Pukf dpm pack 1.0

[Imshi-Yallah 0]

I guess you've never been used to better

Company level, was an error I forgot you don't use anything in the Carl Gustav class since LAW80, or a 60mm/51mm mortar anymore.

So yeah stuff from support company, MILAN, GPMG SF and 81mm.

[Tom Norman 0]

A MILAN's supposed to be on the cards, dunno about the GPMG SF...how would you do that in flashpoint? What would you make different? And why would you want to? You'll never be in a firefight in flashpoint at a range of over 300m anyway. If you just mean you want a mounted version, then have no fear, the LandRover pack has every possible variety you could ever want. An 81mm mortar might be nice...but CoC have already provided us with one so it'd be pointless us making our own for use with UA...unless you want one that can be broken down and carried. Although I never really saw the point of that, it's a little too much for flashpoint...who'd want to haul around mortar componants for the duration of a mission and eventually just sit there manning it?

[Imshi-Yallah 0]

Something that the AI could use to whack out a treeline would be nice, would there be any point in integrating those AI mortar scripts into and addon?

Yeah the WMIKs will do nicely, what about an LSV?

(Who actually uses the LSV for milans, is it an Air assault role thing?)

[Gollum1 0]

Mortars are routinely carried along for assaults in FDF MP. More useful than portable MG's for assaults, while MG's are better for defence. Mortars can make all the difference, and because of their range you actually have to move them close to the target, about a forest or hill away. It just wouldn't work if the mortar is firing from 1 km away, from the base that is normally the starting point, way too inaccurate. In short, carriable mortars are very useful and should be included IMO.

[NeMeSiS 11]

well, the only thing i wanna know right now is how u set the dirtection for stabbing the bayonet in the right way, so when an enemy is left of him, he will look to the left and stab, not just stab in the direction he was already facing if u know what i mean  

EDIT: fixed an enormous amount of grammar rhingys, and probably still forgot some  

[miles teg 1]

I'd just be happy to see a sniper rifle on these lads... or is there one and I missed it in the map editor???

Chris G.

aka-Miles Teg<GD>

[NeMeSiS 11]

I'd just be happy to see a sniper rifle on these lads... or is there one and I missed it in the map editor???

Chris G.

aka-Miles Teg<GD>

yea i missed that 1 too

[Jinef 2]

Excellent work.

It really is nice to see some decent British kit coming out.

I am waiting for the slight bugs in Bayonet to be fixed before I make any missions though.

About the mortars:

FDF mortars are nice and simplistic and functional in the game, but they have a range of 900 metres. However i'm sure that if you are willing to make some British 60mm and 81mm mortar models and get in contact with the chain of command, they will fit you out with a mortar system capable of delivering death out to 6kms (81mm) using either WGL style user controlled interface or UA 1.0 AI controlled system. Considering Mortars and Artillery are the basis of the British military (All mobile forces flank and trap enemy, keep them suppressed with their own weapon systems until the artillery smashes the shit out of them). You guys made the MLRS for CoC, unite once again

Oh and if you want voices for the UKF mod's voice pack that you will be doing then just ask

Share and enjoy!

[Placebo 29]

Oh and if you want voices for the UKF mod's voice pack that you will be doing then just ask

Ditto

[Tom Norman 0]

Well cheers fellas, but there's more than enough British blokes on the team to fill that role.  Some of them are current/ex-squaddies, with signals experience too, so we can handle any voice bits we need doing ourselves.  I actually joined UKF to do that originally (arty signalling to be precise), but they had no use for a voice-over bloke (for the same reasons I just specified)...so I was recruited as their chief scripter instead.

I'll be sure to pass on your requests for a mortar...turns out I was wrong, you all seem very keen to crew a mortar.  Why don't you join the artillery?  We do it better than any mortar platoon anyway.  

Sniper rifles are supposed to be a work in progress as far as I'm aware.  I'm a little puzzled about this myself to be honest, I saw a (badly textured) sniper rifle amongst the early infantry...I think that might have been the BIB infantry at the time...dunno what happened after that.  JONNO!

The LSV is not planned as far as I know.  We've got rovers and their sporty GPMG mounted brother so we're not really bothered about making that slightly bizarre vehicle.  It's an airborne thing anyway.  Next release should be the landies (which are ready to go as far as I can tell), after them is the Warrior (and probably OPV variant with built in fire mission options - as written by yours truly), then I expect work on the Chally 2 will resume (dessie version first for some reason, but I'm sure that will change).  There's also an AS90 in the works.  The model looks good but the textures are non-existant...personally I don't think it will ever get off the ground because arty is so much easier (and more effective) to represent with scripting.  Plus, CoC just isn't very realistic...not for British arty anyway.  

