What algorithm assigns the frequency channels?

The simulation runs DSATUR (degree of saturation), a well-known greedy graph-coloring heuristic. It repeatedly picks the uncolored tower whose neighbours currently use the most distinct channels (breaking ties by raw degree), then assigns it the smallest channel number not already used by an active neighbour. When call traffic activates or deactivates towers, only the towers touching a conflict are recoloured, leaving the rest of the network undisturbed — a realistic incremental repair strategy rather than a full re-plan every tick.

Why do conflicts appear at all if the algorithm is correct?

Each tower keeps a sticky legacy channel from the last time it was active, mimicking how real base stations retain their last-assigned frequency plan. When a previously quiet tower reactivates near active neighbours, its old channel can now collide with one of them. That collision is a genuine, visible conflict until the solver runs its repair pass.

How is the interference graph built?

Every pair of towers whose ground-plane distance is less than the interference radius is connected by an edge, representing overlapping coverage cells that would cause co-channel interference if assigned the same frequency. This is the graph that must be properly coloured: no two towers joined by an edge may share a channel while both are actively carrying traffic.

What does the traffic load slider control?

Each tower has an independent simulated call-traffic waveform. The load slider shifts the activation threshold, so a higher value means towers spend more of their cycle above the threshold and are active more often — increasing how frequently the coverage pattern, and therefore the active interference graph, changes.

Is DSATUR guaranteed to find the minimum number of channels?

No — optimal graph coloring is NP-hard in general, so DSATUR is a heuristic, not an exact solver. It typically performs very well in practice and often matches or comes close to the chromatic number on the geometric interference graphs generated here, but it is not guaranteed to be optimal on every random layout.

Home▸AI & Machine Learning▸Cell Network Load Balancer — Graph Coloring Live

📡 Cell Network Load Balancer — Graph Coloring Live

Watch a graph-coloring optimizer reassign frequency channels across overlapping cell towers to eliminate interference, balancing load as simulated call traffic shifts across the network.

AI & Machine Learning3DAdvanced60 FPS

⚙ Under the hood

Watch a graph-coloring optimizer reassign frequency channels across overlapping cell towers to eliminate interference, balancing load as simulated call traffic shifts across the network.

Three.jsWebGLAIMachine Learning

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install