Jump to content


Photo

Steel Beasts Pro P E, Version 4.1

July 2019

  • Please log in to reply
147 replies to this topic

#81 Chris Werb

Chris Werb

    In Zod We Trust

  • Members
  • PipPip
  • 11,007 posts

Posted 04 July 2019 - 0435 AM

Hi Ssnake, I'm looking forward to your next revelation re 4.1. :)
  • 0

#82 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 04 July 2019 - 0725 AM

Currently I'm trying to compile the Release Notes. Three years of changes have accumulated; even if maybe 30% of the bugfixes were regressions from refactoring that do not need to be described, it's still a rather tedious and time-consuming activity that largely prevents other things, such as making yet another video. I will make at least one more, promise, and hopefully I'll find the time for one or two tutorial videos as well (especially WRT the differences in file handling of map packages). But, Release Notes first.


  • 0

#83 Chris Werb

Chris Werb

    In Zod We Trust

  • Members
  • PipPip
  • 11,007 posts

Posted 04 July 2019 - 0749 AM

Currently I'm trying to compile the Release Notes. Three years of changes have accumulated; even if maybe 30% of the bugfixes were regressions from refactoring that do not need to be described, it's still a rather tedious and time-consuming activity that largely prevents other things, such as making yet another video. I will make at least one more, promise, and hopefully I'll find the time for one or two tutorial videos as well (especially WRT the differences in file handling of map packages). But, Release Notes first.

 

I hope you will include the funnies which have become a staple of SB release notes and make them unusually readable :)


  • 0

#84 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 04 July 2019 - 0846 AM

Seems to be the only way to make a document like that consumable at all. Also, it preserves my own sanity while writing it. Or it's my insanity leaking through the cracks, who can say?


  • 0

#85 Chris Werb

Chris Werb

    In Zod We Trust

  • Members
  • PipPip
  • 11,007 posts

Posted 04 July 2019 - 0929 AM

The word "borkened" must make at least one appearance :)


  • 0

#86 Chris Werb

Chris Werb

    In Zod We Trust

  • Members
  • PipPip
  • 11,007 posts

Posted 07 July 2019 - 0948 AM

I have just read your description on what you had to do to implement suspension movement over high fidelity terrain on the SB forum. I had to read it three times and let it sink in for a few minutes thereafter to comprehend it. I'd really like you to re-post it here to give an idea of the work involved in continually upgrading SB and your team's achievements in doing so.


  • 0

#87 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 07 July 2019 - 1558 PM

I'm proud of the team's achievements, yes. At the same time, for the end user the critical element is that old scenarios still work, that the new version largely behaves like the old one. I don't want to gloat in "lookit muh new technology!!!" style. I will take the greatest pride in knowing that there were massive changes underneath, and they are largely imperceptible for the average user.

 

The vehicle suspension wasn't the greatest challenge. we basically had to repair the damage that the introduction of a disruptive new technology created first. So the question ultimately is whether the new technology - ultra-high resolution terrain - was worth it. I gave a presentation of that at the NATO CAX Forum last September. I can't say that I have proven with rigid science that it is, but there are strong hints from early adopter customers that 2m resolution LIDAR data (or better) really DO change simulation outcomes substantially. In one experiment that I ran I compared simulation outcomes of exactly the same scenario where I let a single M1 run the gauntlet of a company-sized static defense with ATGMs and tank gun fire (I cherry picked the ammunition to be as harmless as possible to generate as many shot events as possible).

So, in the 10+ iterations I had nearly 1500 rounds of tank ammo fired, and almost 1,000 ATGMs.

In a DTED-3 based terrain database of the same terrain there were basically two peaks of engagement distances, at about 2,000m and at around 900m. The average engagement range was 1200m for tank gun fire, and 1300m for missiles.

