News:

Precision Simulator update 10.175 (29 May 2024) is now available.
Navburo update 13 (23 November 2022) is now available.
NG FMC and More is released.

Main Menu

Statistical randomness

Started by Hardy Heinlin, Mon, 9 May 2011 23:05

Hardy Heinlin

In PSX there'll be the option to trigger malfunctions at random.

Similar options were available in PS1. I remember in the early forum years there were some discussions about the probability; some users found that the malfunctions occured not often enough when, for example, 1 event per hour was set. Other users found they occured too often.

And I used to explain that random means random, and not exactly 1 event per hour. The distribution of events along the time axis is diffused over many hours, not just over one hour. It may happen that there'll be no event in 10 hours, or 10 events in 1 hour -- it's unpredictable. 1 event per hour is just an average measured over several days or weeks ...

To make this principle more clear, I called it in PSX "statistically random". But I fear it won't make it clearer :-)

An alternative mathematical solution to guarantee 1 event per hour would be this: If no event occured in the last 59 minutes, then force one in the 60th minute. -- But then it's predictable! Not good.


Cheers,

|-|ardy

frumpy

Quote from: Hardy HeinlinTo make this principle more clear, I called it in PSX "statistically random". But I fear it won't make it clearer :-)

I suppose ;)
Maybe an explainatory sentence like "instrument will fail within the
next XX minute(s)" might do it. But thats a lot of text for a simple
idea. Let them pilots find out what "statistically random" means *g*

Shiv Mathur

Quote from: frumpyMaybe an explainatory sentence like "instrument will fail within the
next XX minute(s)" might do it.

As I understand Hardy, that's not correct.  There is no guarantee
that it will fail within x minutes.

Shiv

Mundyas

Hi |-|

I think people get confused when statistics is mentioned and tend to think it is too hard (for them ). Though most guys seem to understand easily enough the vital statistics of an attractive woman and no doubt vice-versa!!

Random is as random does! (as the old expression goes)

Why not just rename it random failure(s) and as you will have the function separately (as you have in PS1 now)  to say have an engine fire in say 132 seconds it will be fine for me at least!

I always liked to know the engine problem was coming on the simulator as then I was ready for it. In real life it obviously just happens as in NZ a few days ago and the pilot training kicks in for a once in a lifetime event.

 /-\\  for Andrew

Shiv Mathur

Quote from: MundyasWhy not just rename it random failure(s)


It's all explained in Hardy's opening post  ... people
were misunderstanding what 'random' implies.

Cheers,
Shiv

Mundyas

Hi

Yes, Shiv you are right of course. I should have read |-| more closely.

It just seems to me that calling it random failure(s) is less confusing. And that is the conundrum that |-| faces!

Well guys hope you are flying today

EGKB - just now CAVOK 15C lovely weather.

Andrew

Will

I thought of a way to have it both ways.  THere are two legitimate desires here.  Desire #1: Some people will want to experience a failure on a flight, but they won't want to know when it's going to happen.  A great example is engine failure... they may want to fly in PSX specifically to train engine failures, but they might not want to know exactly when it's going to happen.  They would be disappointed if they flew for an hour and no failure occurred.  Desire #2: Some people will want to approximate the randomness of the real world... maybe there will be a failure somewhere, maybe not.  These people want the underlying randomness, but won't care if there's a failure in a specific flight or not; they would just want for not every flight to be mechanically perfect.

So here's how to combine them:

( o )  No failures.
( o )  ____  failure(s) within the next ____ minutes.
( o )  Failure rate of ____ event(s) per hour (on average).

The middle option defines set parameters: If you ask for 1 failure within the next 60 minutes, then it satisfies desire #1, the person will get their engine failure at an unspecified point within the next hour.

The third option sets a rate.  If you enter a rate of 1 failure per hour, you wouldn't be guaranteed any failures at all, even over 10 hours... but the longer you flew, the closer you'd get to an overall average of 1 failure per hour.  Importantly, you could enter decimals in the second option, such as 0.25 events per hour.  (Alternatively, you could require whole numbers and as for a number of events per 100 hours, or 1000 hours.)

