Collision Avoidance, Cones of Uncertainty, and Appropriate CPA

[Paul Elliott]

Originally Posted by Dockhead

If you are trying to pass as close as two cables from a fast moving, large ship, however, you cannot ignore the antenna placement problem (a Triple E is two cables long). It's frustrating that we don't get that part of the static data, on our Class "B" sets. I think one would need to assume that the antenna is somewhere in the accommodation block, and add the estimated distance from there to the part of the ship you expect to come closest to, to your CPA.

We *do* receive that data from the ship's AIS, regardless of the type of receiver or transponder we have. And Class-B transponders also do send antenna position data. When we see the ship's length on our AIS plot we are actually seeing (GPS distance from bow + distance from stern).

The problem is that the this data is entered by human beings and is occasionally in error. I've seen Class-A dimensions that show the ship wider than it is long -- obviously a data-entry error. I wouldn't trust my life on AIS antenna position information. Another problem is that this static data is only transmitted once every six minutes. Once you do receive it the dimensions aren't going to change, but it can still take an unnervingly long time for this static data to show up.

One more issue is that the ship's heading is sometimes not included in the AIS dynamic message. I don't know why, but perhaps the ship's fluxgate compass is AWOL? I do know that I've seen zeroes or NA data in that field, while the COG and SOG are being properly sent. Obviously bad heading data would completely screw up an automatic aspect-based CPA calculation.

Display of the GPS position will also depend on the plotter or program you are using. There is an IMO standard for target icons, but I don't know if the antenna position is displayed. On NavMonPc (which I wrote pre-standard), the ship outline can optionally be displayed relative to the GPS position. My CPA calculations are done from GPS position, not bow/stern/beam. That could be done, but I wouldn't trust the results (false sense of security?)

[evm1024]

Rod,

I think you missed my question.

Rod, what do you think about opening up your proposed CPA to 200' does it make it better or worse than 180'?

Can you take a moment to answer if 200' would make a better or worse crossing?

[ramblinrod]

Originally Posted by evm1024

As for the one proposed crossing where I set up for a 2 cable CPA to the oncoming ship - I would not give it a B. At best it is a D. The crossing situation as a whole sucks - but you know that....

Since you proposed a solution for the scenario, (brave) I am responding.

You give your 180' CPA solution an A+, all the while ignoring the real problems in actually getting that close to the 4th or 5th ship in line.

It doesn't have any real bearing on the outcome, and I know why you are suggesting I would be crossing in front of the 5th ship, (despite some erroneous claims by others, I do have imagination and can envision things) but that is not necessarily the case. The first ship of the convoy could be the ship my course would cross behind.

And ignore the very real problem of that ship taking collision avoidance measures and throwing your crossing into disarray - can you say SNAFU nay make that TARFU.

Regardless, what solution was planned, this is a real possibility. Whether planning to be 180 ft, 600 ft, or 1800 ft aft of the lead ship, if the ships make collision avoidance maneuvers the plans are out the window. Again, in all probability, if the ships were going to take evasive measures, they would have done it long before I got to within 180 ft of the port stern corned of the lead ship. Chances are, if they hadn't seen me yet, and did at that last moment, the whole scenario would play out, before their actions could even be effected. It is possible that someone could have seen me earlier, and had planned a maneuver but hadn't executed yet, but if so, I would have heard the 5 short blasts, and known to abort before even getting near.

You fail to realize that the goal in this cross is not to get as close to the leading ship as possible so as to keep as far away from the trailing ship as you can.

Indeed the goal is to get as far away from the leading ship as possible while maintaining a safe distance from the trailing ship. It is a balance where the risks of the leading ship and in balance with the risks of the trailing ship..

I believe the primary goal of any crossing, is to get across without being whacked. I believe that for this scenario, my solution stands the better chance, but yours may work.

I should point out that in the 600' CPA I did the math on we end up being about 1 nm ahead of the trailing ship when crossing the ships centerline.

Yes, but considerably less from the following ship starboard course line. You also end up being much greater than 600' from the lead ships starboard course line, so in my opinion you gave up too much room. Your opinion may vary, that's fine.

It is plainly clear that you have not been 180' away from a 600' long ship doing 20 kts. No need to answer that question. Your silence speaks volumes.

