AIS crossing Shipping Lanes

[dunedin]

Agree that software is catching up fast, not that surprising given the relative newness of AIS to the mass leisure market.

That said, the fact that a gaggle of big US/Japanese tech company cannot sort a "major but easily addressed omission" contains a clue to a different explanation. For the aforementioned reasons it turns out that this is an explicit choice on their behalf made for both legal and safety reasons. Raymarine has recently come up with something much more useful and significantly more complex solution in that it attempts to depict the implicit error terms around the vector crossover points. Others will surely follow (perhaps they have?) in implementing more secure solutions than the deceptive attraction of a simple bearing at CPA angle.

Sorry, but I can't see why it is "legally and safety" wise OK to give a CPA distance, but not to give a CPA bearing.
Nor can I see it is difficult mathematically or processor wise to calculate the bearing at CPA, as this is part of finding the lowest distance apart.
Still seems an odd omission to me.

 

[dunedin]

It doesn't! In fact you'll love the new AIS functionality with the only caveat that you may have to update the software to Lighthouse II version and this may in turn stress the CPU on an E Series. The Raymarine tech guys in Fareham and will I'm sure be able and happy to advise.

Sadly Raymarine haven't been willing to port Lighthouse software for the E series, only the newer (lowercase) e and now eS series (demonstrating their commitment to ensuring rapid obsolescence of expensive electronics). Hence my list about the confusing differences in AIS vector for example.

 

[lw395]

TBH I've always got by just using manual methods on anything that has a CPA of less than 0.5nm or so.
Is the bearing to target rising or falling?
But then I tend to cross the channel from the Solent. It's not that hectic.

 

[Roberto]

In the absence of a built-in system indicating whether a crossing will be ahead or astern, it helps to get back to basics: monitor the change of bearing indicated in the target info, just as you'd do with a hand-bearing compass. The bearing's trend will tell you how you'll cross and the only algorithm required is already in your head. (If it isn't, perhaps you shouldn't be at sea)

Granted there might be a bit of overload on a busy day in The Solent, but for crossing shipping lanes, it's fine.

(Apologies if this was already suggested: I lost interest in the squabble.)

Hi Mac,

+1, that is what I do, the reason being my gps dongle sends unsmoothed cog/sog data to opencpn with about 1Hz frequency, in close quarters this means all sort of vectors jumps with cpa all over the place; by writing down the bearings of any targets there is no doubt whatsoever as to what will happen.

All the best, r.

 

[Lon nan Gruagach]

OK ..... I'll take the bait Dougal.

I'm struggling to make the connection between AIS and the paddlewheel.

Richard

There is not, he just wants to sound like an expert but is not.

https://help.marinetraffic.com/hc/en-us/articles/205426887-What-kind-of-information-is-AIS-transmitted-

SOG is transmitted. OK, so not speed through water and not from paddle wheel. But if it reports that wrong then your receiver will have its speed and therefore future position wrong.

In addition his claim that because radar sees the other it becomes obvious whether the other ship is passing ahead or behind is just as fanciful.

Well I guess not everyone understands vectors and relative velocities.
Its how the radar displays the information. so when I say

With radar your boat is the reference and other vessels tracks are displayed relative to you

You would probably miss "displayed relative to you", but hey ho. The radar displays you as a stationary dot with all else moving past you (their course and speed minus your course and speed). The result being which end of you it will pass. OK some more difficult calculation, but simply if it's radar track is up the screen (normal orientation is you are pointing up) you pass behind it.

 

[jonjo]

Its how the radar displays the information. so when I sayYou would probably miss "displayed relative to you", but hey ho. The radar displays you as a stationary dot with all else moving past you (their course and speed minus your course and speed). The result being which end of you it will pass. OK some more difficult calculation, but simply if it's radar track is up the screen (normal orientation is you are pointing up) you pass behind it.

Right so for Radar to outperform AIS we start with an initial contact splodge placed on an electronic chart image deduced from a fat radar beam ping. The bearing is itself dubious because it is meaningless on a pitching yawing yacht unless corrected by a gyro dampened heading sensor. We then expect the electronics to lay down multiple instances of these inaccurate splodge contacts and then over time when enough data is available guess some probable heading and speed.