Just something to think about....
Will /Chicago /USA

Jeroen Hoppenbrouwers

#7
I think that a more usable second type of failure would not be a hidden random value that still is guaranteed to happen within one hour, but instead an indirect variant.

No failures: clear. Nothing will ever break.
Random failures: clear. Things may break if you fly long enough, possibly a year depending on the setting.
Guaranteed failures within a set time period: use the existing 'failure in 123 seconds' option, but either:

1. Have somebody else set the value for you (instructor style, especially when multiple failures occur).

2. Have a 'popup style' automated instructor select a seconds value for you that satisfies your one-hour desire, but don't show this value (by first closing or masking the instructor screen and only then filling it in).

This second automated instructor option would not change anything at all in the existing setup, it would be a way to roll a dice to set the number of seconds without seeing the outcome. Once rolled and set, don't open the failures screen, as all stuff will be plainly visible. And when you save the situation, all settings will be saved, unlike with random (where the random will be saved but not the actual value as that remains unknown until the failure actually happens).

Sounds complex? Possibly. But the implementation may be very simple: if you put in, say, -3600 seconds as the trigger time, PSX will upon saving of the malfunction screen replace each negative value with a random value between 0 and 3600. And this gets saved.

It won't even conflict with existing negative values (that signal that the malfunction has already occurred).


Jeroen

Will

I like that, Jeroen, it's intuitive and sounds like it would meet the needs.

I thought of a fourth need, though.

1. No failures.
2. XXX Guaranteed failure(s) within the next YYY minutes.
3. Truly random failures per unit time, which may include no failures at all.
4. Chance of a event-triggered malfunction being triggered or not triggered.

For #4, the new one, let's say that my training is going well except that I can't fly properly when I lose an engine at V1.  If we set the sim to fail an engine at V1 on every takeoff, I get used to predicting it, and I'm prepared.  That means I perform artificially well.  But if the sim were to fail an engine at V1 only 15% of the time, it would be more of a surprise and I'd learn how to handle the event when I wasn't expecting it.
Will /Chicago /USA

Jeroen Hoppenbrouwers

#9
Hm, this kind of specific training scenarios may need a true artificial instructor. This starts sounding like a job for an add-on. PSX may restrict itself to providing a proper instructor station, but may leave the artificial instructor to an add-on.

Real sims don't generate training scenarios. Instructors do this.


Jeroen

Hardy Heinlin

#10
Quote from: Hardy HeinlinAn alternative mathematical solution to guarantee 1 event per hour would be this: If no event occured in the last 59 minutes, then force one in the 60th minute. -- But then it's predictable! Not good.
Well, OK ... it is good if this "forced" method is optional. Thanks for the inspiration, guys :-)

Actually, random generators occur at three different places on the Instructor:

- In each single malfunction item, in a pop-up menu
- On a general page similar to PS1's "Random malfunctions ..." page
- On that page with the title "Self-induced"

Now, on the general Random page I'll include both options: forced and non-forced random.

In the individual single malfunction items I'll replace the ...
"Statistically random: [ ] Events/day"
... by ...
"At random within next: [ ] minutes".

I won't add it, I'll replace it (keep it simple, don't bloat the software): The "Statistically random" feature will still be available on the general Random page, and there it makes more sense because that's the area for the very unpredictable stuff, while the single items are rather for the forced stuff. You won't sit down and spend half an hour programming random settings in each single item, you will do this on the general Random page by a few mouse clicks.

(The general Random page is under construction. Nothing to show there yet.)


Cheers,

|-|ardy





Will

#11
Hardy, on the second photo you posted, can you add a right-most column that says "Probability"?  That should be self-explanatory, but in case it's not, what I mean is that with a probability of 0%, you'd get the same as having no boxes checked in that row, i.e. no failure would ever be triggered.  With a probability of 15%, then after 1000 events, the failure would have triggered roughly 150 times.  Finally, with a probability of 100%, the failure would always be triggered at the respective event.

That would be so useful.
Will /Chicago /USA

Hardy Heinlin

#12
That's exactly what I want to add to the general Random page, Will.

The Self-induced page is for guaranteed failures (for which it makes no sense to be triggered by an airspeed or altitude etc. "Trigger the hot start at 12:55z or at 5000ft" makes no sense).

