Hideous Destructor 4.10.0b

Projects that have specifically been abandoned or considered "dead" get moved here, so people will quit bumping them. If your project has wound up here and it should not be, contact a moderator to have it moved back to the land of the living.
Fallingferret
Posts: 37
Joined: Sat Jan 14, 2017 11:53 pm

Re: Hideous Destructor

Post by Fallingferret »

it works with the folder inside just the same. it just needs to be a pk3 file
AldenBlueten
Posts: 7
Joined: Sun Sep 02, 2018 10:41 pm

Re: Hideous Destructor

Post by AldenBlueten »

Ah. Okay. Seems like DoomLauncher is slightly buggy with zips. Atleast I now know about renaming the file to a pk3
Gideon020
Posts: 558
Joined: Mon Sep 28, 2015 3:23 am

Re: Hideous Destructor

Post by Gideon020 »

Man, I keep forgetting that I need to strip off armor when I get hit by a fireball. Keep winding up with 153% burns that not even a med-kit can fix. :D
Mikro
Posts: 8
Joined: Fri Dec 15, 2017 5:31 am

Re: Hideous Destructor

Post by Mikro »

@Gideon020 dude, I got up to like 280ish the other day. luckily I found a soulsphere in the level :P
User avatar
Matt
Posts: 9696
Joined: Sun Jan 04, 2004 5:37 pm
Preferred Pronouns: They/Them
Operating System Version (Optional): Debian Bullseye
Location: Gotham City SAR, Wyld-Lands of the Lotus People, Dominionist PetroConfederacy of Saudi Canadia
Contact:

Re: Hideous Destructor

Post by Matt »

User avatar
mrtaterz
Posts: 236
Joined: Wed Feb 14, 2018 8:47 pm
Graphics Processor: nVidia with Vulkan support
Contact:

Re: Hideous Destructor

Post by mrtaterz »

Would it be possible to add in Stealth features like suppressors (attachments/items you pick up for the SMG and pistol maybe) and silent takedowns with melee at some point? A stealth playthrough of this game would be great.
User avatar
Matt
Posts: 9696
Joined: Sun Jan 04, 2004 5:37 pm
Preferred Pronouns: They/Them
Operating System Version (Optional): Debian Bullseye
Location: Gotham City SAR, Wyld-Lands of the Lotus People, Dominionist PetroConfederacy of Saudi Canadia
Contact:

Re: Hideous Destructor

Post by Matt »

Cover fire (intimidation) and exploiting an opponent's aggression I understand the basics of, but the dark arts of how not to be seen are a total black box for me.

The hardest things to model are camouflage and messing with expectations - trying to ambush a player in deathmatch makes sense to me since you know where people tend not to look (through narrow windows, around certain corners, corners opposite from where the big weapons or powerups are, areas that don't flow very well so people tend not to move through them), but I don't even know where to begin coding anything remotely not-completely-insane that would have a similar heuristic.

(Incidentally, I am completely unable to play stealth games against AI at all, except in some RPGs where I draw aggro from one member of a group at a time in what amounts to a blatant AI glitch exploit.)
mumblemumble
Posts: 927
Joined: Fri Aug 23, 2013 1:59 pm

Re: Hideous Destructor

Post by mumblemumble »

Honestly, camo is taking it much too far, doom is almost entirely interior city / techbase with very little camo, so I don't think its all too important, compared to LOS handling peripheral vision handling, and detail levels compared to distance / movement / cover.

Metal gear solid is a good example, even sons of liberty had a highly effective setup, and even if you didn't include enemies sweaping areas, I believe the following could work, just as an idea.

Give each enemy at least 2, possibly 3, vision "cones". First is very pointed, aimed, second is at a 90 degree cone typical of MGS. 3rd if added, would be like second, but more severe. I suppose instead of cones it could just be a degree check outside vision, but then I imagine vision would need to be rendered more. Perhaps instead of just a cardinal direction, also rendered vertical vision.

For calculations of visions, take into account, if possible, following factors. Which vision cone its in, if target is moving, if observer is moving, distance, and, if possible, amount of cover (i doubt this could ever detect things like vines covering windows, but solid architecture would be a start.) and lighting, depending on the monster type.

How I imagine it could work for a basic premise, is give every technically visible unit a "point value" based on conditions. Being up close, exposed fully, full lighting, and looked directly at, you would have multiple negative "stealth" scores quickly adding up, to the point its almost entirely impossible to hide, where it immediately goes to full on detection because you stick out like an idiot.

