Collision Avoidance, Cones of Uncertainty, and Appropriate CPA

[Dockhead]

Originally Posted by Exile

For me, and having to learn so many different technical, previously unknown-to-me boat-related subjects a bit later in life than many, I've found it far more rewarding to acknowledge being wrong (or just not knowing) than trying to appear smarter or more knowledgeable. In the first instance I usually become far better educated; in the latter I just remain . . . well . . . sorta dumb uninformed. But to each his own, it is the internet after all. The upside is that, personalities aside, CF has always been the best resource for me for just about everything pertaining to my boat, as this thread amply illustrates.

Sure!

The fear of appearing to be wrong, is a bigger obstacle to learning, than even being stupid is. We should not fear being wrong -- we should EMBRACE it. Being wrong is glorious -- you're the one getting the most out of the conversation. Admit it joyously. I am wrong on here just about every single day, on one thing or another, and I am infinitely grateful to those who point it out.

[Juho]

Originally Posted by Dockhead

At 5 knots, he can only move 50 meters (!) in the time it takes the ship to pass him by, if the ship is 200 meters long. So he must already be within 100 meters of a line extending along the side of the ship, by the time the bow of the ship passes him, in order to barely get to 50 meters from his quarter as it flashes by.

If one wants to have max perpendicular distance to the (square) bow of the ship, then one should be on the ideal course (with speed = 5 knots) already in when the bow of the ship passes the sailboat. It seems that you can have max 355.3' (108.3m) perpendicular distance to the (square) bow, if your target is to have 180' CPA to the corner of the transom.

A practical procedure, to cut as close to the (square) aft of the ship as possible, while maintaining also maximum distance to the bow, could be as follows.
1) Know your own max SOG in the intended direction.
2) Get the SOG and true course of the ship from AIS.
3) Count your ideal crossing angle, transform that into a true course, and take that course.
4) Use AIS to stay at approximate collision course (keep CPA to the transom corner somewhat smaller than your planned final CPA) with the ship. Lower your speed if you are approaching the ship too fast. If that is not enough, you may run parallel to the ship for a while and then return to the ideal angle. If your CPA gets essentially larger than you planned it to be (even with max speed and ideal course), you may may need to pick another target ship, if next ships are about to come too close.
5) When you get closer to the ship, stop following AIS, keep your ideal course, and adjust your speed according to your gut feeling (watching carefully the wake and other possible problems).
6) When your bow points to the nearest corner of the transom, just keep your max speed and ideal course if the state of the sea allows that (you can't cause a collision any more) until you have fully crossed the shipping lane. If the next ships on the lane have different speed than the first one, you may need to use another ideal angle and turn to the corresponding new ideal course (this is already perfectionism, and usually not really needed).

These rules apply to both 180' CPA crossings and one mile CPA crossings. If the next ships are not too close, and you don't want to cause stress to the ship, and you don't want to risk your own boat, you should pick a big enough target CPA value.

If one wants to run parallel to the ship in order to maintain better control (not trusting the AIS based approaching approach), and also avoid getting too close to the bow, maybe one should aim at such a distance (considerably bigger than the planned CPA) to the route of the ship, that one can run parallel with the ship until the bow of the ship is about to reach the sailboat, and then turn to the ideal course (that will take the sailboat to the planned distance from the corner of the transom). Maybe the AIS devices could calculate and display also the smallest achievable CPA (if course of the vessel is changed) for this purpose.

Originally Posted by Dockhead

AIS gives most precise data, but it treats the ship as a mathematical point, and our devices don't have access to data about location of the GPS antenna.

I think AIS actually has or may have such information available. The specs seem to include that option. One can indicate the position of the GPS antenna, but not the shape of the ship. Square transom can however be derived from the transmitted LOA, beam and position of the GPS antenna. In Marinetraffic some vessels have a shape and GPS antenna location. I don't know if any AIS devices use this (possibly available) information when calculating CPA.

[Dockhead]

Originally Posted by Juho

If one wants to have max perpendicular distance to the (square) bow of the ship, then one should be on the ideal course (with speed = 5 knots) already in when the bow of the ship passes the sailboat. It seems that you can have max 355.3' (108.3m) perpendicular distance to the (square) bow, if your target is to have 180' CPA to the corner of the transom.

