Roach Crowd Sim

Rough concept

Similar to boids except that the cockroachs' bodies are parented to an empty/root bone which is shrinkwrapped to a low poly mesh (offset some distance from the floor) which marks the area the roach's are constrained to.

The system uses empties to input data. Custom properties define whether an empty represents an aligner, a control or a boid (roach). Controls (which can be of type 'follow' 'attract' 'repel' 'herd' 'panic' 'sleep' etc) are empties which affect boids within a certain radius during the simulation time. Proximity data is refreshed on a frame by frame basis. Aligners are used during setup to quickly orientate and configure a large number of boids using a weighted linear fall off to interpolate starting direction and speed between multiple aligners.

The roachs have an emotional state which is adjusted each frame by various influences (including how crowded the flock is, whether they have been involved in a collision, if they have strayed from the flock, or if they are under the influence of a controller). Emotional states decay/normalize over time.

Optimizations

Python is incredibly slow. With 500 boids there are 500x499 inter-boid distances (or separations) to measure. These are stored in a half-matrix (1/2 n (n + 1)) indexed by [boid with lower index of pair][boid with higher index of pair]. The separations are also indexed in another dict by their value ie dict[separation distance]=[index a][index b]. Both sortings of the separations are required so although the script is basically holding duplicate data, this will be faster than reindexing the half-matrix by separation size each time we need to find the pair which is closest together, or a list of all pairs which have collided.

Having to