Drone Swarm Formation Control — Distributed Consensus (2D)

Formation shape

Swarm

Drones 12

Comm radius 130 px

Obstacles 3

Formation health

Mean position error

—

Converged

—

Comm edges

—

Graph connected?

—

Avoidance events (last 3 s)

How it works

Each drone has its own fixed target offset from a virtual leader — the geometry that defines the wedge, circle, diamond or line. This is a real distributed formation-consensus control law, not boids: instead of steering by neighbour alignment/cohesion/separation, only one pinned drone (the diamond marker) is pulled directly toward the leader's absolute goal. Every other drone only ever compares its own formation error to the errors reported by drones inside its communication radius (the thin blue lines), and applies

u_i = −k · Σ_j∈N(i) [(p_i − d_i) − (p_j − d_j)] − c · v_i

where p is position, d the desired offset position, and N(i) the current neighbours. This drives every drone's formation error toward the average error of its neighbours; because the pinned drone's error is anchored near zero, the consensus propagates that anchor across the graph edge by edge until the whole shape locks in. Cut the comm radius low enough to fragment the graph (see "Graph connected?") and the formation can drift out of shape in the disconnected piece — a real limitation of consensus control, not a bug.

Obstacle avoidance is a separate repulsive potential-field term added on top of the consensus law, pushing a drone away from nearby obstacles in proportion to 1/distance² — it does not use the communication graph.
Scatter randomises every drone's position so you can watch the consensus law pull disorder back into the exact formation shape from scratch.
Click anywhere on the canvas to send the whole formation to a new leader waypoint.

Click to set a new leader waypoint. Thin lines = active comm links; dashed ghosts = target formation slots.