I simply choose not to answer silly questions.

Rod, what do you think about opening up your proposed CPA to 200' does it make it better or worse than 180'?

In reality, I would not have range determination nor course keeping ability to that level of accuracy. By planning 180 ft, I am most likely going to be somewhere +/- 50 ft by eyeball.

Similarly, at 600 ft, your deviation is likely no closer than +/- 150 ft.

[evm1024]

4th or 5th ship is not counting from the head of the line. It is counting from the ship in front of you when you are 1 nm from the path of the ships. It matters not where you are in relations to the start or end of the line.

If you try to cross with a CPA of 180' even if you could judge that +-1' the results would be the same. The ship that you ended up passing astern of would have plotted you on a collision course from more than 1 nm out. They would have been watching you at 4 or 5 miles out. The question is how close do they let you get before they take evasive action. 180' - Laughable. Do they let you get within 0.25 nm before action? hard to say.

What distance do you think that they will take action?

As a guess I might say that at 0.5 nm out they will take action. 15 degree turn to stb by the 4th and 5th ship. (turn to port, see you in court). And you are given some time to correct your error in predictability.

But again will you please answer would opening your target CPA up to 200' improve or worsen your crossing? Please do take a moment to answer.

[Dockhead]

Originally Posted by Paul Elliott

We *do* receive that data from the ship's AIS, regardless of the type of receiver or transponder we have. And Class-B transponders also do send antenna position data. When we see the ship's length on our AIS plot we are actually seeing (GPS distance from bow + distance from stern).

The problem is that the this data is entered by human beings and is occasionally in error. I've seen Class-A dimensions that show the ship wider than it is long -- obviously a data-entry error. I wouldn't trust my life on AIS antenna position information. Another problem is that this static data is only transmitted once every six minutes. Once you do receive it the dimensions aren't going to change, but it can still take an unnervingly long time for this static data to show up.

One more issue is that the ship's heading is sometimes not included in the AIS dynamic message. I don't know why, but perhaps the ship's fluxgate compass is AWOL? I do know that I've seen zeroes or NA data in that field, while the COG and SOG are being properly sent. Obviously bad heading data would completely screw up an automatic aspect-based CPA calculation.

Display of the GPS position will also depend on the plotter or program you are using. There is an IMO standard for target icons, but I don't know if the antenna position is displayed. On NavMonPc (which I wrote pre-standard), the ship outline can optionally be displayed relative to the GPS position. My CPA calculations are done from GPS position, not bow/stern/beam. That could be done, but I wouldn't trust the results (false sense of security?)

Thanks for this -- really interesting stuff which is new to me.

So let me see if I understand this correctly -- the static data is used to form the ICON? NOT to calculate CPA? Am I understanding that correctly? So CPA is still done to the antenna as if the ship were a mathematical point, right?

As to "My CPA calculations are done from GPS position, not bow/stern/beam. That could be done, but I wouldn't trust the results (false sense of security?)" -- amen! False precision! Do all AIS sets work this way?

I display AIS data in two ways -- on my main nav system, a pair of B&G Zeuses, and at the nav table via OpenCPN.

The Zeuses don't have different icons for different vessels, or even different sizes of icons -- just a carat. O however has an option for "Show AIS Targets Real Size"

[evm1024]

Folks, you can take a look at this crossing with ease. The difference in speed gives a relative course line of 14 degrees ( arc-tan(5/20) ).

So if you draw a line of ships on paper to scale (600' long, 1.25 nm apart) - you really only need 2 ships. With the line of ships top to bottem of your paper.

Then make an angle of 14 degrees to the line of ships.

Your sailboat will be along the angle line. Set the upper end of your angle at any desired place relative to the lead (upper) ship and you will see where your sailboat will be in relation to the ships at any given time in the crossing. It will show the CPA to the following ship. CPA is measured at a right angle to the 14 degree line.

Wow, harder to say it than it is to draw.

[Paul Elliott]

Originally Posted by Dockhead

Thanks for this -- really interesting stuff which is new to me.

So let me see if I understand this correctly -- the static data is used to form the ICON? NOT to calculate CPA? Am I understanding that correctly? So CPA is still done to the antenna as if the ship were a mathematical point, right?

