free autopilot

[SecondBase]

Originally Posted by boat_alexandra

I am looking for suggestions for my free autopilot see:

http://pypilot.org/img/demo0.avi
http://pypilot.org/img/autopilot2.avi

The source code is available:

github.com/pypilot/pypilot

.

Way to share.

Not sure if it will help you but if it does--I have a new in box CanaKit Raspberry Pi 3 Complete Starter Kit - 32 GB Edition. If you are stateside PM me a mailing address and I'll ship it to you no charge.

[seandepagnier]

I sail tomorrow from north carolina to Massachusetts and will try to use the autopilot as much as possible. I have two complete units for testing.

0) raspberry pi with power supply and sd card.

1) I have a simple hat that plugs onto any raspberry pi for the first 10 pins only. This hat has an IMU attached, and connects also the serial port to the motor controller. It can hold an optional ir sensor for remotes.

2) Motor controller. This box has power (2 wires) and serial data (4 wires) going in, and motor drive (2 wires) going out. I used a waterproof plastic box. It fits in a box 3"x2"x1.5" For bigger motors (> 3 amps) need a larger box.

3) linear ram - I have been using simrad tiller pilots that have the fortunate tendency for the magnets to slide out of place, and break the hall sensors off leaving them brainless. Any motor that can move the rudder both ways will work in theory.

4) (optional) hat for raspberry pi using the next 14 pins (pins 11 - 24) to provide a 5 button, and cog lcd user interface.

5) (optional) weather sensors. These plug in usb, and provide wind speed, direction and barometric pressure. Can steer to true wind or apparent wind.

6) (optional) gps This is for steering to gps course. It is possible to use most usb gps, or relay gps in through the network.

6 (optional) usb wifi to allow external control (raspberry 3 and w already have wifi)

7 (optional) ir receiver and remote control. Can use any tv remote with lird to control the autopilot.

8 (optional) smart phone, connect to autopilot ap with wifi, and use browser to configure and control with touch screen.


All of the above is implemented and working well. There are more under development.


Everything is working perfectly on the orange pi zero, which is fast and much cheaper than raspberry pi.

[barnakiel]

(Optional) - you can use smartphone sensors (gps, acceleration, etc.) as input to your AP motherboard.

This is easily done wireless-ly by harvesting your smartphone's sensor data and piping this data over serial BT or TCP (depending on what your Ap softwrae / hardware accepts).

Very easy. Cheap. Most owners have smartphones anyway.

+Hugs,

Cheers,
b.

[jim321]

i've got two or those orange pi zero's could never find anything that worked right on them i wanted to use one for a pihole i cant remember what the problem was, maybe i'll dig one out and give it a try..
running op on a pi3 now and have seen your project on that forum also..
still waiting on bits from china to do more.
good job
jim

[seandepagnier]

Originally Posted by barnakiel

(Optional) - you can use smartphone sensors (gps, acceleration, etc.) as input to your AP motherboard.

This is easily done wireless-ly by harvesting your smartphone's sensor data and piping this data over serial BT or TCP (depending on what your Ap softwrae / hardware accepts).

Very easy. Cheap. Most owners have smartphones anyway.

If you just use whatever sensors in whatever smartphone, you won't be guarenteed the same resolution, and calibration maybe different. Different sensors have different drivers and different resolutions. Many of the compass in smart phones are not very good. This leads to a lot of variability.


The mpu9255 is available for $4 and has good resolution and plugs into the raspberry via i2c, this is what I use. Maybe someday I will support "any" smartphone, but I think it can lead to a lot of problems.

Also, you cannot just "pipe out" on bluetooth or wifi and expect good results. This will add considerable lags and hiccups which greatly decrease ap performance. You don't want an accidental jibe because your bt link reset. I don't even recommend usb for this. I am using only i2c and serial for this critical communication. You can send wind or gps over wifi, and it can automatically fallback to compass seamlessly if the signal drop.

