The Daily Click ::. Forums ::. Klik Coding Help ::. Need some help following an article.
 

Post Reply  Post Oekaki 
 

Posted By Message

Sketchy

Cornwall UK

Registered
  06/11/2004
Points
  1971

VIP MemberWeekly Picture Me This Round 43 Winner!Weekly Picture Me This Round 47 WinnerPicture Me This Round 49 Winner!
6th January, 2009 at 22:24:59 -

Thanks
That explanation might be worthy of being posted in the articles section?

Here's the new and slightly improved example file:
http://cid-b1e7ee094271bbda.skydrive.live.com/self.aspx/Public/LoSeg.mfa

There's one minor flaw that I'm aware of - let's see how long it takes you to figure it out...

Edited by Sketchy

 
n/a

-J-



Registered
  05/10/2008
Points
  228

VIP MemberThe Cake is a Lie
6th January, 2009 at 23:48:43 -

Thanks I was thinking of writing an article on a couple of things (that included) which reminds me, (this is slightly off topic) but do you have any idea on how to get vitalize working within an article?

Edited by -J-

 
n/a ...

Deleted User
7th January, 2009 at 00:18:05 -

@sketchy: thanks alot man, I'll take a look at it tonight.

@julian: That's very useful to know actually. I figured it was either something like that or if you had multiple commands for each or statement such as:

if player does such and such
and such
or
this

if player does such and such
and such
or
if player does such and such
and this

 

-J-



Registered
  05/10/2008
Points
  228

VIP MemberThe Cake is a Lie
7th January, 2009 at 01:21:19 -

Just testing something. Word.

Edited by -J-

 
n/a ...

Deleted User
7th January, 2009 at 06:32:34 -

I don't know if I'm retarded or what but I've done everything exactly as the example shows, but it just gets all jacked up. It's the whole angle part, for whatever reason it calculates stupid, it's not even close. I can't really even describe it. Does the fact that I'm using 32 directions count for anything? I hope that's all it is. I'll give it a shot tommorrow. Thanks anyways for your help guys.

 

-J-



Registered
  05/10/2008
Points
  228

VIP MemberThe Cake is a Lie
7th January, 2009 at 10:51:36 -

Using 32 directions and then setting the "direction" of the object through the event editor won't give you a 360 degree angle number. Instead of using "set direction to", under the Direction tab, go down to the Scale/Angle tab and choose "set angle" and then enter the angle calculations, and then 0 or 1 depending on whether you want speed or quality.

Not sure if that helps but I hope it does

 
n/a ...

Sketchy

Cornwall UK

Registered
  06/11/2004
Points
  1971

VIP MemberWeekly Picture Me This Round 43 Winner!Weekly Picture Me This Round 47 WinnerPicture Me This Round 49 Winner!
7th January, 2009 at 13:06:57 -

Does my example work for you?
If it does, and it's only your implementation that doesn't, then I'd probably need to see your file to be able to help.

My example uses only angles (not MMFs "directions").
If for some reason, you only want 32-directions, it's easy to convert between angles and directions:

Angle = Direction * 11.25

Direction = Angle / 11.25 (kind of obviously)

 
n/a

Deleted User
7th January, 2009 at 18:43:53 -

Nevermind, I've got it. Using the advanced math objects convert directions to angle function, I set the current angle to always be that conversion of the enemies direction. I also needed to tweak the distance calculation a bit, but now it seems to work fine.

For the example though, during each LoS check, I would add "turn flag 0 off" anytime you "deactivate group line of sight" so that way the enemy will stop blinking when you walk through the area. Also, I don't know if you need to, but the last line, "if flag 0 is on" I added the ID check before that, rotations using the direction calculator and "activate group line of sight". I don't know if the ID check or the group activation is necessarry. I'll test it though. I'm sure you know this already but just in case someone stumbles across this thread and goes "gdam it what ter hell!>?lol"

Once again, thanks alot for your help guys, you have no idea how exciting it is to actually have a project going that looks and feels very professional. I'll be sure to credit both of you when I complete this game.

