|
|
05-10-2017, 14:28
|
#16
|
Registered User
Join Date: Aug 2009
Location: oriental
Boat: crowther trimaran 33
Posts: 4,449
|
Re: Autopilot - Open Source
Quote:
Originally Posted by Moo
Motion sensors are already integrated into autopilots and they monitor wave amplitude and the boats attitude using this data to dynamically adjust rudder input and gain settings - I have not seen the code, this is by observation. (I have spent thousands of hours observing autopilots!!!)
|
Which autopilots and what code?
Quote:
Modern off the shelf autopilots have a lot less controls and require a lot less calibration, then those of a few generations ago; which frustrates some. They do learn and adapt to the current conditions and do it quite well. In terms of the rate, amplitude and duration of rudder input as the boat lifts to a wave. (they also automatically work out things like counter rudder etc..)
|
Again what autopilots? My autopilot has many controls at the moment.... If it is automatically optimizing them, you can see what it is doing.
How does it automatically work out counter rudder?
Quote:
No amount of analysis is going to help an autopilot to steer in real sea conditions so I don't think there is much need to worry about mega complex analysis of wave theory. I would not use an autopilot that relied on theory.
The pilot can't disregard secondary, tertiary or giant isolated waves because that is the real world - hence cameras to enable it to see what is coming.
|
Yes, exactly, and this is the plan to incorporate this logic.
|
|
|
05-10-2017, 19:16
|
#17
|
Registered User
Join Date: Mar 2012
Location: Boston, MA
Boat: 1981 Bristol 32 Sloop
Posts: 17,994
|
Re: Autopilot - Open Source
I have a logitech camera to donate, which would work to get started, but you'd need 2 port and starboard and one astern. And the port and starboard would need to gimballed I think. You'd use the cameras facing the approaching wave direction.
But the DSP computing problem is still the same, after the visual processing, as motion sensors. At some point you'll need infrared for nighttime.
Observation: Discussion is discussion, and there is no need for dismisive statements particularly when talking ideas. Smile.
|
|
|
06-10-2017, 05:56
|
#18
|
Registered User
Join Date: Mar 2012
Location: Boston, MA
Boat: 1981 Bristol 32 Sloop
Posts: 17,994
|
Re: Autopilot - Open Source
|
|
|
06-10-2017, 09:22
|
#19
|
Registered User
Join Date: Mar 2012
Location: Boston, MA
Boat: 1981 Bristol 32 Sloop
Posts: 17,994
|
Re: Autopilot - Open Source
Goal: Through video processing techniques, wave analysis and learning algorithms.
1. Establish primary wave direction, amplitude and period.
2. Establish secondary wave direction, amplitude and period.
3. Identify and determine large anomalous waves direction.
1. Video DSP processing with emphasis on lowest possible frame rate & resolution to reduce calculations, but providing adequate contrast and edge definition.
- Extent of view: Large waves (minimum 2 if possible). Medium waves (3 to 4). Small (many).
- Frame rate: 1-5 frames per second
- Resolution: ?
Possible Approach
- Use DSP "Motion Compensation" Algorithms to subtract the differences between frames to establish changing areas of interest and reduce the amount of data to be analyzed.
- Use machine learning Pattern recognition and detection to establish the speed, direction and interval of the Motion Compensation areas. Conference on Computer Vision and Pattern Recognition - topics related to computer vision and pattern recognition—basically any topic that is extracting structures or answers from images or video or applying mathematical methods to data to extract or recognize patterns.
- Try to find or develop an wave or pattern algorithm which can separate the primary and secondary wave patterns.
- From the boat accelerometers get an average wave height and constantly compare that to the Motion Compensation areas. When there is an established change noticed in the motion compensation patterns determine what action to take.
Following the Conference on Computer Vision and Pattern Recognition leads, there is a mesh based approach which might be problematical with typical dark blue waves because there is not a clear edge or pattern that helps to define the wave recognition. The results may be somewhat anomalous in certain light or night time conditions. However this approach is taken with measurement of buildings & spaces which have definitive edges.This approach fits a mesh to a non-rigid surface, and is worth looking into or at:
2005 Best Paper: Real-Time Non-Rigid Surface Detection, Julien Pilet, Vincent Lepetit, Pascal Fua [pdf]
A new point matching algorithm for non-rigid registration - CiteSeerX [pdf]
Fast Non-Rigid Surface Detection, Registration and Realistic Augmentation
Real-Time+Non-Rigid+Surface+Detection Search
The problem with this approach is that it is much more computer intensive and the mesh then has to be interpreted for the basic data needed: wave direction, amplitude and period. However there are know algorithms and we know it has been done. Also Sean has worked a lot with meshes. If the computer can create the waves as simple line drawings, then I think the pattern can be analyzed with some known algorithms (which ones, I don't know).