As for setting the unit's direction to face the enemy, if you send me your email addy, I'll send you a script snippit I put together ages ago and now use quite commonly.  It's a short trigonometry equasion which finds the angle from one point on the map ([X,Y,Z] grid) to another (just picture an imaginary right-angle triangle on the map).  It's easily applied into your own scripts and I often reuse it instead of writing the whole thing out again using new variables (besides, you can just change all the variables in a oner using CTRL-H in notepad).  I've got some others along the same lines if you like.  Trig based script snippits that you otherwise might not be able to get your head around.  Would they be useful to you?

[Jonno 0]

Just to clear up a few points, the Landrovers need the interiors finishing off, they are 1/3 through a retexture, to bring them upto par.

As for mortars, we have no current plans to add them, but a Sniper unit will make an appearance at some stage.

So after we get the v1.1 DPM pack update out, we'll (DeLiltMon and me) be concerntrating on finishing the Landrover interior ready for release.

Then once that is released I expect DeLiltMon will move right onto the LandRover XD Wolf variants, including the WMIK version with the L1A1 HMG mounted.

A few of us will be working on multiple projects, therefore it may take a while untill you see progress, but we should end up having a few releases bunched together.

Not to forget all the other guys beavering away on projects, some trying their hand at texturing for the first time, as its hard to find the talent, so maybe we'll try create the talent.

I mustn't forget to mention the mission makers also working away on creating mission to utilize the DPM pack and Landrovers. But we hope to see some decent user missions too! *nudge*

nuff said, I'll say no more

[lpdotcom 0]

Has anyone made a mission using this addon yet?

[alcazar 0]

Any chance of adding a copy of the configs to the 1.1 update? It'd make it easier to create new groups armed with Kabal's SLR, etc for personal use.

Or even to have a crack a replacing the standard OFP units using OFP:Y2K3 or the like. Now there's a thought!

[Ex-RoNiN 0]

Does the line <table border="0" align="center" width="95%" cellpadding="0" cellspacing="0"><tr><td>Code Sample </td></tr><tr><td id="CODE">[this, N] exec "\ukf_infantry\scripts\init_beret.sqs"

need adapting if used in MP or when the addon is located in a mod folder?

[dinger 1]

Hi, three things:

A) the UKF_CLANSMAN is the wrong case to be recognized by CoC_UA (I suspect the config was changed between last fall and now)

B) We have spoken with UKF about making the AS90 work with CoC UA. The UKF M270 MLRS (Also minus texture) is currently available with the CoC_Arty release.

C) Artillery is certainly easier to represent via scripting, and might be more efficient, if by efficient you mean "10 seconds after a mouseclick the rounds land right on the point of mouseclick." As for CoC arty not being very realistic, I'd love to hear how you feel it can be improved, since we are constantly striving for the most faithful simulations possible in OFP. In beta, we have every major type of indirect fire working and firing every major (and many minor) type of shell and fuze. Yes, the mission reaction times are reduced, and the range and charge settings -- while based on real-world numbers and corroborated by a motley assortment of FACs, ANGLICOs, redlegs, gunbunnies and others -- are simplified (it is a game after all), but I do believe we make a reasonably realistic product (the radio comms need work though).

[Tom Norman 0]

Relax, I was just talking about the VP and fire orders for British arty.  I'm sure it's perfect as far as American proceedures go, but we do it a different way and it would be nice to represent that in a UKF addon.

By 'easier to represent via scripting ' and 'more effective', I don't mean 'death by mouse click' at all.  Your UA system can lob shells all the way accross the map if need be, but they're not very accurate when they get there, the distributions are dodgy as you like and adjustment leaves a lot to be desired.  My script only spawns a shell a couple hundred metres away from the target and launches it at 350m per second at the calculated point of aim.  The end result is a round that will land consistantly within about a metre of the same point of impact every time.  This means I've got a greater degree of control over where the rounds fall and can adjust the simulated inconsistancies as I see fit.

Now don't get me wrong, your script is genius in it's purest form, but it seems factors outside your control (ie. limitations of the OFP engine) are stopping you from getting the results you want...or if they are exactly what you were after, then I think the effects on the ground you have created are too large-scale for a small-scale game like Flashpoint (you know what I mean...it's a big playing field as far as games go...but real arty plays on a far grander scale).  Also, artillery fire is actually quite a bit more accurate than I think you have represented in UA.  In Iraq, we recieved only one or two adjustments for the duration of the war.  Phoenix gave us most of our grids, but when the mission was sent by one of our OPs, he'd just go straight to fire for effect, and adjust that if need be (which hardly ever needed to happen).  I co-ordinated a fire mission in support of two American sea harriers to engage 8 T-55's.  I think they called it a CIAD mission...new one on us at the time, but they wanted us to plonk an illum round down in the middle of the enemy position (functioning at ground level).  We put the first round right where they wanted it.  Then they bombed the bjeezus out of them before we plastered the survivors with 80 rounds of bomblet.  That should give you some idea of how accurate arty fire can be with decent met.