 

Sketchy

Cornwall UK

Registered
  06/11/2004
Points
  1971

VIP MemberWeekly Picture Me This Round 43 Winner!Weekly Picture Me This Round 47 WinnerPicture Me This Round 49 Winner!
7th January, 2009 at 21:36:02 -

Okay, I'll come clean

After further experimentation, it may not be quite as stable as I'd have liked

Yes, you should turn flag0 off, when the player isn't visible. Have you tried it though? I couldn't make it work. To be honest, I've been trying to improve it, and small changes to completely unrelated events seem to totally screw everything up. Good luck though

Using extensions is a good idea. I couldn't get my example to run at all with HWA, until I replaced the angle formula with an extension (weird, cos it worked fine with the normal build).

I think "machine independant speed" is essential. I get the impression it goes wrong if it can't maintain the proper framerate.

There's no need to add "if flag 0 is on" or "activate group line of sight", on the last line - I guess it won't do any harm either though.

The way it works, it only checks the line of sight for one enemy each frame. This means that with 10 enemies and a frame rate of 50fps, it can take up to 0.2 seconds for the player to be detected after they enter the enemy's line-of-sight (0.1 secs on average).

You *could* make it check every enemy, every frame, but there are two main problems with that;

1.) 10x as many enemies take 10x as long to calculate LoS for. It could easily end up slowing down the whole game if you have enough enemies (I suspect that would make it pretty unstable too).

2.) It's a humongous pain in the arse to code. It would require nested fastloops (fastloops inside fastloops inside fastloops). I spent quite a while trying, but it's beyond me. Don't let that stop you trying though.

I'm quite impressed with it myself actually - I made some of the enemies move around a bit and rotate, and it was almost like a game in itself

I don't know what kind of a game you're making, but you should make it so you can crawl - increasing the minimum range at which you're detected, but making you move slower and unable to perform actions until you stand up. All it would take is 1 extra event and some more frames of animation.

Incidentally, I found pixelthiefs example on my harddrive - couldn't make it work - just kept crashing (again, weird, cos it always used to work).


Edited by Sketchy

 
n/a

Deleted User
7th January, 2009 at 22:28:15 -

Well, I messed with it a bit more, and instead of having the enemy look directly at the player, I've created a target object, that I will most likely replace with a an xy val similar to the detection process. So when the player is in the line of sight, a target object is either placed on the player or created based on if there is one or not. If the enemy is in attack mode when it reaches the target, the target gets destroyed, runing a looking script. If the enemy is not, the target simply goes to the next spot that is currently just a random value around the area of the enemy. Also, instead of telling the flag to turn off during the line of sight checked, I did one of those follow counters.

I also added another event that if you are within say 50 pixels of the enemy and the player speed is greater than 15, then the enemy will turn around. I might make another for the longer distance, but I'm going to have to revamp the code alot.

The biggest problem I'm seeing is that, like you said, there is a lag in the calculations, several times the enemy just kind of forgets what he's doing and starts going "OOH SHINY OBJECT" and stops following the player. While maybe that might be hilarious to have as a lower difficulty style enemy, I'd like to work with it to find a better solution. Also I can see with the visual target object that the target detection is going much slower than the gameplay itself.

Unfortunatly the way it looks is that I'm going to have to get rid of the spread values and make it specific to each enemy, replace the calculations with extentions and see how many enemys I can ultimatly have on screen before it makes a noticable difference. So about 2. Haha.

If you can get pixeltheif's system to work, that'd be awesome, I have a node detection example of his that I'm going to try to work into it somehow.

As far as the style of game I'm making, maybe when I finish up the system a bit more and get something other than test gfx I'll upload it to the projects section.

 

Sketchy

Cornwall UK

Registered
  06/11/2004
Points
  1971

VIP MemberWeekly Picture Me This Round 43 Winner!Weekly Picture Me This Round 47 WinnerPicture Me This Round 49 Winner!
7th January, 2009 at 23:14:27 -

To be honest, I didn't really understand most of that (or any of the first paragraph infact)

"I also added another event that if you are within say 50 pixels of the enemy and the player speed is greater than 15, then the enemy will turn around. I might make another for the longer distance, but I'm going to have to revamp the code alot."

