Modelled on the honeybee "house-hunting" decision studied by Seeley et al.: a swarm with no home must pick one of several candidate nest sites, and it does so without any central controller, purely through positive feedback (waggle-dance recruitment) balanced by negative feedback (stop-signalling). Each frame integrates, per site i with quality qi:
discover_i = k_disc · q_i · U
recruit_i = k_dance · q_i · N_i · (U/T)
inhibit_i = k_stop · q_j · N_j · (N_i/T) (j = the other site)
dN_i/dt = discover_i + recruit_i − inhibit_i
dU/dt = −Σ discover_i − Σ recruit_i + Σ inhibit_i
U is the pool of uncommitted scouts, NA and NB the bees currently committed to each site, T the total swarm size (U+NA+NB). Recruitment is autocatalytic (more dancers → more converts → more dancers), which is what lets a small early lead snowball; cross-inhibition is what stops the swarm splitting — the stronger site actively suppresses the weaker one's support. The swarm commits the instant either NA or NB crosses the quorum threshold.
- Dots — individual bees: grey = undecided scouts drifting near the hive, olive = committed to Site A, gold = committed to Site B.
- Graph — running population share at each site over simulated time.
- Quorum — real swarms typically decide around 15–20 quorum-sensing bees packed at a site; here it's expressed as a % of the swarm for clarity.