When CPA is 0.1NM and we are trying to judge ahead or behind a ship, you would place your faith in vectors derived from splodge extrapolation estimation rather than the instant cartesian maths and trigonometry of AIS.

I recall the early days of AIS on this forum when I was a near lone voice expressing enthusiasm about how AIS was going to change small boat offshore sailing. For each person who got it back then there were 20 cynical radar loving luddites. It is quite a shock to encounter someone like you 12 years later.

 

[Lon nan Gruagach]

Right so for Radar to outperform AIS we start with an initial contact splodge placed on an electronic chart image deduced from a fat radar beam ping. The bearing is itself dubious because it is meaningless on a pitching yawing yacht unless corrected by a gyro dampened heading sensor. We then expect the electronics to lay down multiple instances of these inaccurate splodge contacts and then over time when enough data is available guess some probable heading and speed.

When CPA is 0.1NM and we are trying to judge ahead or behind a ship, you would place your faith in vectors derived from splodge extrapolation estimation rather than the instant cartesian maths and trigonometry of AIS.

I recall the early days of AIS on this forum when I was a near lone voice expressing enthusiasm about how AIS was going to change small boat offshore sailing. For each person who got it back then there were 20 cynical radar loving luddites. It is quite a shock to encounter someone like you 12 years later.

Ahh, so its religious prejudice then?

Is there anything to make you think that radar processing has stuck in your 12 year old stone age?
Lets make a bit of physics, maths and new fangled electronics a bit clearer.
The effect of pitch and roll on a successful ping are an effectively tiny error on the range of a successful ping.
Blob averaging is stunningly accurate, in fact it is a method used to improve accuracy of the majority of measuring systems. Search "least significant digit dithering" for more details.
Gyro dampened heading sensor, yeah, sounds really hi-tech, space ship stuff, good in a put down post. That is until: http://www.digikey.com/product-detail/en/stmicroelectronics/L3GD20TR/497-12081-1-ND/2793125 OHH, $4.29 can we afford it?

 

[dom]

Sorry, but I can't see why it is "legally and safety" wise OK to give a CPA distance, but not to give a CPA bearing.
Nor can I see it is difficult mathematically or processor wise to calculate the bearing at CPA, as this is part of finding the lowest distance apart.
Still seems an odd omission to me.

As you say the calc is not difficult mathematically or processor wise, so it is surely safe to assume that the boys and girls at FLIR, Garmin, Furuno et al are able to do it. So why would they on occasion choose (i.e. not omit) to present a bearing at CPA for legal and safety reasons?

The answer has to do with the inherent error terms. If an AIS set informs us that we have a CPA with a ship of 5miles we will happily ignore it. If however the CPA falls to 0.2m we pull or colreg hats on and start thinking about what to do next. At this stage sailors naturally want a bearing at CPA to know whether they will be ahead of or behind the ship. Their logic is simple: a combination of a CPA and a bearing defines the precise juxtaposition of the two vessels at the TCPA.

This precise positional freeze is however no more than the highest density overlap point of two probability distributions (ship and yacht). All other possible positions are valid, just with a dwindling likelihood as one departs from the maximum density hotspot. If the CPA is five miles the likelihood of a collision will be tiny, but a CPA of 0.2 miles opens up all sorts of non-trivial possibilities such as 0.3m ahead, 0.1m behind, or a direct collision course.

And nobody would be happy with a set which after they clattered a ship informed them, "well there was a 28% chance of such a collision, a 24% chance you were behind and a 48% chance your were ahead; bad luck mate!" And neither would the MCA.
In simple terms shaded high-risk intercept zones are a better way to depict the inherent uncertainty than the faux accuracy of a precise positional fix at the TCPA.

 

[jonjo]

The answer has to do with the inherent error terms. If an AIS set informs us that we have a CPA with a ship of 5miles we will happily ignore it. If however the CPA falls to 0.2m we pull or colreg hats on and start thinking about what to do next. At this stage sailors naturally want a bearing at CPA to know whether they will be ahead of or behind the ship. Their logic is simple: a combination of a CPA and a bearing defines the precise juxtaposition of the two vessels at the TCPA.

