Autopilot Control

[seandepagnier]

Originally Posted by rgleason

mecharduino https://www.kickstarter.com/projects...strial-servo-m

This is the project I find most interesting. It is an efficient way to drive steppers.

Sean are these strong enough and fast enough?Overview of Specification
Microcontroller: SAMD21G18A
Encoder: AS5047D
Motor Driver: A4954
Motor included with Mechaduino 0.1 Servo: 17HS16-2004S1
Magnet: Diametrically Magnetized NdFeBr

It depends on the boat. Bigger boat needs more torque, more difficult conditions needs more speed. I know someone who uses a stepper motor and bicycle chain to gear it down and turn the wheel.

[CareKnot]

Please forgive the delay in responding. I have been otherwise occupied. Let me see if I can clarify some things.

A little over a year ago I was single-handing my sailboat southbound on the ICW through South Carolina. I had no AP and was using OCPN on a tablet. Nauti One was putting along at about 3 knots against the current. It was winter, late at night and I was cold, hungry and tired.

I don't know how I missed it. Maybe I dozed off at the helm. All I remember was the boat lurching to a stop with the bow squarely planted in a large inflated yellow hazard marker. Yikes! That is when I got interested in collision avoidance for marine applications and dynamically updating electronic sharts in real time.

Originally Posted by boat_alexandra

I sailed today for the first time with the autopilot route plugin.

The results were good in "route position bearing" mode. This is the mode that does not use xte, but instead finds the optimal position to steer to that is (I set to 100) meters ahead, and takes that bearing. This means a continuously changing bearing.

I have several bugs to fix, and I did change opencpn, so it won't be available until the next beta series.

Congratulations on the successful shakedown. May I presume that the 100 meter value will end up being a calculated variable based on available room to maneuver and vessel dimensions? It would seem that you would want a different value for various sized boats in different conditions.

Tight channels and meandering rivers would require different tactics than say, the open ocean, right? If I find that an error has placed me out of the channel, I usually either backtrack to the channel or aim for deeper water if the way is clear. Usually I then turn again to resume the route when I reach it. I would think you would want the AP to emulate similar behavior. Is this doable?

Originally Posted by boat_alexandra

This level of precision is nice, but not really useful on an autopilot because the precision of a brushed motor is an order of magnitude less than what the autopilot compensates for in normal operation anyway.

We don't need to print things, or make a laser cutter or 3d printer for example... just move the rudder a few degrees +- 10% is ok. I'm not saying stepper motors won't work, they absolutely do, but generally, unless the stepper motor is closed loop (has shaft position feedback to the controller) it will be less efficient than other types of motors. If it is closed loop, it becomes much more efficient, on par with brushless motors, and also very accurate, but becomes more expensive, maybe worth it. Check out the mecharduino project.

I get that and agree. About the other, I was just shopping ideas. Most of the solutions I read about were closed loop and gave position feedback, so an accurate rudder position indicator and precise control is doable. As far as affordable, I haven't crunched those numbers yet.

Btw, I am not married to the use of stepping motors. Just considered them because of reliability, variable speeds and precision and the availability of numerous RC controllers of that type. I found relatively inexpensive RC solutions that use stepper motors have some advantages worth considering. But like I said, I'm just spit-balling at this juncture.

Thanks for the mecharduino project info. Very interesting stuff. I wonder what they mean by 'less expensive'.

Let me back up for a minute. I'm not sure if you understand my intentions (and forgive me if I stray from the OP). I really wasn't very clear.

So here is my thinking, if a bit ambitious: For my suggestions, I'm focusing more in the 'what', as opposed to the hows or whys of navigation or AP. AP solutions for OCPN, no matter the technical options, would ultimately benefit from more chart data and more accurate chart data. This is why I mentioned a continuous scanning laser device to collect vector data (aside from it's other uses more suitable to passage making, which I'll get into later).

OVERVIEW
In a nutshell, I am informally evaluating a system of location specific data collection, processing and distribution. The minimum primary table record would contain fields for Lat, Lon, Alt, Date, GMT, Descriptor, Name, Type. Support tables would provide static and dynamic data as needed. These data structures have yet to be fully defined (though I have a general idea what might be needed). Data from other input sources are also needed, but it has not been evaluated or accounted for at this time. Table structures will have to be adjusted accordingly.