I could spend all day on this, but the problem is at the basic level the same for use of cameras and accelerometers, determining patterns and differences, which has a mathmatical/machine learning solution. Math is powerful (and frustrating to some). The camera adds the potential to react before the wave arrives, which is a huge improvement (but not so huge if the smart autopilot is already projecting and anticipating some wave of about the average size.) It also adds considerably more computing load and challenges in the field of computer vision and pattern recognition, which does have a strong body of knowledge.
I'd rather have some cameras on board and a fantastic energy efficient autopilot operating my rudder/wheel than have the added weight of a wind self-steering device on the stern, but we have a small stern.
Sean, I hope you can figure this out.
|
|
|
06-10-2017, 12:02
|
#20
|
Registered User
Join Date: Jul 2017
Posts: 188
|
Re: Autopilot - Open Source
Of course an autopilot set to a compass heading will steer straighter, but steering in relation to the wind course varies because the wind varies. As you say it makes little difference but the big advantage is that it removes the constant sail trim adjustment a compass course demands and that saves a lot of time and effort when offshore. Obviously compass course is needed for near coastline or reefs. I think without a windvane, being able to use the windvane option on your tiller pilot is a big plus.
|
|
|
06-10-2017, 14:55
|
#21
|
Registered User
Join Date: Mar 2017
Posts: 804
|
Re: Autopilot - Open Source
Quote:
Originally Posted by boat_alexandra
Again what autopilots? My autopilot has many controls at the moment.... If it is automatically optimizing them, you can see what it is doing.
How does it automatically work out counter rudder?
|
Raymarine Evo .. virtually no controls. There is a three stage gain setting named IIRC Leisure, Cruising, Performance and one other gain setting buried deep in a menu.
No idea how it works out the counter rudder but it starts off steering quite erratically until it works out the basics and then it is off.
It needs more controls, sometimes you just want to jump in and tweak things.
Haven't used it on a boat smaller than 55' though although it is designed for the leisure market.
|
|
|
06-10-2017, 14:58
|
#22
|
Registered User
Join Date: Mar 2017
Posts: 804
|
Re: Autopilot - Open Source
Quote:
Originally Posted by rgleason
[FONT=Arial]Following the Conference on Computer Vision and Pattern Recognition leads, there is a mesh based approach which might be problematical with typical dark blue waves because there is not a clear edge or pattern that helps to define the wave recognition. The results may be somewhat anomalous in certain light or night time conditions. However this approach is taken with measurement of buildings & spaces which have definitive edges.This approach fits a mesh to a non-rigid surface, and is worth looking into or at:
|
Don't forget that it is vital that this works at night time. So the cameras will have to be capable of night vision. Thermal will not work it has to be image intensification.
|
|
|
06-10-2017, 15:03
|
#23
|
Registered User
Join Date: Aug 2009
Location: between the devil and the deep blue sea
Boat: a sailing boat
Posts: 20,945
|
Re: Autopilot - Open Source
Quote:
Originally Posted by rgleason
Goal: Through video processing techniques, wave analysis and learning algorithms.
1. Establish primary wave direction, amplitude and period.
2. Establish secondary wave direction, amplitude and period.
3. Identify and determine large anomalous waves direction.
(...)
|
Close. But not good enough.
These are not like sound waves that you can decompose from the compound and get information on the trains. This is so because in case of sea waves the primary and secondary data are averages. You can easily imagine what happens to the decomposing algo when you allow for the quasi normal distribution of actual direction amplitude and period (-s).
gigo
So try the other method. Listen to the waves, build a database and check this periodically for data drift. Then (try to) tie boat g vectors with wave train data that preceded the vectors. Now you are talking.
Mind consequences of the mistakes such a "smart" AP will make now and then. Are they worth the potential gains.
Otherwise NKE would have been there long ago.
What are we at then?
b.
|
|
|
06-10-2017, 15:38
|
#24
|
Registered User
Join Date: Aug 2009
Location: oriental
Boat: crowther trimaran 33
Posts: 4,449
|
Re: Autopilot - Open Source
Quote:
Originally Posted by Moo
Raymarine Evo .. virtually no controls. There is a three stage gain setting named IIRC Leisure, Cruising, Performance and one other gain setting buried deep in a menu.
|
I think having only 3 settings plus a hidden one to tweak the autopilot sounds really lame.
Quote:
No idea how it works out the counter rudder but it starts off steering quite erratically until it works out the basics and then it is off.
|
So no one knows how it works? Sounds like it's impossible for anyone to make use of this logic or improve it.
As for the camera at night... Yes well maybe a problem. It could have a spotlight I suppose. In any case, at least it could work during daytime which is better than nothing. It's also a problem to see the waves at night as a human.
The camera also adds the possibility to learn faster, and calibrate the other sensors better.
|
|
|
06-10-2017, 16:13
|
#25
|
Registered User
Join Date: Mar 2017
Posts: 804
|
Re: Autopilot - Open Source
Quote:
Originally Posted by boat_alexandra
I think having only 3 settings plus a hidden one to tweak the autopilot sounds really lame.