However let's say you are in an alcove (map02 waterway alcove) and a shotgun guy walks past down the hall. Assuming you are crouched (reduced profile) still (blend into his motion blur of peripheral vision) hes moving (peripheral vision is further hampered) its very likely, even if in his peripheral vision, that you would be missed, even at very close range.

At long range, obviously not everyone has eagle vision, especially in peripherals, so unless looking DIRECTLY at the player, or there's significant movement away from cover, I figure one could be somewhat undetected over 1024 units away. Movement, direct vision, and obviously noise would negate this, but it could be possible, weighing distance vs all other factors. You could make this only calculate if player is WITHIN Vanilla vision so as not to completely lag it up, so its only maybe a dozen calculations at a time, and perhaps once enemies get positive I'd, drop the calculation until LOS is broken.

Honestly, bigger than stealth, would be a patrolling AI, how the AI would find a player that it suspects is there, or has recently escaped. But I imagine you could do this if you added waypoints for "player last seen here" or "shots fired in earshot". When either happens, you could have enemies attempt to navigate to the point, and investigate within say, 200 units, generally, from where it happened, for a brief point, and on edge, before relaxing somewhat, and going back to an idle wander / scan los.

.... Really, line of sight, and predicting players that aren't visible are the 2 main things needed for stealth. Its ironic you mention the video about not being seen, because that's really what it boils down to for stealth. Even in non traditionally stealth games, like tf2, stealth can be used by anyone for a huge advantage. I used to play scout And would intentionally go when enemies were distracted to attack. This may seem like just ambush, but ambush is very much based on stealth, people good at ambush are usually good at stealth, and stealthy people are usually good at ambush, at least within such game mediums.

So I'd focus on something of a more realistic LOS system, accounting for distance, cover, lighting and movement, as well as peripheral vision issues, and just see how that alone functions as a means of stealth, you might be surprised.

https://en.m.wikipedia.org/wiki/Peripheral_vision

Its important to note with peripheral vision, that central vision is extremely high resolution, and detailed, while peripheral vision is incredibly low resolution, but sensitive to movement. So it's kinda like the center of view is high Definition, semi peripheral view is lower definition, but able to make out shapes, colors, ect, and extreme peripheral view (the 180 degree scan on vanilla doom monsters) is incredibly low definition, but picks up on movement particularly well, especially if the movement is the odd one out (peripheral vision doesn't work well while running for instance, because everything seems to be moving, so smaller movements are lost in the signal noise).

Always, who knows if this is helpful, but I really would focus on modeling a sort of vision first, because stealth by and large is determined visually, and while camo would be cool, you need to start with more fundamentals, like size of object, movement of object, distance of object, and where the object is in peripheral view. Sure, colors, textures, all play factors, but I think peripheral vision emulation would be easier to do for now.

Edit : actually I'd highly encourage messing with wolfenstien 3d for this reason, camo aside, the low resolution is a good emulation of stealth. Ever confuse a potted plant with a brown shirt in earlier levels? This is intentional, because at low resolution, a brown shirt and plant look identical, and distance effectively reduces resolution along with peripheral view, and with no movement, it's even worse. Creatures are very visually adapted to detection of movement, as movement indicates threat or prey. But movement mixed in with other movement (the viewer walking, or a busy environment) gets lost in all of the information unless you have specific information scanned for. In peripheral view, moving cars, machines, other people, and an enemy all look very similar, and central view is constantly moving to scan and collect information, but you can only see 1 spot at once.

In the end, its like you take any visual, and as you scale the distance, you scale the resolution back, and as you go into peripherals, you also scale resolution back. So you can kinda use this as a bench mark if you think about it, how far away x how far into peripheral = general stealth, plus lighting, plus any observed movement.
User avatar
Matt
Posts: 9696
Joined: Sun Jan 04, 2004 5:37 pm
Preferred Pronouns: They/Them
Operating System Version (Optional): Debian Bullseye
Location: Gotham City SAR, Wyld-Lands of the Lotus People, Dominionist PetroConfederacy of Saudi Canadia
Contact:

Re: Hideous Destructor

Post by Matt »

Just a word on camouflage: I've been playing some Freedoom lately with the new shotgunguy and zombie, and I must say they are much better camouflaged than their Doom equivalents. The greebled blobby greys make them stand out against the backgrounds a lot less - not enough that you'd really think anything of it just looking at them, but enough to give them enough of a split second to get the drop on you.

Of course, it could just be me not being as intuitively used to the sight of them as I am to the Doom monsters, but there's still at least definitely a difference between the SPOS guys and the PLAY/CPOS guys.