Anyway...like I said, I'm sure it's spot on for American arty proceedures, but I do this stuff for a living and I can see it doesn't follow quite the same lines as the way we do it in the British army.

And please...what does a bunny know about fire dicipline and data calculation?  They're button pushers and breech creatures.  The only reason God gave them heads is so they'd have some place to put their helmet, thus freeing up their arms to carry more heavy things.  If you have any plans on representing British fire dicipline and proceedures in UA, then you'll want to get hold of a British CP Ack or the like.

[denoir 0]

My script only spawns a shell a couple hundred metres away from the target and launches it at 350m per second at the calculated point of aim.  The end result is a round that will land consistantly within about a metre of the same point of impact every time.  This means I've got a greater degree of control over where the rounds fall and can adjust the simulated inconsistancies as I see fit.

The neural net that does the ballistic calculations in UA can give as precision of less than 1 m at distances up to 32 km. It can always hit within a couple of meters. The precision and bias error are artificially added to simulate the correct PE/CEP as defined by DoD standards for the weapon systems simulated.

The PE/CEP  are defined as that 50% of shells should land within a circle with the radius of the CEP value. Note that the table above is only for spotting accuracy. In addition you have the precision error (weapon precision that is individual to the piece) and bias error (external factors such as pressure variations, chaotic wind patterns..).  Overall the UA system is capable of precision of an order of magnitude better than a real system.

As for more precision, you have guided payloads such as the copperhead (laser) or STRIX (IR). These and several other will be included in the next release of UA.

To give you a practical demonstration of the precision:

The target: An innocent bus stop:

The target position is marked with the flag pole:

We call in some shells in low angle trajectories (high angle is more precise). The 155 mm (virtual in this case as this is just the neural net running) howitzer is at a distance of 6,6 km and with a difference in hight of 85 m. The elevation is 26.7 degrees and the muzzle velocity is 300 m/s:

Splash:

Closer inspection of the hits:

As you can see, precision is no issue

[dinger 1]

haha.

Yes, I agree that often artillery is simply "too big" for the OFP game. The CPU can't handle more than a platoon on platoon engagement in most cases.

And, having done an artillery script myself, I can appreciate many of the benefits -- you just create the shells in the pattern you like and go home.

UA accuracy is actually pretty good.  Denoir's neural net gets accuracy within 1m.  By the time you rig it to UA units and OFP's firing system, it's more like 5-10m (Depending on how well you calibrate it --- heh I posted at the same time as Denoir. He'll be hurt to know that OFP's firing system isn't quite as accurate as his neural nets).  To see that, place the zero bias and zero precision error logics on the map.

As for why it's so wild, well, while the artillery pieces in the pack vary in when they were first deployed, we wanted to keep the simulation focused on something that was on par with the 1985 theme.  And even so, we've adjusted down the error for the next version, so you usually don't need to make adjustments unless the target's moved.

Yeah, I spoke to a marine last fall who was pretty big on SEAD fire missions: suppressing fire , then one illum on deck (night) or smoke (day), then more fire....  That's easy to do.

Edit: and I'll add, that UA is flexible enough to support entirely different procedures.

[Tom Norman 0]

Excellant...you just said exactly what I wanted to hear.  Greater accuracy in the next version, more ammunition natures/fuze settings and scope for British proceedures.  Perfect.  It'd be great to see copperhead make an appearance too, it'd be an excellant tool against enemy armour.

Maths is flawless, but like you said, flashpoint buggers it all up.  Shame really.  I'm definately going to try out this zero bias and zero precision error thing you've mentioned.  How did you get the shell to fly the way you wanted too?  Do any scripts act on it during flight, or are it's ballistics all set up in the config to provide that kind of stability and consistancy?

And one more point...SEAD mission, that's right!  Suppression of Enemy Air Defence is what Ginge said that stood for...which confuses the hell outta me because we were engaging T-55's.   I guess it's just about keeping their heads down while air power does it's stuff.

[denoir 0]

How did you get the shell to fly the way you wanted too?  Do any scripts act on it during flight, or are it's ballistics all set up in the config to provide that kind of stability and consistancy?

It's ballistics all the way. Once the shell is fired there are no scripts modifying its trajectory. It has however nothing to do with the config. OFP has a physicis engine that is complex but consistent. Since for ballistics it includes drag, it is impossible to describe a general trajecotry with a simple equation. Instaed we used neural networks to solve the problem. Neural networks are non-linear adaptive systems that learn from data. For the artillery we did a data capture of random shell firings and recorded the firing point, the touchdown point, the muzzle velocity, the elevation angle and the difference in heigth.