DATA COLLECTION
Data collection would be automated and transparent to the end user. Seperate functions would allow optional operator contributions.

DATA PROCESSING
Some extrapolation of data is possible. Unknown opportunities will be presented as the system evolves. Appropriate changes will need to be made.

Examples:

1. Where GPS marks the approximate 'center' of a location, laser scans can provide hyper-accurate boundary mapping. This includes everything from marinas to single pilings to uncharted pipes.
2. Vessel location data could be accurately calculated from established data points.
3. Dynamic objects of all types can be tracked and courses projected independent of AIS data. Approximately 3m error on vessels at about a mile. Plenty of time to get out of the way.
4. Imaging and overlays, even in extremely poor visibility, even through darkness, smoke, dust, fog, etc.
5. Where OCPN is mapping static object boundaries with every laser equipped unit, the location/object data layer rapidly becomes very robust and very accurate in high traffic areas. Further investigation would be needed to properly evaluate performance and application.

Later on, might there be an opportunity for OCPN to collect on government grants to supply cartographic data to the government and to fund laser units for participating vessels?

Originally Posted by boat_alexandra

As for sharing the data. I have considered the idea of using wifi for this for passing vessels, but generally it's impractical. It might be more useful in anchorages.

I was thinking of something a little different. There are numerous digital radio modems available that are easy to install, configure, are feature rich, transmit anywhere from 2-3 mile line-of-sight to 100 miles depending. The smaller units can be had for well under $30. This one is rated at about 5 miles and sells for around $20: https://www.ebay.com/i/192161756489?chn=ps

Two boat moving at 10 knots in line-of-sight conditions over a 2 mile distance would afford about 10 minutes to pass each other and then pass out of range. That should be adequate for updating local charts, though 5 miles would be better

I would suggest OCPN handle the unit identification, verification, handshaking and data transmission. Comparing chart layer indexes checksums would determine if chart layer data transmission is necessary.

If yes, records are isolated for transmission, based on index comparison. Only missing or records that do not match are updated. Original records are retained.

If no, transmit next set of indexes. Continue until update complete or radio link is lost. The actual data requirements for transmission are very small. I estimate that numerous charts could be evaluated and transmitted fairly quickly.

Alternately, Wifi or cell data can update online databases where any user has access. But there are a lot of details to work through for this approach. Again, spit-balling.

Originally Posted by boat_alexandra

Would the laser be dangerous? Scanning the waves would be useful, would it get returns from water?

Lasers are safe if standards are met. I am looking at are low power units with frequencies not in visual range. It's not like we are trying to bounce one off the moon. This is fairly close line of sight stuff.

In theory, anything that reflects a usable amount of light can be measured and tracked. So yes, waves are definitely on the table.

Originally Posted by boat_alexandra

Already GPS is more accurate than your corrections from a moving vessel in most cases. Now there are european and russian satellites to make it very accurate.

Not sure if that is completely true, but I agree in general practice.

With a laser, accuracy generally decreases as range increases. If I am scanning nearby object for the chart overlay, we can be talking about as little as +/- 2cm.

The speed of light is 299,792,458 meters per second or about one foot per nanosecond. A boat wouldn't move much in the time it takes for light to make the round trip to a target 1/4 mile away. If the motion is perpendicular to the light source, the effect on accuracy would be nearly undetectable. If not, it is correctable.

Originally Posted by boat_alexandra

Having eyes on the autopilot is a long term goal of mine. The vision would be used in the same way as image recognition is utilized by driverless cars.

There is another device I am considering. We should talk.

Originally Posted by boat_alexandra

There is a serious issue with mis-trusted data, and poorly calibrated sensors. Depth sensors need to know the water temperature and to a lesser degree the salinity.

These factors change, and even if the temperature is measured at the surface, it cannot be easily measured for the entire water column.

Furthermore, even with tidal corrections, low pressure systems also affect water levels, as well as rainfall. It is difficult to deal with this. The best bet would be to map relative water depth finishing the survey at the starting point.

Consider the case of navionics sonar charts. They tried to do what you describe, and failed at it, causing numerous boats to run aground. Now the data might be interesting but cannot be considered reliable.

I understand your reluctance to bring this functionality online, and rightly so. However, if we already have the data, it might be worth examining the issue further. Averaging large data sets, removing record that correspond to anomalous events, etc., could be a good start. In the end, all we want to do is correct or augment existing chart depth data in relatively shallow water.

