CPA calculation - ship sizes - closer proximity

[rgleason]

http://www.sailvalis.com/NavMon/AisDetail.html

Found this useful and thorough from Paul (NavMonPc)

If no gps position info defaults to port bow.

-gps position info is 4 distance to of ship


Sent from my SCH-I545 using Cruisers Sailing Forum mobile app

[rooiedirk]

I'm not aware of any other program doing so, but it would make sense if:
cpa < 3, 4 or 5 times target length
AND
if on screen the cpa info (in rollover or AIS target info) it is made clear that this cpa is
to bow, stern PS etc.
We should be aware of course, that AIS as a collission avoiding tool on such a short distance has it disadvantage. Not to say is dangerous.
Dirk

[Paul Elliott]

Originally Posted by rgleason

NavMonPc

Found this useful and thorough from Paul (NavMonPc)

If no gps position info defaults to port bow.

-gps position info is 4 distance to of ship

I'm not sure that that last part means!

In NavMonPc I *don't* do Kalman filtering on the GPS position change, I just use the transmitted SOG/COG.

I also use extrapolated (dead reckoning) positions for the other ships, where I extend the position based on last known position, course, and speed. This reduces the inaccuracy caused by the AIS reporting intervals, but introduces the potential inaccuracy should a ship change course or speed during this interval. The AIS update is faster when a vessel is turning, so if everything is working as it should the CPA/TCPA calculations should be better with the extrapolation than without. The extrapolation can be turned off if you prefer.

In the NavMonPc AIS link posted above I state that a lost-signal target will vanish from the screen after ten minutes. This may have been the case in early NavMonPc versions, and is (I think) the default, but longer "afterlife" durations can be selected. The target remains greyed-out. I think the standard for a silent target is the triangle with the slash through it, but I didn't know that at the time.

I designed all this without benefit of the ECDIS guidelines, so no doubt it could be improved.

In NavMonPc I calculate CPA/TCPA for every target, not just those within a particular range. I could probably slightly reduce the CPU load by being more selective, but honestly the CPA/TCPA calculation is pretty simple, and if there are 100 or fewer targets I doubt if the extra computational load is noticeable. I've had thousands of targets on-screen (using a worldwide internet AIS feed) and the CPA/TCPA calculations are a very small percentage of the load. Things do start bogging down when I approach 10,000 targets!

And nobody should think that my comments on using eyeballs in close-quarter situations means that I'm not a huge fan of AIS. It just that when I'm in a situation where a particular tool can give me confusing information, I prefer to avoid the confusion and use a more suitable tool. As much as I love AIS, in close quarters (let's call that one mile or less) where I have good visual ability AIS is not my first choice. I may use it to confirm my estimate of target speed, look for hidden tugboats, etc, but I'm not going to rely on the AIS proximity alarm. If it's foggy then I will use radar and AIS, or better yet, be somewhere else if practical.

We might want to look at how Vesper does their AIS alarm processing (Watchmate). I've not used one, but they are very highly regarded. If someone has a good handle on the close-quarters situation I would really like to know about it!

[barnakiel]

Anybody knows what a safe distance (safe passing distance) is to bridge officers of the bigger ships?

I think they must have some raw open water / inshore rules that could help a small craft operator decide if they are seen and avoided.

b.

[MarkJ]

Originally Posted by nohal

Mark...
Are we clear now?

Pavel

I hope so.

Originally Posted by barnakiel

Anybody knows what a safe distance (safe passing distance) is to bridge officers of the bigger ships?

I think they must have some raw open water / inshore rules that could help a small craft operator decide if they are seen and avoided.

b.

Barny,

Theres no rules been or being implemented AFAIK. Companies set their own rules. Most require a minimum CPA of 1 nm in open water, fewer have a minimum CPA of 2 nms.
Many ships go around me at exactly 1 nm.

Mark

[barnakiel]

Originally Posted by MarkJ

I hope so.






Barny,

Theres no rules been or being implemented AFAIK. Companies set their own rules. Most require a minimum CPA of 1 nm in open water, fewer have a minimum CPA of 2 nms.
Many ships go around me at exactly 1 nm.

Mark

Mark,

This is exactly the info I was looking for. Also in 100% sync with our observations in open waters.

We could then safely assume that, in open waters, if one is going for us @ less than 1 Nm CPA and is say X miles from us, they cannot see us.

Now the X could be assumed less than 4 miles when we carry a class B and 2 miles otherwise, in good visibility.

I am not sure if this algorithm is of any use to the rarefied mind of a coder BUT sure it makes a fine decision making mainframe to a half awake skipper at helm of their 'invisible' sailboat!

Cheers,
b.

[rgleason]

A very large destroyer being towed out of Newport ri to be broken up, skipper wanted us to pass 1 nm ahead or pass astern. Did not have ais and it was a close call on passing 1 nm ahead so we went astern. Tugs towing at night call if you are within 1 nm and appear happen have a close cpa. Had this happen once. Was going to get skipper up to tack and the call got him up in an instant.
So it appears to me that 1 nm is recognized as a kind of safety bubble.


Sent from my SCH-I545 using Cruisers Sailing Forum mobile app

[bdbcat]

Folks...

A lot of good discussion about navigational best practice.

But, the questions for O remain undecided.

Here is a proposal:

1. Calculate first order estimated CPA based on GPS-GPS, as a rough estimate, ignoring target size. That's what we do now.

2. If estimated CPA > 1.0 NM, follow current O logic for alarms, etc.

3. If estimated CPA < 1.0NM, and target dimensions are available, then recalculate CPA twice; once assuming the target's GPS receiver is right on its extreme bow, and once on its extreme stern. The two CPA calculations will use the target's stated dimensions, and reported heading.

