high of a starting current. Hopefully I can reduce cost and make weaker ones.. I have two more weaker drive units I have not yet tested.. one can drive a wind vane, the other is based on a broken autopilot.

Currently it uses a simple filter which provides critical dampening.. so it counter steers before it hits the right course so it wont oversteer, and it automatically adjusts the two gains over time to achieve this.

I am pondering the viability of commercializing this. What do people want? What will they pay for? Actually any comments related to this type of work are appreciated.

My initial plan was to offer a waterproof box which has 2 wires going in and 2 wires going out and can handle up to around 30 amps. If you need more power you could just replace the motor controller as they are made in all sizes and extremely cheap. It may be required to have a separate box for the magnetic sensors as they are negatively affected by the high current wiring (this is the current setup) An internal hall sensor can detect motor stall (in less than 1/20th of a second) and an internal buzzer can provide some feedback on various error conditions.

Then _you_ have to take either a broken autopilot, linear actuator, electric winch, hydralic pump etc.. and connect it to the wires, and rig the mechanical parts to make it move your rudder. Brushless motors would be supported as this is what my prototype uses and it is no extra cost or difficulty. Also optionally usb can be connected to control with opencpn to give gps steering and waypoints etc, otherwise just compass.

Then for those who can solder, a much cheaper kit version (thru hole soldering only) could be assembled in a few hours for a much reduced cost, however the user would have to perform a rather arcane calibration procedure themselves to make the magnetic sensors correct. They can calibrate automatically over time.. but I currently know of no way to get better than 2 degrees accuracy in all orientations automatically on a moving boat without prior careful calibration on land.. this trick only works to keep them in tune. It may be possible to get around this with redundant sensors but this is complicated (have not worked it out) and would require 3 $25 sensor boards instead of just 1.

An optional wind vane sensor could be connected.. I built one already using an rc servo.

To take a look at the code and/or algorithms
https://github.com/seandepagnier/cruisingplot

Since all the design would be open, you would be free to modify and mix and match hardware as required for whatever application, eventually support gyros for racing applications, or for people who need to surf the swell with their spinnaker at 12 knots. In fact I have a lot of future ideas not listed here because it would get too far off track.

I will post pics tomorrow.

[noelex 77]

Welcome to the forum.

Sounds like a great idea. These electronics should be much cheaper than the marine companies are selling them for.

An electronic cheap windvane should be very useful. Even people with a commercial unit would appreciate a back up.

[Snowpetrel]

I really like the idea, I love the windvane or trim tab drive idea. A cheap reliable backup would be popular I am sure.

You are in a great area to test it... If it can survive the Karori rip it should be fine anywhere in the world!

[Snowpetrel]

Could the basic software be adapted to use a cheap smartphone as the guts, combining computer, gps and fluxgate compass. complete with screen and touchpad controls. maybe even with bluetooth or wireless conectivity to another cheap phone for remote use, and as mob function... Now that would be cool!

[seandepagnier]

Originally Posted by Snowpetrel

Could the basic software be adapted to use a cheap smartphone as the guts, combining computer, gps and fluxgate compass. complete with screen and touchpad controls. maybe even with bluetooth or wireless conectivity to another cheap phone for remote use, and as mob function... Now that would be cool!

Absolutely. This is a great idea as it would save cost on the magnetic sensors as well as increase functionality (touchpad controls etc) I probably still need a microprocessor to control the motors, and get stall feedback. This type of option would probably be developed later on.. but of course if someone sends me a smartphone to use it would speed this along considerably.

Not too sure I'm going to be doing a lot of testing the kairori rip, especially not during storms as the tidal currents could put me on the rocks!

My current setup is rather simple, just counter rotating pulley on the same drum so the motor winch could be anywhere (in a nice dry place) with just the lines going out connected to the rudder to move it. Of course various configurations are possible and entirely up to the end user. So yes for people who can salvage components and build things, sure they could make an autopilot for under $100 as I mean only to provide free design instructions and software, but for others to avoid soldering and experimentation it would be a little more.. but the great thing is it is modular so if it breaks it isnt likely to cost a lot to fix.

[cas206]

Came across this because I have a tiller drive version of the Raymarine ST4000 gear with a nonfunctional control head/course computer unit. I'm currently trying to determine what is needed for a laptop to read the mag sensor (5 wires + shield) to turn into digital information and what is needed to to control the tiller drive (+/- 12 volt signal). I need to accomplish that before messing about with any software code.

[motu42]

Hi boat_alexandra

I have been developing the other end of such a system. It provides the instruments and chartplotter, plus autopilot control. The front end is anything with a web browser, eg PC, tablet, smartphone.

Pls have a look at Freeboard [MeshCMS] and give me your thoughts.

Freeboard will be open source, on commodity hardware, but many potential users will just want a complete finished box of gear delivered. I'm still thinking about how to deal with that.

Rob

[Jimbo485]

Where are you these days, boat_alexandra?

How did you go with the development of your autopilot?

[seandepagnier]

I wanted to follow up for people interested in free autopilot systems. I have completely rewritten the software, and extensively tested it over the last year and half sailing about 1000 miles.

The computer used could be almost anything, but so far I am testing with raspberry pi and other cheap chinese pi clones that cost $10 or less. They use little power, but provide a small display and keypad, tv remote control (working on wireless), opencpn plugin with integration and web interface for control from devices.

I have two motor controller options. Either an RC drone controller ESC which are available in different sizes I tested 20, 30 and 70 amp. These are reprogrammed for brushed motors, or I have developed a more efficient large controller from scratch intended for hydraulic and large linear drives capable of 60 amps and idle current only 16 mA. There is optional rudder feedback, and all sorts of protection. I haven't damaged any controllers with regular use.

For the actual drive unit, I have used 6 different tiller pilots, wired directly as well as a windshield wiper motor rigged to the tiller using a winch drive. I am working on using battery operated drill motors with ball screw to make a linear ram as well though it won't be as efficient. The windshield wiper motor cost $6 and is more efficient (uses 50-70% power for same rudder speed) than simrad tp22, raymarine 1000, autohelm 2000 and others.

For wheel driven boats I have used windshield wiper motor connected to a toothed belt to turn the wheel on 30 ton steel boat. I will be testing a hydraulic pump drawing 10-15 amps in a few days on another boat.

I have tested on 8 different boats.

For more information: pypilot - open source marine autopilot

[gmakhs]

Interesting