Ragdoll Fighting Game

[axcho]

What do you think about making a ragdoll fighing game for the calculator? It could be in assembly, using just linear math (multiplication and some division). The nice thing about using the Verlet method is that it doesn't require trigonometry! We can also avoid square roots if we can tolerate some springiness in the limbs. And collision detection won't be too big of a problem if we can ignore collisions between certain limbs, and only test between enemies when they are close enough (Special bullet-time feature when you get within striking distance of your opponent! ).

If you know of any good line-drawing routines I could use, let me know. I guess the ragdolls could just be thin stick figures. Instead of sound effects when the sticks hit, the screen could shake and maybe have some particle or plasma effects in the background (using cellular automata or something).

Why not? I mean, if we can do N in TI-BASIC, we can do ragdoll physics in z80!

[CoBB]

My intuition says it's not feasible, although it's probably borderline. It's not hard to do in theory, but it's a big question whether it can be done fast enough.

Let's see... In each frame, you have to update every particle (there are at least six per player), check for collisions (at least six items per player -> 36 pairs to check plus the boundaries of the arena), calculate their impact and draw lines (at least five per player plus the heads). So at a mere 10 fps we have 550000 clock cycles to do all this (50000 cycles are needed to update the screen). The coordinates need to be at least 16-bit wide (you need at least 8 bits of subpixel accuracy for Newtonian physics to work at all, and it even might not be enough; I used 12 for Death Rally and 16 for the y coordinate in Acelgoyobis, otherwise gravity just wouldn't work). This means about 1000 clock cycles per multiplication or division. Deducting about 50000 cycles for all the drawing operations we're down to about 500 multiplications and divisions altogether, let's say 450 if we account for additions and subtractions as well. That might be enough.

[alexrudd]

Go for it. I know I can't.

[CoBB]

Well, here's an excellent long-term goal to motivate your learning.

[axcho]

Each ragdoll will ideally have 11 nodes. When they are out of striking distance from each other, there will only be 11 pairs each to check, since I am only going to enforce collision checks between each bone and its parent bone. When they are close enough, we have... bullet-time!

I think there will be 2 division operations per node (1 for length adjustments and 1 for collision detection), for 2*2*11=44 per frame when the ragdolls are far away. When they are close, well, there's a maximum of 44+2*11*11=286 divisions per frame (probably less since I can stop checking collisions after a specified number of hits). For multiplication, I'm getting something like 10 or more per node. It looks like it might be possible...

[CoBB]

[elfprince13]

thats some in intense calculator gaming.

[alexrudd]

[CoBB]

I believed the head to be firmly mounted to the torso. If it is a distinct particle, then there are 12 particles and 11 pairs per player. Since the figure is a tree (graph-wise), it always has n-1 edges (bones) if it consists of n nodes/particles.

[axcho]

In the original Ragdoll Masters, the head is a separate node. This makes a big difference in the controls because your head is what you accelerate with the arrow keys! I'll experiment with various skeletal and control systems before settling on one, of course. I think I'll make a Flash version first.

I just drew a picture of the ragdoll, and there are 11 nodes and 10 bones, plus a circle for the head.

2 for the two feet
2 for the two knees
1 for the hips
2 for the two hands
2 for the two elbows
1 for the shoulders
1 for the head

That adds up to 11 nodes! Each one has a bone attached to it, which in the head's case is actually a circle, which calls for a different collision detection method. It would be a lot easier to have headless ragdolls...

[EDIT]
I've been playing Ragdoll Masters on different settings, and I noticed that there is actually no collision detection within one ragdoll! All the limbs of one ragdoll can overlap. Instead, the joints each have a preferred angle which they settle into. I think that means I can cut down on collision detection even more, though I'm not sure of the best way to implement the 'memory' joints without trigonometry.

I also realized that I don't have to check collisions between ragdolls twice; all it takes would be one sweep through. The sweep would probably be arranged so important collisions are tested first, like those that inflict damage, so it can stop testing once both ragdolls have scored a hit (not much point in testing for blocking if you are already hit somewhere else). It's unlikely that you would be damaging two places at the same instant anyway.

[IAmACalculator]

You're making a clone of Ragdoll Masters? This could be interesting. ASM or basic? Or have you not decided yet?

[CoBB]

[axcho]

This is most definitely going to be in assembly!

[CoBB]

[Brazucs]

Topic splitted from here.

[Liazon]

wow, i'm just confused. axcho, you're just amazing. I hope this works out okay.

[Fr0stbyte124]

I think air resistance would be negligable, and real people don't typically bounce when they fall to the ground, so you can treat gravity and floor landing as special case. Think crumpling.

[axcho]

I was originally planning to have no recoil from the walls (like Ragdoll Masters), but I thought it might be fun to have the option to bounce off the walls like in the animation Rou Bo. Boundary collisions would have to be handled separately anyway, so the default setting would be recoil from your opponent and none from the walls.

Air resistance would be easy to implement, and it would probably be fun to customize as well. It might also be important in preventing your motion from becoming too erratic and fast, though I won't know until I test it.

[Liazon]

[Super Speler]

What would look cool is if you were to have no air resistance and only friction from ground... bouncing off walls and the ground would never be cooler looking.