4. Of the (now) 3 estimated CPA's, choose the smallest for further alarm processing and reporting in O.

This is a conservative approach, and will not involve excessive computational cost. The object here is to remind one to stay far away from large targets, if possible.

Of course, one is free to configure O's alarm parameters as desired. Nothing new there.

How about it?

Dave

[rgleason]

3. If estimated CPA < 1.0NM, and target dimensions are available, then recalculate CPA twice; once assuming the target's GPS receiver is right on its extreme bow, and once on its extreme stern. The two CPA calculations will use the target's stated dimensions, and reported heading.

I like that approach, in fact I was going to suggest doubling the size of both my boat and target boat, to accommodate the fact that we do not know if they entered the GPS location in the ship properly. - Side to side considerations? Width of ship can make a difference.

Earlier I was not descriptive enough above about NavMonPC GPS location relative to ship bow, stern & sides. NavMonPc allows entry of four separate distances from bow, 2 sides, stern of ship to the GPS location to position the GPS. Where does Opencpn AIS do this? - Forgive me I have not configured an AIS yet, but I have not found it in the Options > Ships > Own Ship or AIS parameters.

For me.
1= good
2= good (adjustible or not?) - Is this a new Close Proximity Calculation parameter?
3= good (Width of ship?)
4= good

So thats my vote.
FYI I found Polar Navy claims more accurate AIS calcs at close proximity.
I will try to find out more about Vesper when I get a chance.

[nohal]

Originally Posted by rgleason

Earlier I was not descriptive enough above about NavMonPC GPS location relative to ship bow, stern & sides. NavMonPc allows entry of four separate distances from bow, 2 sides, stern of ship to the GPS location to position the GPS. Where does Opencpn AIS do this? - Forgive me I have not configured an AIS yet, but I have not found it in the Options > Ships > Own Ship or AIS parameters.

Rick...
Setting Options | Official OpenCPN Homepage

Pavel

[rgleason]

I have written vesper with some questions about close proximity calcs.
Also they have a guard ring and they have "filters" that remove all but the problem targets. Don't know what the filters are.
AIS WatchMate 850 - AIS transponder and display

[Stu Jackson]

Originally Posted by rgleason

I like that approach, in fact I was going to suggest doubling the size of both my boat and target boat, to accommodate the fact that we do not know if they entered the GPS location in the ship properly. - Side to side considerations? Width of ship can make a difference.

Why would you do that for your boat? Compared to shipping, our boats are just gnats. How does this help at all? I understand it for checking large ships, but how does that help THEM to understand YOU?

[rgleason]

Nohal, thanks. ...pretty dumb...did not select from drop down

"Ship Icon Type" "real scale bitmap/vector"

Very sorry.

Stu. maybe we know we set GPS properly so no need for double our boat width and length.

[barnakiel]

@ bdbcat

I would consider, as a brainstorming option, the following order of calculations:

1. check if gps offset data available,
2. if positive, calc CPA, assume worst case scenario, (deduct the bigger offset from the calculated CPA),
3. if negative, calc CPA, assume object diameter X (X to be decided, possibly zero as default)

I hope I am not muddling clear waters.

b.

[Paul Elliott]

Originally Posted by rgleason

Earlier I was not descriptive enough above about NavMonPC GPS location relative to ship bow, stern & sides. NavMonPc allows entry of four separate distances from bow, 2 sides, stern of ship to the GPS location to position the GPS.

I should clarify this. NavMon's GPS location entry is *only* used to draw your vessel dimensions on the AIS screen. It does not use this information, nor the target vessel's size / GPS location information when calculating CPA. CPA calculations use the GPS lat/lon data, nothing else.

If you have an AIS transponder you will need to program that with your vessel dimensions / GPS location. NavMonPc does not send any information to your transponder.

By the way, I use the four distances used to locate the GPS and vessel dimensions because that's the way it's specified in the AIS "Static and Voyage Related Data" message.

I'm not claiming this is best practice, but that's the way NavMonPc works.

A very conservative way to figure CPA would be to (mathematically) construct a circle with a radios of the target's length, and center that on the GPS lat/lon position, then calculate CPA to the edge of the circle. In practice you would just need to calculate CPA from GPS positions and then subtract vessel length from that -- same result as the circle calculation. All bets are off if the vessel length is incorrectly programmed, but then we're no worse off than before.

But in my opinion, this doesn't help much in the multiple-ship close-quarters situations. The problem I used to see in San Francisco bay was that the generation of false alarms was so bad that I was compelled to turn off the alarms entirely and revert to eyeball avoidance. Reducing the calculated CPA by assuming that the GPS position / ship dimensions are incorrect will only make the alarming more sensitive and prone to falsing.

This ultra-conservative method might improve AIS alarm performance in the case of simple crossing situations where the ship is not doing a lot of maneuvering.

One thing I did when developing the NavMon AIS capability (and that was how NavMon got started), was to give the program the ability to "ride along" on any of the vessels being received. I would select a ship, and tell NavMon that it was that ship. NavMon would then use that ship's lat/lon/SOG/COG as its own. I could then play with alarm parameters and algorithms with all sorts of live and canned data. I would save AIS data from my own boat, my land-based receiver, and various network sources, so I had a huge variety of test situations.

Another thing I would do is to create a fictitious route for NavMon and insert that in the middle of my pre-recorded AIS datastream.

This is much safer than trying to put my own boat in near-miss situations to see how the alarms work!