Again, just spit-balling...

Originally Posted by boat_alexandra

The question of hazardous objects can be dealt within opencpn. A camera can see other boats much like a human would, and then alarm is sounded and from there, the autopilot could be commanded to evade.

I was thinking about short term course projections for drifting hazards too. But it's just a thought.

OK, time to go check on the boat. Thanks for all the input. Let me know what you think.

[CareKnot]

Originally Posted by rgleason

mecharduino https://www.kickstarter.com/projects...strial-servo-m

Sean are these strong enough and fast enough?Overview of Specification
Microcontroller: SAMD21G18A
Encoder: AS5047D
Motor Driver: A4954
Motor included with Mechaduino 0.1 Servo: 17HS16-2004S1
Magnet: Diametrically Magnetized NdFeBr

Rick, please check your PMs.

[CareKnot]

This is the idea that Gene and I discussed at length. Ultimately, he passed it along to Uncle Sam, no strings. https://www.google.com/patents/USH1740

The applications for this technology are manifold and extreme, but I have a more mundane suggestion that might enhance smart AP controls.



Cameras developed from this technology have ranged from very high speed to very long distance that produce very detailed images. They incorporate enhancements like specialized electronics, mechanisms and lenses that I don't think will be necessary for this application. Plus, I have some ideas that might gain us a no-royalty agreement should we try to develop this idea further.

Most of what I suggest has been done before, but in different configurations if my information is correct. Anyway, I propose a series of gated CCDs positioned vertically along the axis of a rotating prism. Each linear CCD array is individually matched and aligned to the frequencies desired, IR through ultra-violet. The number of color channels requires further discussion. The prism and CCD arrays are contained in a housing with three thin vertical apertures. It would rotate at a rate designed to expose the CCDs x number of times across a 120° field of view depending, yielding 3 images per frequency that cover a full 360°.

The viewer (or analyser) would combine selected monochrome images into a composite for the best contrast and detail to suit various circumstances. For instance, different materials react to or reflect different frequencies better than others. By cycling through various combinations, floating and fixed hazards might be more easily identified and avoided.

I would like to incorporate the output from the scanning laser I described in another post to add vector data to pixel groups. Limited perspective and resolution 3D images could be constructed in this manner. Perhaps this would be adequate to track large waves that would be desirable to navigate around. No 2D object recognition software required because object are already 3D mapped.

Does anyone 'get' what I'm going after here?

[seandepagnier]

Originally Posted by CareKnot

This is the idea that Gene and I discussed at length. Ultimately, he passed it along to Uncle Sam, no strings. https://www.google.com/patents/USH1740

The applications for this technology are manifold and extreme, but I have a more mundane suggestion that might enhance smart AP controls.



Cameras developed from this technology have ranged from very high speed to very long distance that produce very detailed images. They incorporate enhancements like specialized electronics, mechanisms and lenses that I don't think will be necessary for this application. Plus, I have some ideas that might gain us a no-royalty agreement should we try to develop this idea further.

Most of what I suggest has been done before, but in different configurations if my information is correct. Anyway, I propose a series of gated CCDs positioned vertically along the axis of a rotating prism. Each linear CCD array is individually matched and aligned to the frequencies desired, IR through ultra-violet. The number of color channels requires further discussion. The prism and CCD arrays are contained in a housing with three thin vertical apertures. It would rotate at a rate designed to expose the CCDs x number of times across a 120° field of view depending, yielding 3 images per frequency that cover a full 360°.

The viewer (or analyser) would combine selected monochrome images into a composite for the best contrast and detail to suit various circumstances. For instance, different materials react to or reflect different frequencies better than others. By cycling through various combinations, floating and fixed hazards might be more easily identified and avoided.

What floating and fixed hazards are more visible with IR or UV?

I would like to incorporate the output from the scanning laser I described in another post to add vector data to pixel groups. Limited perspective and resolution 3D images could be constructed in this manner. Perhaps this would be adequate to track large waves that would be desirable to navigate around. No 2D object recognition software required because object are already 3D mapped.

Does anyone 'get' what I'm going after here?

Yes, if you can make a scanning laser this would be useful. It sounds difficult. I'm still struggling to build a sonar.

[CareKnot]

Originally Posted by boat_alexandra

What floating and fixed hazards are more visible with IR or UV?