A practical procedure, to cut as close to the (square) aft of the ship as possible, while maintaining also maximum distance to the bow, could be as follows.
1) Know your own max SOG in the intended direction.
2) Get the SOG and true course of the ship from AIS.
3) Count your ideal crossing angle, transform that into a true course, and take that course.
4) Use AIS to stay at approximate collision course (keep CPA to the transom corner somewhat smaller than your planned final CPA) with the ship. Lower your speed if you are approaching the ship too fast. If that is not enough, you may run parallel to the ship for a while and then return to the ideal angle. If your CPA gets essentially larger than you planned it to be (even with max speed and ideal course), you may may need to pick another target ship, if next ships are about to come too close.
5) When you get closer to the ship, stop following AIS, keep your ideal course, and adjust your speed according to your gut feeling (watching carefully the wake and other possible problems).
6) When your bow points to the nearest corner of the transom, just keep your max speed and ideal course if the state of the sea allows that (you can't cause a collision any more) until you have fully crossed the shipping lane. If the next ships on the lane have different speed than the first one, you may need to use another ideal angle and turn to the corresponding new ideal course (this is already perfectionism, and usually not really needed).

These rules apply to both 180' CPA crossings and one mile CPA crossings. If the next ships are not too close, and you don't want to cause stress to the ship, and you don't want to risk your own boat, you should pick a big enough target CPA value.

If one wants to run parallel to the ship in order to maintain better control (not trusting the AIS based approaching approach), and also avoid getting too close to the bow, maybe one should aim at such a distance (considerably bigger than the planned CPA) to the route of the ship, that one can run parallel with the ship until the bow of the ship is about to reach the sailboat, and then turn to the ideal course (that will take the sailboat to the planned distance from the corner of the transom). Maybe the AIS devices could calculate and display also the smallest achievable CPA (if course of the vessel is changed) for this purpose.



I think AIS actually has or may have such information available. The specs seem to include that option. One can indicate the position of the GPS antenna, but not the shape of the ship. Square transom can however be derived from the transmitted LOA, beam and position of the GPS antenna. In Marinetraffic some vessels have a shape and GPS antenna location. I don't know if any AIS devices use this (possibly available) information when calculating CPA.

As far as I know, all AIS devices treat all targets as mathematical points and calculate CPA etc. based on the GMDSS antenna location. It is true that ship dimensions referred to the antenna location is transmitted as part of static data. As far as I know, this data is not used to "correct" the CPA, and for sure it's not with Class "B" units. Maybe Class "A" units can do this, but I've never seen any reference to it.

And what would be the point? All of the other uncertainties which we have discussed makes it impossible to calculate a CPA with such accuracy. Anyone who has held course for a while with a very close CPA will know this -- what is a 50 meter CPA (say) at one moment, will be 0, 200, 75, etc. the next moment, and will keep jumping around. So you just can't use AIS to set up such a close pass. As someone said -- a 180 foot CPA is equivalent to a 0 CPA.

At least with our Class "B" units, CPA in a really close pass will always be less than the unit calculates -- to the extent of the amount of ship between us and the GNSS antenna. So as Ping pointed out near the beginning of the thread -- if AIS says he's passing closely ahead of you, his bow might actually be passing close behind or be on an exact collision course. Same thing with passing behind. AIS might say we will clear his stern by a cable -- but there may well be a whole cable of ship behind his antenna location. A cable is almost 200 meters; 600 feet. So 180 foot CPA??

[Juho]

Originally Posted by Dockhead

if AIS says he's passing closely ahead of you, his bow might actually be passing close behind or be on an exact collision course. Same thing with passing behind.

The first use for AIS devices that indicate also the size of the vessel could be to stop me trying to cross in front of the bow of a slow ship with my fast motorboat at night in the case that the bow is actually 300 m longer than I thought .

[Dockhead]

Originally Posted by Juho

The first use for AIS devices that indicate also the size of the vessel could be to stop me trying to cross in front of the bow of a slow ship with my fast motorboat at night in the case that the bow is actually 300 m longer than I thought .