While on the general Random page there will be the more unpredictable stuff. There will be several categories (take-off malfunctions, cruise malfunctions, mixed malfunctions etc. pp.).

Will

Aha, okay.  That will be a great boon for people who want to use PSX to sharpen their skills, and don't have the luxury of an instructor sitting behind them.
Will /Chicago /USA

Will

I just noticed you have airframe malfunctions... Excellent!  What on earth are you putting here?
Will /Chicago /USA

torrence

Just to add to the confusion, here's a minor variant that I want sometimes on PS1.  I basically liked the 'random' X failures per 10 hrs (or whatever) since it combined not knowing what would fail with not knowing when it would fail.  The problem was that if you wanted to practice abnormals on a given day, setting the failure frequency high enough to produce some (any) problem in a reasonable time almost guaranteed that you would have a several cascading failures to deal with.  That OK in the sense that it can happen in the real world - e.g. the A380 incident - but not good if you are trying to get accustomed to handling a 'simple' failure thrown at you randomly.  What I was looking for in this case was being able to set the 'rate' of failure so something (random) happened on a short (1-2 hr) flight but after the failure, subsequent failures were inhibited.  

Not sure this is worth pursuing with all the combinations that Hardy is already building into the system.

Cheers,
Torrence
Cheers
Torrence

Hardy Heinlin

Airframe malfunctions: E.g. doors, APU, gear, flaps ...

QuoteWhat I was looking for in this case was being able to set the 'rate' of failure so something (random) happened on a short (1-2 hr) flight but after the failure, subsequent failures were inhibited.
This is a bit like the random categorization in "Severe malfunctions" and "Excessive malfunctions" in PS1. In "Severe" I think it was guaranteed that no more than 2 engines would fail. I think PSX will have some more control tools for random settings.

Creating random settings is like placing furniture in a room sometimes :-) A little bit more to the left ... hm, no ... more right ... a bit more here ... or better a little higher there ... hm ... do I want an engine failure in the evening? Or better later? No.1 or no.4? Perhaps a little fire afterwards ...? I want that one for sure, and this one by surprise ... and the one in the middle semi-surprisingly-pseudo-sure ... :-)


-||ardy

farrokh747

#17
Hi Hardy - the detail here is amazing! - a question: For the builder that will have a remote instructor screen, is there a touch screen layout for the IOS? IE the elements are placed graphically "in boxes" as opposed to drop down menus, etc. Also, is there a weight and CG/balance page in the IOS? (would this be the Load page?)  Apologies if this has been answered elsewhere...

cheers!


Farrokh.

Hardy Heinlin

#18
Hi Farrokh,

the layout is as shown on the screenshot. Are the boxes too small for your touchscreen?

W/B can be set by sliders (touch & drag) and probably also by numerical entry.

...

I just realized there are two ways to run the pop-up option "At random within X minutes". If I enter, say, 5 minutes, the entry field will ...

- ... always read 5 until the malfunction is triggered.
- ... countdown 5 ... 4 ... 3 ... every minute until the malfunction is triggered (it may be triggered anytime, but the smaller the value, the higher the chance, because it must be triggered within 5 minutes).

The former method is a liar. It reads 5 when actually less than 5 minutes are left.

I find the latter the only logical solution. But it has the disadvantage that the user can literally see the probability rising -- so no surprise. On the other hand, if the user is an instructor, it's nice to see what's going on. I could add the option to make the entry unreadable by **** like a password entry. Yaaaaawn ... options, options, options, ... release date: 2020 :-)

No. No option here.


|-|ardy

farrokh747

morning hardy - the Inst stn would be expanded to full screen, so i think the checkboxes should be ok: maybe a keypad in there somewhere? the idea being you don't have to go to the mouse....   :D

Typically, there would be 2 screens at the IS - one for all inputs (fail. pos. etc - usually a touch screen) and another with a map display with nav aids, apt/rwy, ILS, acft path trail, etc

http://www.mechtronix.com/nc/en/photo-gallery/single/?tx_jmgallery_pi1%5BalbumUid%5D=1&tx_jmgallery_pi1%5BimageUid%5D=103