Contrast this with a 78cm resolution LIDAR scan of the same training area; Ammo consumption was cut nearly in half because a lot of micro-obstacles (sand dunes) block the line of sight - obstacles that are simply lost in the lower sample rate of the DTED-3 database with its 10m resolution. The average engagement range shrunk to 500/670m for gun and missile engagements respectively, with basically no hits beyond 900m range.

 

This opens up some interesting questions, such as if "we" (="the west") have attempted to gain standoff superiority for the duel situation because of the Israeli experiences in 1973, and then "confirmed" the importance of standoff superiority in simulations that were all no better than 10m resoluition (and often worse) - in other words, some collective confirmation bias in doctrine and technology development?

I don't have the answer. But this very simple experiment shows that c.p. the change of the terrain database resolution can have a massive effect on the fidelity of simulation outcomes.

 

As much as I would like to claim that I knew this back in 2014 when the decision to go for high res terrain was made, the reality was that I didn't have a clear idea of the outcome except that in real life terrain is often more restrictive than the basic model that we followed since Steel Beasts 1.0 with the basic assumption that every unit could reach any spot in the terrain, at least in principle. But I didn't think that it would have such a dramatic effect on line of sight and engagement ranges. I do believe however that once that you experienced it first hand, you wouldn't want to go back.

 

A similar change (but better contained, with much fewer ripple effects elsewhere in the code) was the transition to the new HE model. And while we aren't finished yet, what we have already shows quite some promise and is in any case much better than what we had before, hands down. Before we basically operated with a hemispherical proximity function to determine the damage that a HE round would create. It's not "entirely wrong" but came with a number of anomalies, not the least making it very difficult to shoot down helicopters, or making it very easy to kill tanks with even 20mm HE rounds if only you knew where to aim (the ground right under the belly of a tank, I think I can now tell this dark secret).

Calculating thousands of fragments in real-time and in a realistic fashion was another major challenge. We have very talented people in the team; much bigger competitors haven't accomplished comparable fidelity.

 

But again, this comes close to the "muh tech" gloating that I don't really like. Our innovation should be driven by training requirements, not the latest graphics card generation and what kind of render technologies it supports. The real benefit of the new HE model is that it's much easier for us now to tailor the terminal effects of explosive projectiles, largely based on open information such as the shell weight, the amount of HE filler, and similarly easy to find data. So, if then something's not quite right, the AAR provides the visual feedback of what happened and customers can then point out what they'd like to see changed, but it largely minimizes subjectivism and guesswork. It gives us and our customers greater confidence in the simulation results.


  • 0

#88 sunday

sunday

    Bronze-age right-wing delusional retard

  • Members
  • PipPip
  • 12,035 posts

Posted 07 July 2019 - 1609 PM

This is true progress, thanks for sharing. Still, it looks like you are not doing simulation of, for instance, HE effects, but computer modelling of those with only a small bit of randomization introduced.

I, quite humbly, suggest you compare the pains your team took to model these things, and compare with the simulations used to guess the influence of CO2 content in Earth's atmosphere with the average warming of the planet.


  • 0

#89 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 07 July 2019 - 1625 PM

...???


  • 0

#90 sunday

sunday

    Bronze-age right-wing delusional retard

  • Members
  • PipPip
  • 12,035 posts

Posted 07 July 2019 - 1904 PM

Computer modelling is not the same as computer simulation, and you have experienced the difficulties of modelling small phenomena.


  • 0

#91 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 08 July 2019 - 0156 AM

Your point being?


  • 0

#92 sunday

sunday

    Bronze-age right-wing delusional retard

  • Members
  • PipPip
  • 12,035 posts

Posted 08 July 2019 - 0431 AM

Never try to make a Prussian look smart when he does not want to :P


  • 0

#93 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 08 July 2019 - 0502 AM

I really have no idea what you're trying to say, and more of "wink, wink, nudge, nudge, say no more" isn't going to help.