The problem with all this is, of course, that one would have to:
1. Check the angle and sprite of the target relative to the actor
2. Analyze the colours and edges of the sprite displayed by the target and then everything around it
3. Subject the resulting edge detection to a heuristic machine learning (which may be pre-primed) to figure out if it's recognized


As for looking for the player using vision cones, depending on how frequently such a check would be made I don't know how much it can bog down the system since we're doing all this in a virtual machine for what may be hundreds of actors in a single tick.

Personally I think a more advanced sight system is way more trouble than it's worth in Doom. The AI opfor is literally demonic which is a perfectly reasonable explanation of them having hidden spiritual "insight" that would let them track you the way they do ingame.
mumblemumble
Posts: 927
Joined: Fri Aug 23, 2013 1:59 pm

Re: Hideous Destructor

Post by mumblemumble »

Actually I think It could be relatively bog free if it used the following system.

If enemy has no vanilla los, do nothing.

If enemy has vanilla los, check for angle outside forward vision. If not within "possible success", do nothing.

If possible, factor in lighting. If impossible, do nothing.

If possible, check target percentage out of cover. If impossible, do nothing.

If possible, engage in combat.

That's 4 checks, most of the time not even being fired for enemies that can't see the player (since it requires vanilla los to even activate). You could also give a secondary state between oblivious enemies, and attacking for "what was that", and have them cautiously look in that direction / examine more closely for a few seconds, so enemies can suspect players, but not see them explicitly.

I know it would obviously add more calculations, but I'm unsure it would make a big impact when we already have dozens of checks every second for heart rate, bleeding, fire, ai, ect.

And I suppose some demon insight would exist for some (knights, limited range detection on higher rank imps, cacos, archviles, ect) but I figure zombies, Mancubi, and others wouldn't exactly be magically adept to detect so well. Especially Mancubi, put together with spare organs. Can't imagine second hand eyes would handle great.
User avatar
ZikShadow
Posts: 564
Joined: Wed Jul 13, 2016 1:43 am
Location: Could be the chair, the floor, or the bed.

Re: Hideous Destructor

Post by ZikShadow »

I'm pretty sure that when it comes to features like that, if it was that simple to implement, it would've been already done. Anyway, there's always Sneaky Doom for the stealth folks, no need to bleed them together with HD.
User avatar
Caligari87
Admin
Posts: 6174
Joined: Thu Feb 26, 2004 3:02 pm
Preferred Pronouns: He/Him
Contact:

Re: Hideous Destructor

Post by Caligari87 »

(crosspost for anyone not on Discord)



8-)
User avatar
Leon_Portier
Posts: 62
Joined: Sun Feb 12, 2017 3:30 pm
Location: Dark side of Bavaria

Re: Hideous Destructor

Post by Leon_Portier »

Working on the bootcamp map, there is something up in the enemy derp when spawned with the ednums. When using

Code: Select all

23016=enemyderp
the error sprite is in its place when using HD 3.5.1 but when using HD 3.5.0 the enemy derp spawns nicely!
The ednum in both releases stays the same.
User avatar
Leon_Portier
Posts: 62
Joined: Sun Feb 12, 2017 3:30 pm
Location: Dark side of Bavaria

Re: Hideous Destructor

Post by Leon_Portier »

Leon_Portier wrote:Working on the bootcamp map, there is something up in the enemy derp when spawned with the ednums. When using

Code: Select all

23016=enemyderp
the error sprite is in its place when using HD 3.5.1 but when using HD 3.5.0 the enemy derp spawns nicely!
The ednum in both releases stays the same.
Update on the upper:
I noticed missing sprites in the console.
User avatar
Matt
Posts: 9696
Joined: Sun Jan 04, 2004 5:37 pm
Preferred Pronouns: They/Them
Operating System Version (Optional): Debian Bullseye
Location: Gotham City SAR, Wyld-Lands of the Lotus People, Dominionist PetroConfederacy of Saudi Canadia
Contact:

Re: Hideous Destructor

Post by Matt »

There's a typo in the DERP spawn state's very first frame. I would have thought this would just show nothing for zero ticks before moving on to the next (correctly typed) frame, but apparently now(?) GZDoom treats it as having no starting frame at all.

No idea why it would have worked in 3.5.0 though. That typo was always there, but only became the first spawn frame in the newest release during the HDUPK overhaul.

Will fix tonight.
Locked

Return to “Abandoned/Dead Projects”