Hunters learned a long time ago that 'brighteners' used in laundry detergent made them more visible to deer. Turns out that deer's vision dips into the violet range and the detergent made their camo 'glow' against a dull background of vegetation and earth. So it is established the certain material is UV reactive. I presume that this phenomena applies across the spectrum, considering that Spectral analysis is an established science. Various materials react with various light frequencies in a way that could be isolated with a device like this.

Could not give you a list off the top of my head, but I'll dig around a bit when I get some time. I was hoping that some members might have some personal experience with the science.

Originally Posted by boat_alexandra

Yes, if you can make a scanning laser this would be useful. It sounds difficult. I'm still struggling to build a sonar.

Probably is. Don't know. I know that different aspects of this idea have already been produced in different products. I presume some product and component research will reveal whether this can be scabbed together or whether it will have to be a 'ground up' project.

[CareKnot]

Correction: Building a scanning laser may not prove to be that difficult. These types of devices are already produced for a number of commercial applications. Modifying an existing product is probably where I would start. I already have a spreadsheet that is used to work out the math for ranging, etc.

I am hoping that I can find similar products to modify for the rotating prism camera. Time will tell. In any case, I will need some help.

[rgleason]

That would be nice, we looked into lidar units for Cars but that apparently does not do well with water.

http://www.cruisersforum.com/forums/...ml#post2495039

http://www.cruisersforum.com/forums/...-191315-2.html

long discussion about it...
http://www.cruisersforum.com/forums/...ml#post2494393

[rgleason]

For weatherfax_pi

travis.yml is 48 lines
https://github.com/seandepagnier/wea...er/.travis.yml

appveyor.yml is 70 lines
https://github.com/seandepagnier/wea...r/appveyor.yml

[CareKnot]

Originally Posted by rgleason

That would be nice, we looked into lidar units for Cars but that apparently does not do well with water.

http://www.cruisersforum.com/forums/...ml#post2495039

http://www.cruisersforum.com/forums/...-191315-2.html

long discussion about it...
http://www.cruisersforum.com/forums/...ml#post2494393

Thanks Rick. I read a lot of the discussion and there is some pretty great stuff.

One of the problems you will run into with most Lidar, especially hand-held rangefinders is that most of the units are very low power and use a diffused beam or incoherent light based on the lower Class standards (see: https://en.wikipedia.org/wiki/Laser_...Revised_system).

There are some workarounds that I would suggest, but I'm waiting to hear from a member that PMed me expressing approval and interest. Said, "...a very wide background as an Engineer in the Aerospace industry." AND "We did mostly IR systems and components, with a little smattering in the other optic domains."

I'll let you know how it goes. Frankly I'm excited!

[CareKnot]

There are a number of patented and non-patented applications of laser technology out there They cover a wide range devices including hand-held rangefinders and sophisticated laser imaging applications. Let me go on record, stating publicly, that I release all ideas I have stated in this thread under Creative Commons, "CC BY 4.0 International", and further state that OCPN may use my ideas as the developers thereof see fit, where other preemptive patents and rights are not infringed.

It is my belief that we can navigate through the maze of applications to develop some of the aforementioned ideas, for the stated purpose of navigation and more specifically, AP control. I do not think that studying the work of others for inspirational purposes, in any way infringes on their proprietary rights. Hopefully, we will be able to produce a 'better mousetrap'.

The following quote is from the article entitled, "Laser Imaging Video Camera Sees through Fire, Fog, Smoke" is of particular interest, in that we are attempting something similar, albeit through different methods.

"Light travels at one foot per nanosecond, so the camera might send out a 10-nanosecond pulse of light and then wait about 50 nanoseconds to open the gate. In the interval, the light has passed through the obscurant, which reflects and dissipates some of it, then bounced off of whatever happens to be around 25 feet away, and returned to the camera.

“When I turn the camera on, all that near-term reflected light is gone,” Billmers says. “I can get rid of the near-term scattering, so I’m not blinded.” The laser light also scatters less than normal white light.

By adjusting the time that the gate waits to open, Billmers can train the camera on objects at different distances, and it is capable of reaching up to a couple of miles."

The ideas in my earlier posts have similarities to this technology though again, are implemented differently and for different purposes. The more I think about it, the more I believe this is doable.

LASER is by definition, "Light Amplification by Stimulated Emission of Radiation". However, MASER ("Microwave Amplification by Stimulated Emission of Radiation") is of equal or greater interest to me at this point.

