Up to 250 colourful birds flock together following Craig Reynolds' three rules of Boids: separation (avoid crowding), alignment (match neighbours' direction) and cohesion (move toward the flock's centre). Hit scatter to spook them!
Emergent group behaviour from three local rules. No bird knows the global pattern — each reacts only to its nearest neighbours.
Adjust separation, alignment and cohesion strengths. Hit Scatter to frighten the flock. Release the hawk to see predator-avoidance behaviour.
Reynolds published his Boids algorithm in 1987. It was so convincing that it was used in the bat swarm in Tim Burton's Batman Returns (1992).
This simulation models collective bird movement using Craig Reynolds' Boids algorithm, which governs each bird's behaviour through three simple rules: separation (avoid crowding neighbours), alignment (steer toward the average heading of nearby birds), and cohesion (move toward the average position of the flock). Despite no central coordination, these local rules produce the sweeping, fluid group formations known as murmurations. Users can observe how emergent complexity arises from simple individual behaviours by adjusting cohesion strength, flock size, and flight speed.
Bird flocking behaviour has fascinated scientists and artists alike since Craig Reynolds published the Boids algorithm in 1987. The model has been used in film visual effects, robotics swarm design, and the study of collective animal intelligence across species from fish to insects.
Boids is a computational model created by Craig Reynolds in 1987 to simulate the flocking behaviour of birds. Each simulated bird (called a "boid") follows three steering rules based only on the positions and velocities of nearby boids: separation, alignment, and cohesion. The result is lifelike emergent group movement without any centralised control or leader.
Use the Birds slider to set the flock size (20 to 250 birds), the Cohesion slider to control how strongly birds pull toward the flock's centre, and the Speed slider to adjust overall flight pace. Click Scatter to startle the flock into chaos, Hawk to release a predator that birds flee from, and Regroup to steer all birds back toward the screen centre.
The separation rule keeps each bird a minimum distance from its neighbours. Whenever a bird detects another within its separation radius (set to 25 pixels in this simulation), it applies a steering force directly away from that neighbour. This reactive repulsion prevents collisions even at high speeds, mirroring how real birds use peripheral vision and rapid reflexes to maintain safe spacing within a fast-moving flock.
Each rule produces a steering vector that is added to the bird's current velocity. Separation computes a weighted sum of inverse-distance vectors pointing away from nearby birds. Alignment calculates the average velocity of neighbours within the alignment radius and nudges the bird toward that heading. Cohesion finds the centroid of nearby birds and steers toward it. All three force vectors are clamped to a maximum magnitude each frame to prevent unrealistic acceleration. The combined update is: new_velocity = clamp(v + w_sep * F_sep + w_align * F_align + w_coh * F_coh, max_speed).
Reynolds' Boids algorithm was first used in film in Tim Burton's Batman Returns (1992) to animate bat swarms and penguin marches without hand-animating each character. It is now standard in video game crowd AI, robot swarm coordination (such as drone light shows with hundreds of autonomous aircraft), traffic flow modelling, and even the design of self-organising sensor networks where individual nodes follow local rules to achieve global coverage.
Biological research largely supports the Boids framework, though real starling murmurations are governed by topological rather than metric neighbourhoods — each bird tracks its six or seven nearest neighbours by count rather than by a fixed distance radius. Experiments using high-speed 3D cameras on starling flocks in Rome (Ballerini et al., 2008) confirmed this topological interaction, which makes murmurations more robust to changes in flock density than metric Boids would predict.
Craig Reynolds, a computer graphics researcher, formally modelled flocking in his landmark 1987 SIGGRAPH paper "Flocks, Herds, and Schools: A Distributed Behavioral Model." However, the natural phenomenon itself has been observed and studied by ornithologists for centuries. Reynolds' breakthrough was demonstrating that three local interaction rules were sufficient to reproduce the complex emergent patterns seen in real animal groups, without any top-down coordination.
The same self-organisation principles appear in fish schooling, insect swarms (locusts, bees), pedestrian crowd dynamics, and even the collective motion of cells during wound healing and embryo development. All share the property of emergent order from local rules. On this site, the Butterflies and Flowers simulation shows related attraction-based movement, while the Forest Fire simulation demonstrates a different class of emergent spatial pattern via contagion rather than steering.
Swarm robotics uses Boids-derived algorithms to coordinate dozens or hundreds of autonomous robots without a central controller. Applications include search-and-rescue drones that collectively cover large areas, agricultural drone fleets for crop monitoring, and underwater vehicle swarms for ocean mapping. The key engineering advantage is resilience: removing one robot does not break the swarm, because each unit relies only on local sensing, just as removing one bird does not disrupt the flock.
Active research areas include understanding how information about a predator threat propagates through a murmuration faster than individual reaction time allows (a phenomenon called the "information cascade"), the role of noise and individual variation in maintaining flock cohesion, and applying machine learning to discover the actual interaction rules animals use from GPS-tracked trajectory data. Physicists also study flocking as an example of active matter — systems far from thermodynamic equilibrium that exhibit phase transitions analogous to magnetism.