Dragon physics

With the switch to the new customizable dragon model, we're also taking the opportunity to explore making the dragon fully physics-based, something that both me and our engineer Georges (currently working on the battle system) have extensive experience with. So what's the difference between that and the way it's currently done?

Some of you may think that the current dragon is physics-based because of how it moves, but it's actually not (at least not in the usual sense of the word). All the movement and dynamics are 'fake' in that it's mostly visual, with no connection to the physics engine. There are no forces involved, no collisions, no way for the body to impact the surrounding environment using physics, or be affected by it in return. All the 'physics' is happening in the context of the animation itself, with no bearing on the rest of the world. There's nothing stopping a limb going through a wall for example. You can still fake these interactions to a certain extent, but it will never match the fidelity of a real simulation.

With the physics-based dragon, it's instead a bit like creating a virtual dragon robot with muscles. If the foot tries to go through a wall, it will stop in place and exert a force on the wall from the leg muscles pushing (potentially breaking through the wall), instead of just clipping through it. Or if you throw an object at the dragon, it will create a realistic impact that displaces the hit body part and propagates through the rest of the body. If the tail swings back and forth, it would slap against any obstacles in the way and bend around them if possible, you get the idea.

There are a few downsides to this approach though (which is why I didn't use it initially, despite having done it before even for dragons). It can be time consuming to implement and hard to make it not look all bouncy and goofy, body parts can sometimes clip though objects, get stuck or end up in other undesirable states that are hard to detect and recover from, things can generally get unpredictable and whacky when lots of forces are involved, and the performance can suffer if you're not careful. On the other hand, when it does work well it provides an unmatched level of interactivity and dynamic gameplay.

Currently this is mostly experimental to see how well we can get it working, but so far the results are looking very cool. I'll be exploring this further next week and maybe start working on some accompanying ground movement too.