As to "My CPA calculations are done from GPS position, not bow/stern/beam. That could be done, but I wouldn't trust the results (false sense of security?)" -- amen! False precision! Do all AIS sets work this way?

I display AIS data in two ways -- on my main nav system, a pair of B&G Zeuses, and at the nav table via OpenCPN.

The Zeuses don't have different icons for different vessels, or even different sizes of icons -- just a carat. O however has an option for "Show AIS Targets Real Size"

I can only describe how CPA is calculated by NavMonPc. OpenCPN or commercial standards-compliant plotters may do it differently. I wrote NavMonPc before the display standards had been finalized.

In NavMonPc CPA and TCPA are calculated based on the AIS target's COG, SOG, and GPS position. I use the same data from your own boat for these calculations. GPS antenna locations (yours, and the ship's) and the vessel outlines are shown graphically on the display if this option is enabled, but they are not used in the calculations. I display the ship's LOA on the target detail window, and this comes from the GPS location data. I show it because it's interesting, and might be useful.

[Dockhead]

Originally Posted by Paul Elliott

I can only describe how CPA is calculated by NavMonPc. OpenCPN or commercial standards-compliant plotters may do it differently. I wrote NavMonPc before the display standards had been finalized.

In NavMonPc CPA and TCPA are calculated based on the AIS target's COG, SOG, and GPS position. I use the same data from your own boat for these calculations. GPS antenna locations (yours, and the ship's) and the vessel outlines are shown graphically on the display if this option is enabled, but they are not used in the calculations. I display the ship's LOA on the target detail window, and this comes from the GPS location data. I show it because it's interesting, and might be useful.

That sounds great, actually! It would be nice to have that data so easily at hand.

I asked that question about whether antenna placement is considered in calculating CPA of a couple of experienced commercial guys last summer. They both replied that they didn't know, and that it didn't matter, because you should never cut it so close that you care about where the antenna is.

I have noticed it sailing close to anchored big ships, where it was obvious that the bow was closer than the displayed CPA would lead you to believe. I don't know whether this is from lack of consideration of antenna placement data, or badly entered data. In any case, I guess it doesn't matter. It's to your point that thinking about it would be false precision.

[Dockhead]

Originally Posted by evm1024

Folks, you can take a look at this crossing with ease. The difference in speed gives a relative course line of 14 degrees ( arc-tan(5/20) ).

So if you draw a line of ships on paper to scale (600' long, 1.25 nm apart) - you really only need 2 ships. With the line of ships top to bottem of your paper.

Then make an angle of 14 degrees to the line of ships.

Your sailboat will be along the angle line. Set the upper end of your angle at any desired place relative to the lead (upper) ship and you will see where your sailboat will be in relation to the ships at any given time in the crossing. It will show the CPA to the following ship. CPA is measured at a right angle to the 14 degree line.

Wow, harder to say it than it is to draw.

That will work!

That's how you plot relative motion, and the line you are talking about, the "14 degree line", is called the relative MOTION line (motion because speed is also part of it, not just course). It's an analogue technique and works. It's easier to interpret if you do it on a plotting sheet or maneuvering board.

You can also do this with math, but it needs to be done right -- you will have to solve 6 right triangles. There have been some calculations on here by home-made methods which do not give the correct result.

Note that you can use a single line with multiple targets only if they are moving at the same course and speed.

[transmitterdan]

You can never be sure of the AIS coordinate relative to the extents of the ship itself. Some reasons have already been given. The time lapse between "now" and the time of the last "known" position, COG and speed creates an ever increasing uncertainty cone for each vessel.

None of the AIS display units I know about move the position of AIS equipped ships along their projected course between position updates. Maybe some high end ECDIS can do that. But the uncertainty in that projected position would be pretty large as we approach the time of the next update from each ship. It seems like a good topic to understand and create rules of thumb that could be used to estimate the uncertainty zone. Sort of like the hurricane path cones many people misunderstand.

It would perhaps be a good feature of AIS displays if they could somehow describe this zone of uncertainty that changes over time. Do any AIS displays try to do that? All the ones I have seen just try to guess the place where the GPS antennas will reach the point of CPA. But they generally imply a precision of that point that is not really supported by the underlying data.