This precise positional freeze is however no more than the highest density overlap point of two probability distributions (ship and yacht). All other possible positions are valid, just with a dwindling likelihood as one departs from the maximum density hotspot. If the CPA is five miles the likelihood of a collision will be tiny, but a CPA of 0.2 miles opens up all sorts of non-trivial possibilities such as 0.3m ahead, 0.1m behind, or a direct collision course.

Hmm an interesting if convoluted theory to explain the omission of bearing @ CPA.

Here is where your theory fails. Given all the positional uncertainty that you allege why would Raymarine be prepared to release a product that tells a customer "don't worry, CPA is 0.5NM" when in fact there is a 25% chance that the yacht will wander 0.1NM under the bow of a ship?

I prefer the simpler screw-up theory of history, it was just a matter of bungled product development in the early stages of the leisure AIS market coupled with a lack of consumer demand. Only regular cross English channel sailors will have routinely encountered the need for bearing @ CPA and this thread illustrates how even some of those were blissfully unaware of the need for bearing @ CPA.

If you can recall the period in question, 2005 to 2010. Raymarine was in a tailspin heading for likely bankruptcy and Garmin was inching its way into the marine market. No doubt there were engineers at Raymarine who thought "surely were are going to surface bearing @ CPA somewhere on the UI" but the technical initiative was lost in a company fighting to survive.

Anyhow the world is all rosy now and in 2016 Raymarine has come out with an early example of artificial intelligence that will assist the leisure boat skipper. Within 10 years it will be possible for a talking Raymarine presence in the cockpit to pipe up and say "skipper there is an interesting 6-ship crossing situation ahead in the TSS may I take the helm?".

 

[[70521]]

Anyhow the world is all rosy now and in 2016 Raymarine has come out with an early example of artificial intelligence that will assist the leisure boat skipper. Within 10 years it will be possible for a talking Raymarine presence into the cockpit to pipe up and say "skipper there is an interesting 6-ship crossing situation ahead in the TSS may I take the helm?".

The skippers reply was not recorded, but the cab boy was offended.

 

[howard easton]

Consider....

AIS says CPA is 0.25 nm. You wish to cross the shipping lane with a bigger margin of safety, say 0.5 nm. So you next need to know if you're going to pass in front or behind the ship, to know whether to increase or decrease your speed (which may be easier than changing course).

I think to do this, AIS/chartplotter needs to indicate whether you or the ship will arrive first at the point at which the respective projected courses cross. But to my knowledge, my Garmin does not display this information.

Have I got this right ? Would this extra bit of information make crossing shipping lanes a whole lot easier ? Or is it already available and I just haven't found it ?

Grateful for advice.

My advice would be not to count on AIS and CPA. And no I'm not a non tech luddite. To the contrary I have a technical degree and have done the maths for calculating the CPA. I just dont believe that the data your plotter / AIS progresses is reliable enough to trust your life to, so I still use hand bearing and eye sight as well. Whats more I still carry paper charts and use them as much as the plotter.

I used to teach navigation and was regularly horrified by the blind belief of students in modern tech. People thinking they could go into a harbour entrance at 10 kn in thick fog on GPS alone believe it or not. Their problem wasnt that they were too technical but that they werent technical enough to realise the confidence limits.

 

[lpdsn]

Anyhow the world is all rosy now and in 2016 Raymarine has come out with an early example of artificial intelligence that will assist the leisure boat skipper. Within 10 years it will be possible for a talking Raymarine presence into the cockpit to pipe up and say "skipper there is an interesting 6-ship crossing situation ahead in the TSS may I take the helm?".

The main technical challenge is how to make the slot that takes your credit card in those circumstances waterproof.

 

[maxi77]

My advice would be not to count on AIS and CPA. And no I'm not a non tech luddite. To the contrary I have a technical degree and have done the maths for calculating the CPA. I just dont believe that the data your plotter / AIS progresses is reliable enough to trust your life to, so I still use hand bearing and eye sight as well. Whats more I still carry paper charts and use them as much as the plotter.

I used to teach navigation and was regularly horrified by the blind belief of students in modern tech. People thinking they could go into a harbour entrance at 10 kn in thick fog on GPS alone believe it or not. Their problem wasnt that they were too technical but that they werent technical enough to realise the confidence limits.

