ACO / Ant System Explorer

Ant Colony Optimization — layered combinatorial search and graph-based TSP

© 2026 Theodore P. Pavlic  · MIT License
© 2026 Theodore P. Pavlic  · MIT License
ANT COLONY GRAPHpaused
Select τ pheromone or η prior view to interact
OBJECTIVE HISTORY
Reference Key
mants per iteration (colony size)
αpheromone exponent in τα·ηβ
βheuristic exponent in τα·ηβ
ρevaporation rate: τ ← (1−ρ)τ, floored at τ_min = τ_init. At ρ=1.0 all paths reset to the uniform baseline each iteration (near-random with low Q). Range 0–1.
Qdeposit scale — one ant on the optimal path deposits Q per node per iteration. Vary Q to control how quickly pheromone differentiates; vary ρ to control how long it persists.
τPheromone — learned trail strength. Higher τ means more ants chose that path historically. Decays via evaporation; built up by deposit. Click in τ view to inject.
ηPrior — fixed heuristic attractiveness, independent of pheromone. All presets start uniform (η=1); use the η prior view to add bias. Click to boost, right-click to decrease.
Statistics & Status
Iter. best — best objective in the current iteration. Fluctuates; shows — until the first iteration completes.
Glob. best — best across all iterations; never decreases. Persists even as pheromone evaporates — this is the elitist archive.
2nd / 3rd best — 2nd and 3rd distinct solutions by objective value, not by τ. Can differ from τ view because pheromone decays. Shown in top paths view.
Tour — one ant's complete solution path: a choice at each decision layer (layered mode) or a full circuit through all cities (TSP). "Paused mid-tour" = ants are partway through building their current solutions for this iteration.
Algorithm — click to expand
Application — TSP is the original ACO formulation
Objective function
Controls
speed 1.0×
Statistics
Iter. best
Glob. best
2nd best
3rd best
View
Pheromone memory
Node trails  ↔  Edge trails
Parameters
m8
α1.0
β2.0
ρ0.10
Q1.0
© 2026 Theodore P. Pavlic  · MIT License