Hi Peter and David, This is very interesting for me. Ultimately, my project with these things is for building a model sailboat that is semi-autonmous. Give it a list of waypoints, and it will navigate to them. This is already implemented commercially for model planes, and there are various attempts at trans-atlantic crossings with model sailboats (so far unsuccessful). To begin the learning process, I am using arduino and a wind direction instrument to automatically adjust the sheet for the apparent wind. And to get more realtime practice with the arduino, I am adding to the NEMA-based system on Cats Paw to derive tilt-compensated true heading info. A tilt-compensated Hall effect compass module from Sparkfun generates the NEMA code for magnetic heading, and the RMC sentence from a little GPS unit provides the local magnetic variation. So that allows construction of HDG (magnetic heading + deviation and variation), which standard nav programs like MacENC or iNavX can use to get true headings. I have not even started putting together an autopilot for the model (on Cats Paw I use a Raymarine wheel pilot), but would love to learn more about how you implement the "learning" steps in having the autopilot make the appropriate rudder corrections for a certain course error. My guess is these projects will not be of interest to lots of the Stus-list, so perhaps we should start communicating offline. Interested?
Eric Frank Cat's Paw C&C 35 Mk II Mattapoisett, MA > From: David Paine <paineda...@gmail.com> > To: cnc-list@cnc-list.com > Subject: Re: Stus-List Homebrew autopilot > Message-ID: > <cantme359m12gxtortu3jrnd7qgbjh+5g03wyp-amz1txfov...@mail.gmail.com> > Content-Type: text/plain; charset="windows-1252" > > Hi Peter, > > I built a homebrew below deck autopilot for my C&C33 a couple of years ago > using a linear actuator, a mechanical clutch, a H-bridge-like serial power > controller from Pololu, an arduino, a transreflective serial LCD display, > four buttons, and most important of all, a digital compass with serial > communication (NMEA 0183) operating at 10Hz which, along with a sparkfun > gyro (15 degrees/sec) provides heading information. I bout everything from > ebay and probably have $250 and untold hours invested) The programming is > fairly easy with arduino since there is a PID controller library available > (use two, one for the positioning of the actuator, the other for correcting > for the course error (the difference between the intended course and the > present heading. > > I've used my autopilot on many long solo trips including one 80 m mile > offshore run with 20 kn wind directly astern and a large swell off the > quarter (during which I was able to sleep for four hours as the system was > so stable. > > Hints: Forget using GPS (or wind) alone to set the course you need much > faster response time and that is why you need a 10 Hz heading signal. I > thought about using a Kalman filter but soon learned that Kalman filter > math is serious juju and was very glad to have stuck with PID control -- we > only have two degrees of freedom its not like it's a quadracopter! The > worst part of my system as it stands now is the mechanical clutch to engage > and disengage the actuator -- one day I will find a hydraulic actuator > which will allow a solenoid to disengage the actuator. Raymarine actuators > have an electromechanical clutch which is also nice but consumes more > current than I like. Also, try to use components with serial > communications protocols -- these will have on board filtering and make the > whole exercise much easier. > > > Best, > > David > > > > > On Tue, Jan 21, 2014 at 2:24 AM, Colomba <colo...@peter-gottwald.de> wrote: > >> Good day >> >> I`m wondering if there is anybody who want to build a DIY autopilot? If >> so, it would be interesting to discuss/combine ideas. >> >> In the moment my idea ... and what I`ve done so far is following: >> - Controll unit with a raspberry pi >> - language Python (because I really do not know C to handle this properly) >> >> The idea in the moment is to use this year the old wheel pilot (Navico) as >> steering unit. If the controll unit will work fine, I will change the >> steering unit next winter and will mount a linear drive under deck. >> >> What I`ve done up to now ... >> I took the code of FishPi autopilot as basic to change to the things which >> I think are usefull. >> For time being I integrated a compass/magnetrometer/accelerometer module, >> called cmps10. It`s cheap and doing it`s job. >> For GPS signal I`m using a simple USB GPS mouse to figure out >> Lon/Lat/speed/course >> ... not done yet ... integrate a reader for wind direction. On board I >> have the NASA clipper instruments and will read this datas via Atmel >> microchip, connected to the controll system via I2Cbus. Atmel Afinity 1327 >> Chip is programmed and working, but up to now no test if it`s working with >> the NASA clipper instrument, due to the fact, mast is down for winter... so >> I have to play arround first to test it on board, but ... it`s cold >> outside, so I`m still waiting. >> >> Rudderangle, just working, with a potentiometer and signals collected by >> GPIO pins of raspberry pi. >> >> Route planing .... FishPi provide the idea of integrating routes in gpx >> format. I think that idea is ok for an autonomus vehicle, but I do not like >> that idea. I`ve tested to run opencpn and FishPi parallel on the raspberry >> .... and it works, so I take the signal from opencpn for heading to next >> waypoint via UDP and can use opencpn. >> >> For output, Ive integrated a motor controll via SDA network, which should >> work with a PA with the wheel unit. Not tested yet. >> >> There are a lot of things to do to get everything working nicely. >> - define sailing situation with no go areas (i.e. downwind or app +-38? >> head wind) >> - define a "learning tool" to define rudder gain vs heeling and >> acceleration, my idea is to create a simple database and to store real life >> datas. >> - implement a kalman filter or a complementary filter (which I prefere for >> time being) >> - of cause ... to find a nice housing and screen to have the stuff mounted >> at wheel also to sea seamap at wheel >> >> One may say, hey, you can by this stuff, exactly what you want ... >> correct. But first of all, the hardware will cost me as DIY project about >> 100 ? instead of min 1000 ? ... the other thing is, it makes fun to create >> the own platform. >> >> So, if anybody is interesting to join this project, he/she is wellcome ;-) >> >> best regards >> Peter >> C&C34 Colomba _______________________________________________ This List is provided by the C&C Photo Album http://www.cncphotoalbum.com CnC-List@cnc-list.com
