Multi-Agent System

News & General Discussion
1

I’m currently coding my final year project at University. The basic outline of the project is to create a village simulation with roughly say 700 - 1000 villagers, maybe more.

Currently, I’ve started coding in java, with a grid being populated by different villagers. However, to actually create a village within unity, and have people walking around would look much more impressive.

The villagers would need to use AI to go about their daily lives within the village, i.e. work, sleep, socialise.

The ‘presentation’ can be minimum, I could simply use squares/spheres to represent the villagers.

Therefore, is it possible to:

  1. Create a simulation with 1000+ objects within a scece, but use a lower draw call to render them. I.e. if a script was being run on 1000+ objects at the same time, but the camera wasn’t drawing them, would there be any performace slow down (does performance mainly rely on rendering, and not processing of scripts).

  2. If I chose to use actual models of the villages with a polycount of say 1k per person. Would I be able to render say 100 at any given time, or would a draw call of 100, just destroy a mid-range spec pc?
    (If I chose to use a simple cube, with 6 polys would the same draw call of a 100 be possible, or even the full 1000 draw call, suceeding this would there be headroom, on a mid range spec pc)?

2

  1. Objects not in view of a camera aren’t rendered. (Objects in the view frustum but behind other objects don’t count, and are still rendered.)

  2. If you’re referring to animated models, keep the number of bones to a minimum, and use a skinning weight of 1 bone per vertex, and it should be OK. 100 cubes is no problem; 1000 cubes should be OK, especially if you use vertex lighting only and not pixel lighting.

–Eric