Indeed.

I got a really good lesson on this while sailing fast through the North Sea, sailing through an anchorage full of large tankers.

I thought that a cable ought to be enough CPA passing ahead of an anchored vessel.

Wrong!!!

Good thing I had my eyes open.

We have margins of errors and there is such a thing as a "safe CPA" exactly to deal with all the things we don't or can't know. Like where the AIS's GPS antenna is located, to name just one.

[Paul Elliott]

Originally Posted by Juho

The first use for AIS devices that indicate also the size of the vessel could be to stop me trying to cross in front of the bow of a slow ship with my fast motorboat at night in the case that the bow is actually 300 m longer than I thought .

Both Class A and Class B AIS include the GPS location relative to bow, stern, port beam, and starboard beam. However, this information is contained in the Static Data messages (msg 5 for Class A, msg 24B for Class B), which are only transmitted once every six minutes, and on alternating channels. Sometimes messages are missed due to weak signals or noise. It can take a good while before this GPS location information becomes available. If you still have a single-channel AIS receiver, it can take twice that long.

Also, sometimes this data is incorrectly entered. I don't see obvious errors as often as I used to, but I still wouldn't bet my life on this being correct.

[markpierce]

I always aim well behind ships in near-collision events, but that doesn't prevent me from altering course to pass within a couple hundred of feet behind them when that suits me.

(Always wonder if/why sailboats don't want to adjust sails for changes in course. In my sailboating days, frequently altered course for wind changes.)

[Kenomac]

Originally Posted by Dockhead

Sure!

The fear of appearing to be wrong, is a bigger obstacle to learning, than even being stupid is. We should not fear being wrong -- we should EMBRACE it. Being wrong is glorious -- you're the one getting the most out of the conversation. Admit it joyously. I am wrong on here just about every single day, on one thing or another, and I am infinitely grateful to those who point it out.

In over 540 posts, it seems everyone has forgotten that stopping a boat is also an option when one realizes that crossing a the bow of a larger boat might be too close. Also turning and running a parallel course while the ship passes is also an option, which then allows the smaller boat to pass the stern safely immediately after the ship passes. Even a sailboat under full sails can turn 180 on a dime just by spinning the wheel or moving the tiller.... without even touching the sails... just tack into the wind, change direction.... problem solved.

The dangerous pass doesn't have to be made just because the course has been set.

I'm not implying that dockhead is "wrong," but why haven't these two options been addressed? Instead the thread has taken on a "do or die" tone. Do as the smart people do, or be some kinda idiot and possibly die.

Small boats can stop, they can run a parallel course and they can turn 180 degrees to avoid potential crossing miscalculations the last time I checked.

But then again... I could be wrong.

[Dockhead]

Originally Posted by Kenomac

In over 540 posts, it seems everyone has forgotten that stopping a boat is also an option when one realizes that crossing a the bow of a larger boat might be too close. Also turning and running a parallel course while the ship passes is also an option, which then allows the smaller boat to pass the stern safely immediately after the ship passes.

The dangerous pass doesn't have to be made just because the course has been set.

I'm not implying that dockhead is "wrong," but why haven't these two options been addressed? Instead the thread has taken on a "do or die" tone. Do as the smart people do, or be some kinda idiot and possibly die.

Small boats can stop, they can run a parallel course and they can turn 180 degrees to avoid potential crossing miscalculations, the last time I checked.

But then again... I could be wrong.

Stopping or taking way off is specifically mentioned in the Rules, and in some cases is a great way to deal with a dodgy situation.

Likewise, making a hard turn or a 180.

But in order to know that such a maneuver will be effective, you have to know for sure that you are not passing ahead. With vessels moving at a similar speed to yours, this is pretty straightforward -- you can generally see with bare eyes what needs to be done.

The whole problem, which we've been discussing in this thread, concerns crossings with vessels in open water with a big difference in speed.

A collision course bearing with a vessel on a perpendicular course and making the same speed will be 45 degrees ahead of the beam. You can see the vessel's quarter. If you change course to aim at the quarter, then problem solved.

