Hi, I was just wondering if anyone has messed around with different collision methods. I saw some mention of ideas for hitboxes on this board not long ago. Has anyone come up with a good way of getting a character to respond differently if he's hit in different areas of his body? For instance, in close combat? I've been trying to add boxes to the fists and feet of my character using ent_vertex, but am getting lag despite all efforts and suggestions (thanks a million to everyone who's been helping me with that in that topic), so I was just wondering if there are any other simpler or better methods that people have already used?
Thanks for any suggestions,
Ambit

Okay, this is my idea for a fighting collision system; I'm not sure it's possible though (and can't mess around with it for a fortnight as I'm going on holiday): I model a very basic, very low polygon squarish basic human model, then take note of all the vertices (which are kept to a minimum). This model can then be stretched to match a similar shape to all the models in the game, and then have the skeletons of the different animated models assigned to it (I use MilkShape) so that it can match the size and animation of each fighting model in the game. The differently stretched and animated blocky models (which all share the same vertex numbers) are then attached to the different character models at their origins, but hidden (will that affect rendering time? It shouldn't if it's hidden, right?). Then I hope to be able to use the ent_vertex() command to calculate if certain vertex numbers are near other vertex numbers, or between the positions of other vertices, during certain (eg punch, kick etc) animations (ie by having an invisible identical blocky character attached to each character, you don't have to work out the vertex numbers of each complicated model), and the character can be animated and damaged etc accordingly.

Sorry, that is long winded and sounds incredibly complicated, but does that sound like it might work to any experts or experienced 3dgs users out there? Like I say, I can't test this theory out properly for a coupla weeks.

Cheers,
Ambit

you could try a cheap method. just have a variable saying whether the character is ducking, standing, jumping, high attack, low attatck, etc. Then if the models collide, (or just have a distance variable saying if within x distance, and player is performing y maneuver), just run hitmyhead_if_crouched animation if player1 was doing a low attack and player 2 was doing a crouching, or run falldown animation if player 1 was doing a low attack and player 2 was standing; etc.

[This message has been edited by hedshrinker (edited 31 March 2001).]

Hmm... that sounds like a lot less work and like it makes a lot more sense! How accurate would the collisions be, though? When a fist hits a head for instance, wouldn't it really just be checking the bounding box of both models, and therefore not look as accurate?

Thanks for the help,
Ambit

I think hedshrinker means that you shoiuld simply ignore, or combine the collisiondetection with simply measuring the distances between different entities in their different states.
This way u can achieve "pixel perfect" results.

Cool, thanks. So the distance variable would be changed with each move (ie punch will be closer than kick)? That sounds like it should work, provided the characters are facing each other (and my intention is for the player to auto-face and circle each opponent).

Hi, sorry to bring this up again, but I've been away for a fortnight... Could someone please clarify this - ie the best way of making "pixel perfect" fighting collisions? I'm not exactly clear on how this "cheap" method works... Thanks for all the help.

Measure the distance from a vertex on fighterA's hand to a vertex on fighterB's head, and when the distance equals < 5 (or whatever you choose) make a hit.

I think that is what they are getting at

------------------
"Dare to Do, what others Dare to Dream..."

Currently Recruiting for a Commercial CRPG Project based on a Published Pen & Paper Game:
www.talislanta.com
www.tamerlin.com
www.talislanta.com/talislanta/images/talpromo-56.ram
See our thread in 'User Projects' Forum

Thanks again, ASallade. But to do this you would need to know the vertex numbers on all the parts of all the models, wouldn't you? ie having variables for the face vertices, chest vertices, hand vertices etc for each different character model. Which is why I thought of attaching the dummy model to all... And which in retrospect seems a silly way of going about it. I thought they were suggesting a way of just using distance from origin - but there would be no way of getting "pixel perfect" or accurate collisions between fist and face for that, would there? As always, any suggestions or pointers to my stupidity are welcome and appreciated...

You would need to keep the Vertice Numbers of all the targets you want to track in a text file, and load them with the model.
You could probably have just a few and do fine...

head, chest, stomach, groin, left upper arm, left lower arm, left hand, right upper arm, right lower arm, right hand, right thigh, right knee, right shin right foot, left thigh, left knee, left shin, left foot.

Its a big project your looking at, coding for realtime Martial arts is never easy.

Best of luck!

------------------
"Dare to Do, what others Dare to Dream..."

Currently Recruiting for a Commercial CRPG Project based on a Published Pen & Paper Game:
www.talislanta.com
www.tamerlin.com
www.talislanta.com/talislanta/images/talpromo-56.ram
See our thread in 'User Projects' Forum