AIS is a good lesson in the uncertainty principle which tells us we cannot know precise position and velocity simultaneously. AIS has fairly large uncertainty that changes with time. So that has to be understood before we can effectively use it to plan. I have always believed AIS allows better planning than eyes alone since it can see over the horizon. AIS can really improve the knowledge about any situation. But the data isn't accurate enough to let us get really close to another vessel.

[Stu Jackson]

Originally Posted by StuM

All these hypotheticals about CPAs to two ships in line in a totally unrealistic scenario are really a waste of time.

The bottom line is that anyone who has actually been anywhere close to a large ship traveling at speed is well aware that attempting to get anywhere within 50 metres of one in a slow moving small craft would be madness. And certainly not prudent seamanship!

I have been fascinated by this discussion and have been following it daily.

So, I gave Rod's "concept" a real life test. One thing that I have been meaning to say during this dialog, and that others now have, is that there have been two “trains” [sic] of thought on this scenario:

1) First is the 180 foot...pass astern close to the first vessel. Period.

2) Second are the second and following ships, which affect the amount of time to cross behind the first ship before the second one "gets there" --- which is mostly, I believe, Rod's entire argument for "needing" to cross close behind #1. But not an issue in this exercise.

For the moment, please, folks, indulge me and please forget about the second ship(s). OK? Thanks.

So, I tried Rod's "method" about two hours ago (yesterday afternoon actually). I was heading north on the west side of Swanson Channel in the Gulf Islands. A BIG ferry that goes across the Strait of Georgia to the mainland was making its way out of the terminal at Swartz Bay, heading north, too.

I have made this ferry trip many times (on the ferry!) and I am comfortable saying that at this point in his trip he was most likely going 15 knots. In the grand scheme of things, less than the 20 knot scenario, but still going three times MY speed.

Beaver Point was to my port, at the narrowest point between Saltspring Island to port and North Pender Island to starboard of both me and THE ferry. Only one, now, please remember. Thanks. That distance on the chart is 1.8 nm. I was quite a lot east of Saltspring, perhaps a third of the way across the "neck." Let's assume for this exercise that the big ferry did the same on his side. Therefore, dividing 1.8 nm by three = 0.6 nm away from each other at the "neck." Having sailed on SF Bay for 40 years, I got pretty good at judging distances for ships of this size, and I have been here actively cruising these Gulf Islands for one summer season, dodging ferries left and right. I'd guessed about a half a mile before I checked on the chart. I also raced for a number of years back there, and am quite familiar with turning marks and sailboat speeds at those marks.

I know the ferries size, speed and routes, which make it easier, but, again, bear with me - and use that assumption again that the ferry is on rails. This exercise is one of a 5 knot sailboat crossing a 15-20 ship. Period.

I also recognize that this is NOT in open sea, he had a route/channel to take, etc. But the ENTIRE purpose of this exercise was to see what happens - in the real world.

I saw him coming at between 5.5 and 6 nm as he rounded Portland Island.

So the “0” CPA would be a line between Beaver and Mouat Points if I was aiming for anything along his length and "trying to tuck in behind him."

He had 5 miles to get to that “0” CPA at 15 knots or 20 minutes. I had essentially 0.5 nm to go at 5 knots or 6 minutes [(60 * 0.5 nm) / 5 knots]

I didn’t start “aiming” at his port quarter until he was about 2½ miles away, or about half the distance from when I first saw him and got this idea. So it should take him 10 minutes at 15 knots to get to the “0” CPA point, and I had 6 minutes to get there.

Well, whaddaya know: He was LONG GONE by the time I got anywhere close!!!

Why? Any number of possible reasons. I was a bit chicken – I don’t have Rod’s pucker factor! I may have been going less than 5 knots when I first turned to do this exercise, and then sped up to 5 knots. The ferry was going faster than I thought – I checked AIS this morning and by the time those guys get around Portland Island and pass Moresby they are doing 20 knots already. Really. That would have made his Time from 2½ miles decrease to 7½ minutes from 10 minutes. I may have been a bit further away than just a ½ nm. He may not have been on a straight course, and could have been bending to starboard as he rounded N. Pender Island at Mouat Point. Maybe he saw me coming and turned even a little bit to starboard.

Point being that even if you tried to do what Rod has been proposing, and as others have mentioned, only a brain dead conning crew on a ship would let you even get there.


