An improved ant colony algorithm and simulation

We demonstrate a novel ant colony system with dynamically varied parameters and a penalty-reward function, which is based on the basic ant system (BAS) algorithm, also presented is its application to solving complex TSP problem. Our new algorithm has two important features, the first: a perturbation factor formulated by inverse exponent penalty-reward function is developed; the second: a corresponding transition strategy with random selection is designed. Numerical simulation demonstrates that our new algorithm has much higher convergence speed and stability than BAS algorithm, and brings along good effects of reducing CPU time, and preventing search from being in stagnation behavior.