  • 0

#94 DB

DB

    Crew

  • Members
  • PipPip
  • 11,382 posts

Posted 12 July 2019 - 0800 AM

Just guessing here...

https://www.research..._and_simulation

Not quite sure, but a model used inappropriately can generate unrealistic results. I'm not sure that comparing HE fragment models with climate change models is particularly helpful here.
  • 0

#95 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 12 July 2019 - 1123 AM

Yeah, we can collectively guess what he meant to say. He could simply explain himself, but apparently he doesn't want to.


  • 0

#96 sunday

sunday

    Bronze-age right-wing delusional retard

  • Members
  • PipPip
  • 12,035 posts

Posted 12 July 2019 - 1228 PM

Just guessing here...

https://www.research..._and_simulation

Not quite sure, but a model used inappropriately can generate unrealistic results. I'm not sure that comparing HE fragment models with climate change models is particularly helpful here.

 

The point is the difference between modeling, like finite element modeling of a mechanical piece, where one knows that the underlying equations could be solved in a finite time, and simulation of a system whose underlying equations are known to be unsolvable, or chaotic, thus could not be modeled. The "chaotic" bit means that a very small deviation in the initial conditions of the model gives widely varying results, in orders of magnitude. One example is the famous "butterfly effect".

One needs a background in numerical calculus to understand that, or a mind sufficiently open.

 

I've done both, modeling and simulation, of stochastic systems related to fail models of fault-tolerant systems. For instance, you could compute the mean time between failures (MTBF) of one such system by solving a system of linear equations once, or by simulating the evolution of the system along time a number of times high enough to obtain an acceptable interval of confidence. In those systems the inputs are the topology of the system, the failure rates of each individual component, the repair rates of each component, the repair policies, and the error mitigation policies. Were one of such systems chaotic, a small variation on one of those inputs would result in MTBFs that could range from seconds to years.

 

Then one reads something about "Black Swans", and failure engineering goes back to divination, entrails, and voodoo, apparently. But that is matter for another thread.


Edited by sunday, 12 July 2019 - 1257 PM.

  • 0

#97 Ssnake

Ssnake

    Virtual Shiva Beast

  • Members
  • PipPip
  • 6,861 posts

Posted 12 July 2019 - 1641 PM

I never had the desire to win the competition for the Alpha Nerd position so I'll happily cede that to you. But you still haven't explained what you meant with

 

I, quite humbly, suggest you compare the pains your team took to model these things, and compare with the simulations used to guess the influence of CO2 content in Earth's atmosphere with the average warming of the planet.

 

My compiler just returns a giant WTF.


Edited by Ssnake, 12 July 2019 - 1642 PM.

  • 0

#98 Chris Werb

Chris Werb

    In Zod We Trust

  • Members
  • PipPip
  • 11,007 posts

Posted 12 July 2019 - 2209 PM

Well that was..... different.


  • 0

#99 RETAC21

RETAC21

    A la lealtad y al valor

  • Members
  • PipPip
  • 13,678 posts

Posted 13 July 2019 - 0353 AM

I think that what Sunday was to say is that most game will say, 20mm shell = 10% chance of a kill, while SB will work by saying: a 20mm shell hits a 10mm plate at 90º penetrating to hit the next plate at 80º causing a failure to penetrate but causing spall, etc, etc. resulting in a non-kill.


  • 0

#100 sunday

sunday

    Bronze-age right-wing delusional retard

  • Members
  • PipPip
  • 12,035 posts

Posted 13 July 2019 - 0435 AM

I think that what Sunday was to say is that most game will say, 20mm shell = 10% chance of a kill, while SB will work by saying: a 20mm shell hits a 10mm plate at 90º penetrating to hit the next plate at 80º causing a failure to penetrate but causing spall, etc, etc. resulting in a non-kill.

 

Also that. The importance of modeling the terrain in high resolution also shows how assessment results do differ with the accuracy of the definition of constant conditions.

 

Let me insist, I think Steel Beast is an excellent product, was an excellent product, and should be an excellent product in the future.


Edited by sunday, 13 July 2019 - 0519 AM.

  • 0