But even if they did, it is really hard to set this up when you're going 5 knots and they are doing 20. Even 15!


Again, forget the whole second train of ships. It’s a smokescreen, or strawman. It is NOT pertinent to a 5 knot sailboat crossing a ship doing 20 knots. Ain’t gonna happen.


I, too, am surprised by Rod's lack of answer as to whether or not he has ever tried this, dead conning crew on the ship or not!


As many of you know, I have a great deal of respect for BOTH Dockhead (his CTS exercise is a textbook example of taking his knowledge and sharing it, as is this thread) AND Rod (his electrical knowledge and sharing of it is superb).

In this case, developing a theory of a train of ships deflected, I believe, from the whole 180 feet passing distance. And the work done by those who provided the plotting tables and graphs is truly appreciated.

What I ALSO learned is that the ferries here, the big ones, truck along very quickly and much faster than I originally thought. Contrary to the slower speed of the shipping in the channels I was used to in San Francisco (and let’s NOT get into channels, TSS, harbors, etc., please) it increased my awareness of the importance of speed, time & distance and safe passing. I also realized I am not going to ever try this again, either!!!


Thanks again so much for this entire thread. Extremely useful, helpful and important for safety.

[StuM]

Originally Posted by ramblinrod

I simply choose not to answer silly questions.

In reality, I would not have range determination nor course keeping ability to that level of accuracy. By planning 180 ft, I am most likely going to be somewhere +/- 50 ft by eyeball.

So now we're down to accepting a potential 130ft (40 metre) separation.

It has become blatantly obvious that you have no experience of what you speak.

[Paul Elliott]

Originally Posted by transmitterdan

None of the AIS display units I know about move the position of AIS equipped ships along their projected course between position updates. Maybe some high end ECDIS can do that. But the uncertainty in that projected position would be pretty large as we approach the time of the next update from each ship. It seems like a good topic to understand and create rules of thumb that could be used to estimate the uncertainty zone. Sort of like the hurricane path cones many people misunderstand.

As an option, NavMonPc does move AIS targets along their ded-reckoned path. I call it "extrapolated" position, and it merely uses position, SOG, and COG to advance the icon. I ignore rate-of-turn.

Obviously the longer the interval between AIS updates, the less reliable the DR position might be. However, nearby ships generally send reports often enough that this doesn't become a big issue. I do display the time since update on the target info window so you can decide how much to trust the displayed position.

[Juho]

Originally Posted by Dockhead

By "min allowed distance to approaching ship" -- you mean the decision point, right? The last moment to leave a collision course before executing a collision avoidance maneuver? If I'm understanding that correctly, then I think that's about right. That's near the end of the normal stand-on, give-way maneuvering in open sea without complicating factors (like other ships).

From the drawing point of view the "min allowed distance to approaching ship" or some similar parameter must be there to limit the size of the danger zone from the front. The idea is that when the ship is pointing at you close enough, or there is a risk that it will point at you, you want to get away. That parameter determines the point where the expanding curved section of the danger zone turns into the narrowing straight segment of the danger zone (3 NM in the examples).

Other approaches (alternative parameters) for determining this point could be "orthogonal distance from the estimated track" or "angle of the approaching ship in relation to the boat when the ship is clearly visible". The latter of those is quite vague, but boats without AIS and other modern equipment might use also such measures.

The danger zone can be seen as the area where you want to act right away to avoid risk of collision. That could correspond to the colregs end of "stand-on give-way area". If we want a close match, the parameters of the danger zone should be trimmed closer to "must turn" rather than "might already turn". The stand-on/give-way area could be another zone around the innermost area (maybe determined by some appropriate time to perform the required moves and some other parameters).

Note that when making the decision to move when fishing on the shipping lane, the decision should not be delayed until one feels "too close to the approaching ship" (e.g. 3 NM) but when there is a risk that this feeling may come later unless one moves now.

I find it interesting that you modeled possible turns by the ship. A really important reason for taking early and positive action is to give yourself room to respond in case of an expected change of course of the ship.