That part I get. It's like the enemy would be able to hear you running, right?
There are lots of little things like that, which you could add very easily for extra realism. Crawling was one such idea. Also, the enemy could have a narrower field-of-vision when the player is not moving - it's much easier to notice things in your peripheral vision if they're moving. Other ideas taken from "Commandos" - have the player leave tracks in snow/sand; make the player able to move and hide bodies.

I'm not sure why the lag should be a problem. Every 1/10th of a second is pretty often. If the enemies are getting distracted too easily, can't you just put in some kind of a delay before they stop following?

"Unfortunatly the way it looks is that I'm going to have to get rid of the spread values and make it specific to each enemy, replace the calculations with extentions and see how many enemys I can ultimatly have on screen before it makes a noticable difference. So about 2. Haha."

btw: It turned out, the reason it wasn't working in the HWA build of MMF, is that the latest "regular" build is actually more recent - HWA just doesn't have the ATan2 function yet



TRY THIS: http://cid-b1e7ee094271bbda.skydrive.live.com/self.aspx/Public/LoSeg3.mfa

Edited by Sketchy

 
n/a

Deleted User
8th January, 2009 at 01:37:01 -

Oh well basically with the target object, it's just a separate object that acts as the enemy's current goal. Right now it just kind of goes around randomly and the enemy chases it, I want to make up a kind of network of active objects that the target moves to, then when the enemy reaches that target it moves on to the next one. But it sees the player, the target moves to the player and should stay there untill the player is out of the line of sight, and if the player gets far enough away, the enemy will run into that target, stop, look around, then go do whatever. It's basically a precursor to more complicated tasks for the enemies, like reporting in, taking a break, whatever I think of in the future, and also most importantly, taking the players to jail.

This build looks alot better. I'll see if I can implement most of it into my system.

 

Sketchy

Cornwall UK

Registered
  06/11/2004
Points
  1971

VIP MemberWeekly Picture Me This Round 43 Winner!Weekly Picture Me This Round 47 WinnerPicture Me This Round 49 Winner!
8th January, 2009 at 18:06:13 -

Yay!
The definitive version;
http://cid-b1e7ee094271bbda.skydrive.live.com/self.aspx/Public/myLoS.mfa

I finally fixed all the bugs - it's now rock solid, and more efficient too.
It comfortably handled 20 on screen enemies (didn't try with more than that), and that's on my old 700mhz pc!

I don't remember everything I changed, but I know I added lots of "flag 0 is on" conditions all over the place, so it doesn't go through the whole rest of the process once it's already found a LoS impossible.

Also, and this one took me FOREVER to figure out - "CurrentHeading" mustn't get too big. Basically any time you change it, add "mod 360" to the end and it'll be ok.

eg;
to rotate...
Always -> Set CurrentHeading to (CurrntHeading + 1) mod 360

Edited by Sketchy

 
n/a

Deleted User
8th January, 2009 at 19:02:47 -

Top notch my friend. Thank you for all your help. You should upload it to the download section for the rest of the other chitlins to look at.

 

Sketchy

Cornwall UK

Registered
  06/11/2004
Points
  1971

VIP MemberWeekly Picture Me This Round 43 Winner!Weekly Picture Me This Round 47 WinnerPicture Me This Round 49 Winner!
8th January, 2009 at 19:14:05 -

Thanks. I probably will.

I've done some more thorough testing - on my pc (700mhz like I said before), it can maintain 50fps with up to about 275 enemy objects on screen. That's with no "cone-of vision" objects, and with the enemies just slowly rotating. (now I'm just boasting )

Still, I reckon if the rest of your game is well coded, I don't think the LoS part is going to slow it down much.

Better solution to "CurrentHeading" problem - add a line at the beginning of the Line of sight group:
always -> set "CurrentHeading" to "CurrentHeading" mod 360
much simpler, and definitely won't mess anything else up

Edited by Sketchy

 
n/a
   

Post Reply



 



Advertisement

Worth A Click