One of the big problems with any AIS I have used at sea is there is little or no smoothing in the data displayed, in some respects this is good but it also makes and CPA data potentially unstable. It is how ever a good warning system particularly in a busy area but it does then need good old fashioned seamanship to use that info. It does make life easier if you understand how it works and use the information sensibly. I was taught collision avoidance on the bridge of a submarine with only a compass for comfort reporting every vessel that would pass within 2 miles to the captain, who would often then check your interpretation till he was happy you always got it right so can certainly do it all by hand so as to speak, but enjoy the greater freedom to enjoy life technology can give you

 

[dom]

Given all the positional uncertainty that you allege why would Raymarine be prepared to release a product that tells a customer "don't worry, CPA is 0.5NM" when in fact there is a 25% chance that the yacht will wander 0.1NM under the bow of a ship?

I prefer the simpler screw-up theory of history, it was just a matter of bungled product development in the early stages of the leisure AIS market coupled with a lack of consumer demand. Only regular cross English channel sailors will have routinely encountered the need for bearing @ CPA and this thread illustrates how even some of those were blissfully unaware of the need for bearing @ CPA.

If you can recall the period in question, 2005 to 2010. Raymarine was in a tailspin heading for likely bankruptcy and Garmin was inching its way into the marine market. No doubt there were engineers at Raymarine who thought "surely were are going to surface bearing @ CPA somewhere on the UI" but the technical initiative was lost in a company fighting to survive.

Anyhow the world is all rosy now and in 2016 Raymarine has come out with an early example of artificial intelligence that will assist the leisure boat skipper. Within 10 years it will be possible for a talking Raymarine presence in the cockpit to pipe up and say "skipper there is an interesting 6-ship crossing situation ahead in the TSS may I take the helm?".

To answer your question and as previously described: a CPA typically represents a semi-circle around the target at the TCPA. The introduction of a bearing at the TCPA defines a straight line and the intersection with the semi-circle in turn defines a precise point. The manufacturers (sensibly IMHO) do not want colreg decisions to be taken upon this singular intersection point, which is why FLIR have gone for a broader shaded intercept risk-zone. The respective mariners can then choose what to do about it.

As for the maths; surely nobody can believe that Garmin, Raymarine, Furuno et al are incapable of performing and coding a straightforward GCSE level vector calc?

Finally I know AI sounds ultra-clever and it is. It is normally defined as a machine capable of acting as a 'rational entity', as opposed to one which follows a series of predefined calculations. I have never heard FLIR, Garmin or indeed any of these guys referring to their AIS software as in the sphere of artificial Intelligence where the machine takes its 'own' rational decisions designed to minimise the probability of an encounter with a target vessel(s). Just imagine trying to defend that one in court if the machine misinterpreted a tricky Rule 10 situation.:ambivalence:

 

[jonjo]

As for the maths; surely nobody can believe that Garmin, Raymarine, Furuno et al are incapable of performing and coding a straightforward GCSE level vector calc?

Well yes indeedy.

That you felt compelled to make that assertion of the blindingly obvious to promote your line of reasoning indicates indicates you have not assimilated what others are saying.

Finally I know AI sounds ultra-clever and it is. It is normally defined as a machine capable of acting as a 'rational entity', as opposed to one which follows a series of predefined calculations.

A slightly odd definition, better stated as "as machine behaviour that emulates the decision making of an intelligent being'.

I have never heard FLIR, Garmin or indeed any of these guys referring to their AIS software as in the sphere of artificial Intelligence where the machine takes its 'own' rational decisions designed to minimise the probability of an encounter with a target vessel(s). Just imagine trying to defend that one in court if the machine misinterpreted a tricky Rule 10 situation.:ambivalence:

Your earlier posts lead me to suspect your perception of Raymarine software is about 16 months behind the curve, a lot has happened in that period.

My comment about AI is forward looking. Even today we are a small step away from an AI power helm being able to steer across a shipping lane more competently than some posters in this thread.

Your comment about AI and rule 10 illustrates you do not have an insight as to how a future digital helmsman would be programmed. Expert systems of the 1980's were deemed to have failed with their manually crafted logical reasoning. If Raymarine were to start development of an AI powered robot helmsman they would adopt neural nets configured for deep reinforcement learning as recently demonstrated Google's Deepmind team in London.