The best easy approach that I could find was to use two parameters, "max error in the observed course of the ship" and "min distance at which the ship might turn 90°". The first one determines the angle of the long curved line when it leaves the side of the ship. It could include a quick surprising turn of the ship. The second one describes the uncertainty of ship movements in longer scale. It determines the radius of the long curved segment. It might be possible to estimate the closest point where the boat could turn 90° e.g. based on observations of the AIS history of the boat. Variation of the speed estimates was not used in these calculations (the user was supposed to take it into account when estimating the turning distance).

But this parameter will be so different from the others, that I think it would be useful to model them separately. I think the first question we want to answer is "How much CPA do I need in a given situation to avoid him under any combination of the normal variations and uncertainties about his and my course, speed, position, etc."

Yes, one could say that it is all about estimating the closest acceptable distance. Also the danger zone could be seen as an extension of the CPA area around the ship. It just extends much more in the direction of the bow of the ship than in the direction of the aft and sides. Allowed CPA can thus be said to have different values in different directions. The boat needs to adjust its speed and course in order to avoid the danger zone.

The danger zone could be divided also in two parts. The part close to the ship (3 NM in the example) is the area where you don't want to be in, and the rest is area where you are still ok but having too high risk of ending up inside the first area.

(Note that in the drawings in about 25° direction the danger zone consists of two parts, one next to the ship and one further away. "Allowed CPA" in this direction is thus not just one value but a richer description of allowed distances. I'm not sure if this is the most natural approach psychologically or when estimating the real life probabilities of ship movements. But so far it seemed at least to carry some sensible message of possible unexpected changes in the course and track of the ship.)

All this data on acceptable distances, intentions of the other vessel etc, and their relation to colregs makes me think if it would be possible to develop a semi-automated system that would coordinate the understanding of the two vessels on when one should act, what the plans of the other vessel are etc. It should be really natural to be of any help (i.e. not just new gadgets and new confusing information). That could be a good portion of food for thoughts for some time .

Are you based in Helsinki? I'll be in Helsinki next week on business. I would be happy to sit around a table with you with maneuvering boards and bottle of Koskenkorva, if this sounds like fun and you can spare the time.

That would be fun. But right now I'm at the countryside, and next week we will be lifting our boat out of the water and preparing it for the winter in the Turku region, so next week may be too difficult, but maybe next time.

[Juho]

Originally Posted by Stu Jackson

So, I tried Rod's "method" about two hours ago (yesterday afternoon actually). I was heading north on the west side of Swanson Channel in the Gulf Islands. A BIG ferry that goes across the Strait of Georgia to the mainland was making its way out of the terminal at Swartz Bay, heading north, too.

That was an interesting and enlightening story from real life. It is not easy to keep up with those fast vessels.

I didn’t start “aiming” at his port quarter until he was about 2½ miles away, or about half the distance from when I first saw him and got this idea. So it should take him 10 minutes at 15 knots to get to the “0” CPA point, and I had 6 minutes to get there.

Well, whaddaya know: He was LONG GONE by the time I got anywhere close!!!

This brings up one important point that has not been discussed much yet. Colregs say that one should make clear moves in such a way that the other vessel understands our intentions. You did right when you didn't cause too much confusion to the ferry.

If our intention is to cross the track of the other vessel from behind the vessel, it is obvious that the polite and correct way to do it is to change the direction of our vessel so that it points behind the other vessel, not in front of it. At least that is what I regularly do when I meet with other vessels, and we must decide and indicate to each others how we plan to cross. Turning my bow so that it points to the other side of the other vessel would carry a false message of my intentions.

If we plan to cross the track of the fast vessel as close to its aft as possible, one may have to take a 90° course against its route (in front of its bow). Or actually the most efficient angle (to cross the track as close to the transom as possible) is even slightly more towards the direction where the other vessel is going.

And since out plan is to cross right next to the transom of the other vessel, our course is about the same that we would use when trying to cross in front of its bow. Such a course would be really confusing to the other vessel. In such a situation it would make sense to the other vessel to react, and maybe turn or speed up or slow down. That would be quite bad behaviour on out side.

Maybe one should take also this kind of obvious colregs clarity requirements into account when counting what kind of manoeuvres are possible and what kind of actions we must avoid. Keeping one's nose on the correct side with respect to the other vessel could be one such requirement. At least we should not keep the other vessel uncertain of which side of it we plan to cross (VHF might help if we really want or have to make this kind of moves).