But it's a whole different thing if you're making 5 knots and he's making 20 -- the case we've been discussing. In that case, on perpendicular courses, he will bear 14 degrees ahead of your beam. Even worse -- if his course is 14 degrees off perpendicular, then you will see him right on your beam. It's very hard to perceive such a small different of his course. Is he passing ahead? Behind? You can't tell. You can't tell from AIS, either, to better than a ship's length and probably several. Maybe you're passing slightly ahead, in which case stopping might put you on a collision course.

That's why it's so dangerous to sail into a crossing with a fast moving vessel with a small CPA. It's why good seamanship demands passing at a safe distance (not measured in feet ), and taking action in good time. That's required by the Rules, too.

[Juho]

Originally Posted by Paul Elliott

Both Class A and Class B AIS include the GPS location relative to bow, stern, port beam, and starboard beam. However, this information is contained in the Static Data messages (msg 5 for Class A, msg 24B for Class B), which are only transmitted once every six minutes, and on alternating channels. Sometimes messages are missed due to weak signals or noise. It can take a good while before this GPS location information becomes available. If you still have a single-channel AIS receiver, it can take twice that long.

Also, sometimes this data is incorrectly entered. I don't see obvious errors as often as I used to, but I still wouldn't bet my life on this being correct.

Yes, incorrect data could be a big risk. Luckily the actual GPS position is probably pretty much always correct, except that it could be outdated (after missing some messages), and that could lead to accidents.

I checked some large ships in the English Channel in Marinetraffic. They all seemed to tell the location of their GPS antenna (in addition to their length and width). I checked also some ships in ports. The location of their GPS antenna made sense in the sense that I could not point out any obvious mistakes, like e.g. their bow being clearly on dry land.

Probably large commercial vessels take care of their broadcasted information better than small recreational vessels. That's good since the big ones are the only ones where the size of the ship and location of the GPS antenna can make a big difference.

At least in the Solent surprisingly many recreational vessels seem to broadcast also the location of their GPS antenna. If they broadcast their length but not the location of the GPS antenna (or if that data seems dubious), maybe one should calculate CPA based on the assumption that the vessel is round, with radius of the length of the vessel. To be safe, I guess also outdated GPS data (few expected messages missing) could be handled the same way, displaying an estimate of the area where that vessel might be right now, and taking that into account also in the CPA calculations.

[Dockhead]

Originally Posted by Juho

. . . To be safe, I guess also outdated GPS data (few expected messages missing) could be handled the same way, displaying an estimate of the area where that vessel might be right now, and taking that into account also in the CPA calculations.

In my opinion, that would be trying to get more precision out of the system than is justified. There are inherent limitations on the accuracy of AIS CPA calculations, which we've discussed.

It might be helpful, on the contrary, to display this imprecision with some kind of colored zone around the vessel showing the estimated cumulative error.


Speaking of missing messages -- Class B sends position updates only every 30 seconds, so it can be hard to see our maneuvers. Many Class B sets are badly installed and/or with bad antennae, so with only 2 watts of transmit power, even these infrequent messages often don't get through.

The Class B specification is severely crippled due to (apparently) a patent dispute, which has now been resolved, so we can now buy greatly improved SOTDMA Class B sets. Anyone else planning to upgrade this winter?

[Juho]

Earlier I claimed that the ideal crossing angle (deviation from 90° towards the direction of the traffic) when crossing a unidirectional shipping lane, and knowing the speed of the first approaching ship (v1) and our own speed (v2), is atan(1/sqrt(k^2-1)), where k = v1/v2.

That works when v1>v2. In the case that v2>v1 we can use a simpler formula asin(k). We do have more freedom here (to use also other angles) since now our fast vessel can get away from any situation if needed. With this formula I could slip between two ships that are only 10 m from each others (aft to bow) with my sailboat (assuming that it is fast, powerful, nimble and stable enough) .

But what if v1=v2? What is the ideal crossing angle then? If someone wants to give me an opinion on that, I'll delay my own answer to that question until tomorrow .