The robot would be fed 10,000's of real life scenarios from historic AIS traffic data and the learning would be inferred slowly from gameplay where the no.1 objective is not to be swatted by a big ship. The fully trained robot would know nothing about ColRegs, it would just reliably mimic the helming of someone who did.

I hope I have not ruined someones future patent application by putting that into the public domain.

 

[lw395]

Hmm an interesting if convoluted theory to explain the omission of bearing @ CPA.

Here is where your theory fails. Given all the positional uncertainty that you allege why would Raymarine be prepared to release a product that tells a customer "don't worry, CPA is 0.5NM" when in fact there is a 25% chance that the yacht will wander 0.1NM under the bow of a ship?

I prefer the simpler screw-up theory of history, it was just a matter of bungled product development in the early stages of the leisure AIS market coupled with a lack of consumer demand. Only regular cross English channel sailors will have routinely encountered the need for bearing @ CPA and this thread illustrates how even some of those were blissfully unaware of the need for bearing @ CPA.

If you can recall the period in question, 2005 to 2010. Raymarine was in a tailspin heading for likely bankruptcy and Garmin was inching its way into the marine market. No doubt there were engineers at Raymarine who thought "surely were are going to surface bearing @ CPA somewhere on the UI" but the technical initiative was lost in a company fighting to survive.
......?".

I agree with reservations....
Calculating bearing at CPA relies on the cosy assumption that the vectors of both vessels will remain constant.
Yachts are notorious for not keeping a constant speed or heading, and even tankers steer.
Lots of yachts have completely rubbish heading sensors in this context. So you have to rely on GPS SOG/COG which tends to be historical data.

This means there are errors involved. So it becomes a problem to report 'you are going behind it' on the same level of certainty as 'you are gonna cross its bow'.
The UI becomes subtly complex when it tries to tell us anything other than a cold hard number for us to process mentally.
ISTR conversations with people who have worked at Raymarine doubting the ability of the average buyer to correctly interpret bearing at CPA.
I think displaying the tracks on a yacht-centred, course-up mock radar display is wonderfully clear, but incomaptible with the 'steer the little boat around the picture' ideology.

And can we have one that understands the racing rules too?

 

[Lon nan Gruagach]

The robot would be fed 10,000's of real life scenarios from historic AIS traffic data and the learning would be inferred slowly from gameplay where the no.1 objective is not to be swatted by a big ship. The fully trained robot would know nothing about ColRegs, it would just reliably mimic the helming of someone who did.

Oh dear oh dear oh dear.. not he Show a neural net a gazillion pictures and it can do anything garbage.
The probability that a neural net will develop a catastrophic psychosis tends towards 1.
simple case: counting people in a crowd.
The chances of a scenario on the lines of every picture you show happens to have equal numbers of red jumpers and yellow jumpers and the AI counts red twice and ignores yellow, is alarmingly high.

The fully trained robot will screw up.

 

[dom]

...Your earlier posts lead me to suspect your perception of Raymarine software is about 16 months behind the curve, a lot has happened in that period.

Even today we are a small step away from an AI power helm being able to steer across a shipping lane more competently than some posters in this thread.

The robot would be fed 10,000's of real life scenarios from historic AIS traffic data and the learning would be inferred slowly from gameplay where the no.1 objective is not to be swatted by a big ship. The fully trained robot would know nothing about ColRegs, it would just reliably mimic the helming of someone who did.

I hope I have not ruined someones future patent application by putting that into the public domain.

Too dumb to solve a basic piece of Euclidean geometry, but smart enough "to boldly go where no sailor robot has gone before". Wow!

As to your fear of ruining a major patent application with your random tech musings. Well I expect our friends in Silicon Valley will get over it in time :encouragement:

 

[johnalison]

I have to admit that I don't see the value in knowing the exact bearing of a vessel at the CPA, unless I suppose I intend firing on it. Most of the time, I am content to know that it will pass ahead or astern, and it is usually self-evident from the screen that it will be "on the starboard bow" or "abeam to port" as the case may be.

 

[lpdsn]

It's probably far easier to automate the ColRegs than to programme up self-drive cars. No doubt the techniques developed for self-drive cars will eventually feed across.

What must on no account happen is to allow these AI systems to ever access an on-line ColRegs debate. They might get funny ideas.

"Might is right, Dave..."