Boids Swarm — Flocking & Emergent Behavior

Boids (Craig Reynolds, 1987) shows that realistic flocking behaviour — birds, fish, drones — emerges from just three local rules applied independently by every agent. No global controller, no leader, no shared map. Each boid only perceives neighbours within a fixed radius and acts accordingly.

For a boid $i$ with position $\vec{p}_i$ and velocity $\vec{v}_i$, the three steering forces are:

Separation (avoid crowding, radius $r_s$):

$\vec{F}_{\text{sep}} = -\sum_{j \in N_s} \frac{\vec{p}_j - \vec{p}_i}{|\vec{p}_j - \vec{p}_i|^2}$

Alignment (match neighbours' heading, radius $r_a$):

$\vec{F}_{\text{ali}} = \frac{1}{|N_a|}\sum_{j \in N_a} \vec{v}_j - \vec{v}_i$

Cohesion (move toward group centre, radius $r_c$):

$\vec{F}_{\text{coh}} = \frac{1}{|N_c|}\sum_{j \in N_c} \vec{p}_j - \vec{p}_i$

Net steering: $\vec{F} = w_s\,\vec{F}_{\text{sep}} + w_a\,\vec{F}_{\text{ali}} + w_c\,\vec{F}_{\text{coh}}$

1. For each boid, find all neighbours within the perception radius.
2. Compute the three steering vectors independently.
3. Blend them with tunable weights $w_s$, $w_a$, $w_c$.
4. Add the net steering force to the current velocity; clamp to a maximum speed.
5. Update position: $\vec{p}_i \leftarrow \vec{p}_i + \vec{v}_i \cdot \Delta t$.
6. Repeat for all boids every frame — emergent flocking appears instantly.

Boids logic controls crowd simulation in films (Lord of the Rings battle scenes), drone light shows (Intel's 2000-drone displays), multi-robot search-and-rescue teams, and traffic flow simulation. The rules are so simple they run on microcontrollers in real robot swarms.

1. If you set the separation weight $w_s = 0$ but keep alignment and cohesion, what behaviour do you expect?
2. Why is the Boids model called an example of emergence? Could a central controller produce the same result more simply?