🧮 Modelling a Beehive: Agent-Based Simulation Engine

About this simulation

Agent-based models (ABMs) recreate colony life not by solving one big equation, but by giving each of thousands of individual bees its own simple rules, then letting the hive-level behaviour emerge from their interactions. This scene shows the machinery researchers actually rely on to make that tractable: instanced rendering for the agents themselves, a spatial hash grid for fast neighbour lookups, and a diffusing pheromone field standing in for chemical communication.

How it works

• Spatial hashing — instead of checking every bee against every other bee (an O(n²) blow-up), the ground is divided into a grid of cells sized to the neighbour radius. Each agent only checks the handful of bees sharing its cell and the 8 cells around it.
• Pheromone diffusion — foragers returning from the flower patch deposit scent onto a coarse grid, which is blurred (diffusion) and decayed (evaporation) every frame — a lightweight stand-in for the reaction-diffusion equations used to model real trail and alarm pheromones.
• Genetics — a queen mates with many drones, so a colony is really a patchwork of half-sister "patrilines." Switching the colour mode tints each bee by its patriline instead of its current task, revealing the genetic mosaic underneath the behavioural one.
• Instancing — all bees share one draw call via THREE.InstancedMesh, which is what actually lets a browser animate thousands of agents at interactive frame rates.

Controls

• Agent count — how many bees are simulated at once; watch the FPS and grid-cell counters respond.
• Neighbour radius — the interaction/search radius used for both the spatial grid cell size and the highlighted focus bee's neighbour count.
• Pheromone evaporation — how quickly deposited scent fades from the field between the hive and the flower patch.
• Colour agents by — switch between role-based and patriline-genotype colouring.
• Spatial hash grid — toggle the visible grid overlay used for neighbour queries.

Did you know?

Real ABM studies of honeybee colonies, such as BEEHAVE (Becher et al., 2014), track tens of thousands of individual bees and brood cells across whole seasons — something only possible because of exactly these tricks: spatial partitioning, cheap local rules and coarse diffusion fields standing in for costly continuous chemistry.