[Juho]

Originally Posted by Dockhead

In my opinion, that would be trying to get more precision out of the system than is justified. There are inherent limitations on the accuracy of AIS CPA calculations, which we've discussed.

It might be helpful, on the contrary, to display this imprecision with some kind of colored zone around the vessel showing the estimated cumulative error.


Speaking of missing messages -- Class B sends position updates only every 30 seconds, so it can be hard to see our maneuvers. Many Class B sets are badly installed and/or with bad antennae, so with only 2 watts of transmit power, even these infrequent messages often don't get through.

The Class B specification is severely crippled due to (apparently) a patent dispute, which has now been resolved, so we can now buy greatly improved SOTDMA Class B sets. Anyone else planning to upgrade this winter?

Your proposed approach of considering the location of a vessel with missing AIS messages to be fuzzy could be used. If some ship has stopped sending AIS locations 10 minutes ago in the middle of the English Channel, the AIS devices of all nearby vessels could give a CPA warning based on that missing information. We would thus err on the side that vessels with missing information could be closer to us than they actually probably are.

[Lodesman]

For fun I drew it out on a manoeuvring board, with scale boats.

The small boat is roughly 50' in length at 5 kts, and the large is a Neopanamax vessel - with standard dimensions for that class of 1200' LOA and 160' BOA at 20 kts. For the 180' CPA given the pictures show: as the big ship passes the bow of Rod's boat, it is only 500 feet away; midways down and still looking at a wall of steel that's closed into 300 feet; and at CPA. I can imagine that would be the longest, most puckering 36 seconds of someone's life.

Not shown, but with the relative trajectory, the small boat will not have cleared the larger's transom until it's about 2 cables behind.

As others have stated, I think the bridge crew on the big vessel would be quite stressed by this. What has not been talked about would be their likely course of action. I don't know what the SOP is in large commercial vessels, but if it's the same as warships, then the reaction would be to make a port turn in order to swing the stern away from the small vessel. This might cause more problems, because large helm orders cause an incipient yaw in the opposite direction - ie. the stern would swing closer before it started to go the other way. I can't stress enough how dangerous crossing this close would be.

[poiu]

There's a real world example of the sort of situation being talked about here in this MAIB report.

www.maib.gov.uk/cms_resources.cfm?file=/nedlloyd%20vespucci%20and%20wahkuna.pdf

There was a collision between a 47ft yacht and a container ship in fog in 2003, so before the use of AIS on small yachts. The yacht was to port of the ship doing 7.5kt 12deg and the ship at 25kt on 255deg. Both boats saw each other on radar 6nm away. The ship thought they were safe estimating initially the yacht would pass 8 cables ahead (the estimate reducing down to two cables prior to impact), but the yacht thought it was on a collision course and slowed down to give a passing of 1.5 miles to the stern of the ship. The radar data was wrong on both boats, particularly the yacht and both boats took the wrong actions. The yacht thought he was stand on and didn't change course as he could have done. He slowed down into the path of the ship and the ship did nothing thinking it was safe to pass, but it wasn't. They collided, the yacht sank and the crew were rescued from a life raft.

The report addresses the question of what is a safe passing distance. It quotes Cockroft and the IMO conference saying that the distance depends on so many things it can't be quantified. It also quotes a report by the Seafarers' International Research Centre on p22 "...a passing distance of 3 cables or less, is on anyone's measure, a very dangerous occurrence that could lead to a collision with only a very slight change in circumstances."

IMO regs require radar accuracy of only 7 cables in a crossing situation at 3 miles. The report implies an allowance needs to be made for radar errors. I'll bet modern equipment is much better than this and certainly the ship's equipment would have been better then - it was found to be 2 cables as they collided with the ship's radar showing 2 cables distance.

Both boats didn't use relative radar settings and the yacht didn't know how to use his MARPA settings. That didn't help and the ship was going too fast in fog.

There is a useful reminder on the importance of Colregs Rule 19 where in situations like this of restricted visibility both boats are obliged to take avoiding action.

The bottom line is people and machines cock things up, so you need to have an appropriate margin of error. Here 2 cables was proved to be completely inadequate.