When we had the data we let the neural nets learn how to map the desired elevation to the other parameters. Here is for instance how one of the neural nets looks like (for a muzzle velocity of 500 m/s):

With the trained network we input the difference in height between the position of the artillery piece and the aimpoint, the distance to the aimpoint and if we want a high angle or low angle trajectory. The network then outputs the elevation that the shell has to be fired at to hit the aimpoint.

[stevevcb 3]

I have absolutely no idea what that is or what it means, but hats off to you anyway

[DeLiltMon 0]

I have absolutely no idea what that is or what it means, but hats off to you anyway  

I understand some of it but the rest makes me want to sit in the corner and flick my finger up and down past my lips going "bibble bibble bibble" and I've got a degree in computers

[Tom Norman 0]

Now that's some brilliant stuff. Neural networks eh? I understand the three inputs and how they combine to produce the elevation, but what are the three hidden layers? What's going on in there? And what's this about 'learning'? Surely it's just plain maths. Does it piece together a pattern for the effects of drag for various ranges and muzzle velocities?

By the way, I think you've been misinformed, high angle is not more precise. Low angle normally results in a greater error for range, mostly because of the lay of the land. But firing in high angle gives the unpredictable variables like pressure, crosswinds and the like more time to act on the shell. We only fire high angle when using carrier shells like illum, smoke and bomblet (although not so much with bomblet). The lower velocity of a falling shell reduces the chance that the parachute will be ripped off in the case of illum. While in the case of bomblet and smoke, the shell's velocity is not transferred to the carried sub-munitions.

On that note, will you be offering bomblet, smoke and illum in the next version?

[denoir 0]

Now that's some brilliant stuff.  Neural networks eh?  I understand the three inputs and how they combine to produce the elevation, but what are the three hidden layers?  What's going on in there?  And what's this about 'learning'?  Surely it's just plain maths.  Does it piece together a pattern for the effects of drag for various ranges and muzzle velocities?

I don't want to hijack this thread by going off into detail explaining how neural networks work, but I'll try to give you the basic idea behind it.

In the picture that I posted in my last post, you can see two elements - the circles that are called neurons and the links between them. A neuron works in  that way that it sums all the signals it recieves, applies a simple function (in this case tanh) and outputs the result. Every link has a weight attached to it, which is simply a number that scales the signal that passes through it. So each neuron in the network recieves a number of signals, scaled by the weights, sums them and applies a function and outputs the result. Ok? A neural network is a structure that has many interconnected neurons as you can see from the picture.

So you input a value at the start of the network and the signal goes through it - through all of the links and neurons and in the end you get one or more outputs.

The "learning" consists in adapting the weights in the network so that it outputs the correct thing. This is done by comparing the output of the network to some value that you want it to be. You then calculate the difference between the desired value and the actual output. That difference is then used to modify the weights on the links in the network in such a way that it minimizes the difference between the desired output and the actual output. In the arty case the desired output is the elevation.

So what learning or training does is adapting the weights inside the network to produce the desired response. Now, the neat thing is that through training the network captures the principles at work from the data. It generalizes. You just have to give it enough examples that capture the model you wish to create and the network will generalize when it operates on previously unseen examples.

A nice property of nerual nets is that they are universal approximators, meaning that they can construct any function with arbitrary precision. It can learn the y=sin(x) function or how to drive a car, it doesn't matter. More complex functions require more neurons, but the operating principles are the same.

You asked about the hidden layers - never mind that. It's only a designation of the layers that are not input or output layers. (A layer is a row of neurons. All neurons in one layer connect to all neurons in the layer in front of it).

Some simple to understand references:

Light Description of Neural Networks

What is a Neural Network?

Crash Introduction to Artificial Neural Networks

Quote[/b] ]By the way, I think you've been misinformed, high angle is not more precise.  Low angle normally results in a greater error for range, mostly because of the lay of the land.  But firing in high angle gives the unpredictable variables like pressure, crosswinds and the like more time to act on the shell.  We only fire high angle when using carrier shells like illum, smoke and bomblet (although not so much with bomblet).  The lower velocity of a falling shell reduces the chance that the parachute will be ripped off in the case of illum.  While in the case of bomblet and smoke, the shell's velocity is not transferred to the carried sub-munitions.

Hehe, I'm talking about OFP. You don't have pressure variations etc The difference in OFP is that at low angle you have a larger velocity parallel to the ground than at high angle. A very small difference in elevation at LA can make you overshoot your target while HA is less sensitive. In real life HA is of course more imprecise because just as you said of amospheric conditions that occur at different altitudes.

Quote[/b] ]

On that note, will you be offering bomblet, smoke and illum in the next version?

Yes, yes and yes.

see: http://www.flashpoint1985.com/cgi-bin....t=39757

[jtec 0]

My brain hurts after all that. Nice work by the way i appreciate the work you it is mind boggling to me