By using multiple continuously scanning beams of multiple frequencies of visible and invisible light in both coherent and incoherent alignment and filtration based on various angles of phase and return timing, a broad range of options are opened up to us in the development process. It give us the ability to pick and choose what information to display and how to display it.

Rather that producing the type of imaging associated with color cameras, we would be replicating highly accurate monochromatic vector data that could be recombined in a manner consistent with the goal of ranging and identifying objects based on materials. This seems a much more useful set of tools to a navigation system and to a mariner in general.

In this application as with all Laser usage, safety is a function of how much energy is focused at a specific area over time. This is especially true of visible frequencies of light. It seems to me that using current technology, we can select from visible and invisible spectrums depending, choose how we focus that energy and limit its time on target based on what is needed for a particular purpose. Limiting pulses of coherent energy to picoseconds in length and shorter is routine.

These observations gives us the option to increase energy output of a pulse and tighten the focus of the radiated energy, all the while remaining well within current safety standards, per ANSI Z136 series of standards. I hope this clarifies some of the issues we have already discussed.

[CareKnot]

Corrections to the Above:
Where is said, "I release all ideas I have stated in this thread under Creative Commons...", I should have included, "As I have already recorded elsewhere..."

Where I said, "Rather that producing the type of imaging associated with color cameras, we would be replicating highly accurate monochromatic vector data that could be recombined in a manner consistent with the goal of ranging and identifying objects based on materials...", I should have included, "...,size, shape and other factors."

[rgleason]

Very Interesting ideas Phillip. You obviously know far more about this than me. You mention Microwave as a possible spectrum. How would that help?
Are there devices that can be used now, or do these need to be built new?

[CareKnot]

Originally Posted by rgleason

Very Interesting ideas Phillip. You obviously know far more about this than me. You mention Microwave as a possible spectrum. How would that help?
Are there devices that can be used now, or do these need to be built new?

Hi Rick,

Yeah, I talk a good game, but truth be known, I am woefully behind the times. In another life I developed new applications and devices for dedicated microprocessor systems and wrote machine code in hexadecimal. Though not an engineer nor a technician, some strange devices have been cobbled together by my hand.

I have a unique form of dyslexia that affects my abilities in mathematics. While I have difficulties in that area, there are other things that I instinctively 'know'. When I started out sailing and racing tornado class cats, my instincts made me a reliable tactician. Fortunately for me my teammate was an excellent helmsman and a very technical sailor. We made a great team. That is where I excel; as part of a team.

More recently I worked with my friend Gene, developing a new approach to vocoders for cell phones and some other technologies too complex to detail here. Though extremely successful in our implementation, it did not turn out well for us. That is all I will say about the matter.

I was once known as an 'idea man' that could make connections that others missed or flat out could not see. My meager accomplishments building early computers and control systems were no match to the really talented people with whom I worked. When I had the occasional flash of inspiration, it was usually other people that worked out the details.

To answer your question, yes many technologies are built from semi-modular components. So in my estimation, significant portions of what I describe can be assembled from available inventory. On the other hand, some portions will need to be designed and built from scratch. In any case, prototyping could be costly - or not. It just depends on the approach we choose and who steps up to help. It's no secret that there are more ways than one to skin the proverbial cat.

When I mentioned microwave frequencies, you may safely assume I meant radar, though not in the traditional sense. This is something that needs a lengthy and in depth explanation that is best left for later. Right now I'm very tired.

Even with all the latest refinements, I always viewed radar as a power hungry shotgun approach that could be more economically implemented when paired with the right technology. Some of this I will explain soon. But the rack is calling. Be well.

[Douwe Fokkema]

Originally Posted by boat_alexandra

So it's possible to send +1 -1 +10 -10 commands with a nmea2000 message to raymarine?

Almost. There is a proprietary PGN that allows setting the heading of the Raymarine EV-1 pilot when it is in auto mode. This is PGN 126208 with a data length of 14. In canboat format this pgn looks as follows:
Z,3,126208,7,204,14,01,50,ff,00,f8,03,01,3b,07,03, 04,06,f0,3d
The last 2 data bytes indicate the heading to be set in thousands of radials, least significant first (divide by 174.53 to get degrees). So the above will set a heading of 91 degrees.
This is tested with canboat and a Actisense NGT-1 and functions flawless.