So no one knows how it works? Sounds like it's impossible for anyone to make use of this logic or improve it.
|
Well obviously somebody knows how it works. FLIR and Raymarine employees I suppose - I'm just commenting in a general way as a driver of yachts equipped with it. I am an electronics engineer but have not dug into this; after all it is not fitted to my own boat
Quote:
Originally Posted by boat_alexandra
As for the camera at night... Yes well maybe a problem. It could have a spotlight I suppose. In any case, at least it could work during daytime which is better than nothing. It's also a problem to see the waves at night as a human.
The camera also adds the possibility to learn faster, and calibrate the other sensors better.
|
Yes I agree that it could be very useful in the calibration stage, except for a spotlight which won't work very well even if it is IR as flying spray in breaking wave conditions will blind it; if visible light it will blind the humans aboard. Perhaps some kind of other vision system such as acoustics, or RF ranging would be better. Less easy to do DIY on a budget though.
|
|
|
06-10-2017, 17:14
|
#26
|
Registered User
Join Date: Aug 2009
Location: between the devil and the deep blue sea
Boat: a sailing boat
Posts: 20,945
|
Re: Autopilot - Open Source
"... So no one knows how it works? ..."
Which reminded me of another great quote:
"Everybody agrees something can't be done and then comes the guy does not know that it cannot be done and he does it".
Clearly. It can.
Cheers,
b.
|
|
|
06-10-2017, 17:20
|
#27
|
Registered User
Join Date: Aug 2009
Location: between the devil and the deep blue sea
Boat: a sailing boat
Posts: 20,945
|
Re: Autopilot - Open Source
If seeing only the nearest wave helps then possibly proximity sensors like in self parking cars could do.
https://www.nytimes.com/2017/05/25/a...ving-cars.html
?
Expect drt cheap lidars within next 5 years.
b.
|
|
|
06-10-2017, 17:46
|
#28
|
Registered User
Join Date: Mar 2012
Location: Boston, MA
Boat: 1981 Bristol 32 Sloop
Posts: 17,994
|
Re: Autopilot - Open Source
IR Light, Acoustics, or RF...How about Lidar? (the sensors are getting better and will be cheaper because of Robot cars..)
https://oceanservice.noaa.gov/facts/lidar.html
https://qz.com/1088163/robot-cars-ne...ght-have-them/ “Everyone’s trying to throw more and more software at what’s essentially a hardware problem,” Russell says. “You see some of these big companies hiring thousands of software engineers to try to make sense out of just a few points that are out there, and to do a lot of guessing.”
Up on the hill, I can see a large black target I thought was for tanks. Russell says it’s actually a very dark black facade that his team uses to calibrate how well Luminar’s sensors can detect objects that reflect next-to-no light.
...
Luminar announced in April, along with its funding, that 100 of its boxes were being tested in the wild, and that it would eventually start manufacturing 10,000 units later this year at its Orlando facility. The company has been working privately with four partners, but on Sept. 27, it announced that Toyota Research Institute (TRI), the division of the Japanese automaker that deals with robots and self-driving cars and is run by the former head of the DARPA Robotics Challenge, will be publicly testing Luminar’s technology in its new autonomous vehicles.
The Lidar seems to have everything that is needed already built in to the hardware and software. (Maybe Luminar would have some beta device that could be tested?) I don't know its soft point or how much energy it uses.
But final choice of sensing device, does not need to be made right away, I think, maybe pick some other device that works and learn how it can improve the AP.
What is Lidar? Noaa.
https://oceanservice.noaa.gov/facts/lidar.html
https://www.sparkfun.com/products/14032
|
|
|
06-10-2017, 18:56
|
#30
|
Registered User
Join Date: Mar 2012
Location: Boston, MA
Boat: 1981 Bristol 32 Sloop
Posts: 17,994
|
Re: Autopilot - Open Source
Quote:
"White Water
Weather, snow-free, leaf-off, etc.
Good water column clarity
What about state of the sea surface?
Topobathy LiDAR – when is a good time to conduct a survey?
Large ocean waves are great for surfing, not for LiDAR!"
--Yes, that's when you want to see the ground below! If you are looking at the water itself, they should be fine. It all depends on what you are trying to do.
https://geozoneblog.wordpress.com/20...than-it-looks/
Also Eye Safety!!
. And I appreciate the eye-safety problem as a new laser safety officer would not let my beam-spread eye-safe laser be used because the laser INSIDE the lidar was not eye safe
|
|
|
|
|
Thread Tools |
Search this Thread |
|
|
Display Modes |
Rate This Thread |
Linear Mode
|
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
|
Advertise Here
Recent Discussions |
|
|
|
|
|
|
|
|
|
|
|
|
Vendor Spotlight |
|
|
|
|
|