I just sailed 1 week using the autopilot in the ocean. Average power consumption was 4 to 5 a/h per day from 12 volts. This is using a very inefficient ram. With a very efficient ram, and smarter brain, this could be 1a/h per day. I can say it works much less than with the original simrad brain, it doesn't move at all when the boat rolls. It probably uses about half the power for the same course keeping ability.

I found that I can get average power consumption very low if I accept course errors. They are still small enough to keep the sail full, but the autopilot doesn't work to compensate wave motions. This is how I always sailed with the wind vane, and it's working. Using 1.4 watts average steering 4.5 knots wing and wing.

I found that upwind, or in very light conditions, the autopilot can use very little power, about half a watt, thats less than the raspberry itself! If the boat is balanced, the autopilot doesn't do much until a wave knocks it or something.

In stronger conditions, 20 knots or more, running downwind, the tiller pilot begins to struggle a lot. It's using about 5 watts or more. I think if the brain were smarter it could do much better.

[barnakiel]

Originally Posted by boat_alexandra

If you just use whatever sensors in whatever smartphone, you won't be guarenteed the same resolution, and calibration maybe different. Different sensors have different drivers and different resolutions. Many of the compass in smart phones are not very good. This leads to a lot of variability. (...)

Choices, choices, horses and courses.

I simply pick up whatever the phone offers (mostly gps and accelleration) and format it for whatever the target device wants. This is simple. BT or wifi link to the brainbloc then. But my toys are just that - I do not ask them to drive my boat while I take rest.

I am 100% with you on some sensors' accuracy. I have found some smartphone gps sensors quite bad. Then again, their accelerometers seem pretty impressive. Not below the sensors I use on Arduino.

You are building the hardware so you know 100% what works, what not. Ideas at time may sound good and only then we discover that they do not work in real life environment. Been there too. Many of my original ideas were bad, few were fine.

I am most impressed with your skills. PLS keep on rolling as elsewhere the world is right now going a bit nuts about "I , me , mine , and myself". Building for others, sharing, seem at an ebb.

We need good APs as much as we need attitudes like yours.

Cheers,
barnakiel

[jim321]

where can i find the image you referred to in a previous post for opi zero.

" 3) The autopilot can use the much cheaper orange pi zero instead of the raspberry pi. I can provide debian images for this. "

thanks
Jim

[seandepagnier]

There should be 3 images soon:

1) tinypilot -- highly stable (runs on ramdisk) linux distro for raspberry pi. It may support orangepi someday
2) debian/orange pi -- basically just debian on the orange pi with pypilot installed and running automatically as a service at boot
3) openplotter -- it is planned to support pypilot, soon. Not sure if the next release or when exactly as I do not maintain openplotter.


For all other systems, you can simply pull the code from github and install it.

I didn't upload any images just yet. It is extra work to make the image, package, compress it, and upload it. I am waiting until I can get some kind of web store going with "kits". I have 3 autopilots right now, with 2 more very soon. You can PM me if you are interested in my hardware, but keep in mind it is still very experimental.

Of course you are free to build your own, and I am glad to support this, but there are many pitfalls, it's probably easier and saves a lot of time to get hardware from me.


I finally tested in protected waters in drifting conditions running wing and wing, catching puffs of wind (sails not always full) and the power consumption for the motor dropped to 0.2 watts average. The wind vane cannot steer in these conditions, as I no longer have a light air vane, and my homemade vane rudder is much heavier than the original. Even if the wind vane were sensitive enough to work, it would not be able to steer nearly as straight. The boat speed was 1-2 knots.

I am working on a high power controller for boats 10-50 tons. My rating is very conservative. The controller I have overheats at 50 watts continuous (I consume 3-4 watts at 20 knots on my 4 ton boat), only because it's in a sealed plastic box. The higher power controller will handle 5x the power of this one. Also 24 volts anyone??

I am working on implementing a neural network to make machine learning possible. This will make it automatically learn, but also be able to take sensors not normally used, like a camera looking at the waves.

[NahanniV]

Why the special version of the MPU9255 board ?



Is this open source as well ?

Would it be better to move the compass to a location remote from the electronics ?

[rgleason]

The shop cart is empty now.