lilinitsy replied to lilinitsy's topic in Game Development & Technical DiscussionHey, Gaius, nice mod! I skimmed it - it's a lot of JS, which I don't know very well - and it's cool. You should make a display video if you want to show off the differences between your mod and the standard. I'm mostly interested in making the clustering / flocking behaviour without a formation more realistic, disregarding improving other aspects for now. In a strict implementation, I shouldn't have to actually end up modifying the costs of pathfinding though. Many more advanced crowd simulations are either based on a spring-force type model, or a velocity space model, and the path doesn't exactly change. The units will move a little differently, but their goal will be to move along the path, while the crowd simulation aspects (forces, velocity spaces, etc.) will keep them very slightly offset from the path. It should result in a better looking group movement without a formation.
lilinitsy posted a topic in Game Development & Technical DiscussionHello, all. I'm new to 0. A.D. and interested in contributing. Mainly, the crowd simulation and the way that units move by default without a formulation and the way they bunch at one location is quite unrealistic. This can be solved with the pretty heavily researched topic of crowd simulation. I've implemented Boids crowd simulation (a pretty basic one that looks reasonably good) before and would probably like to reimplement it for 0. A.D (https://pdfs.semanticscholar.org/4183/faeef708a56b988742b5572fce9174caec7b.pdf). The codebase is pretty massive, and I've noticed that there is a lot of somewhat relevant AI code in source/simulation2/components. I'm mostly wondering if anyone can point me more specifically to where I'd look to begin modifying how units move and adding some better crowd simulation.