A New Ant Colony Algorithm Based On Dynamic Local Search For TSP

As the traditional ant colony algorithm(ACA) for solving TSP(traveling salesman problem) is easy to fall into a local optimal solution and slow convergence, also the quality and stability of solution is not ideal. To solve these problems, an core improved ant colony algorithm based on dynamic local(DLACA) search which achieved by scout ant, the three auxiliary strategies are city classification, enlarge the gap of pheromone and clear pheromone after inherit is proposed.City classification means that at the beginning of the algorithm, we should classify all cities into several classes. cities of different types represent different levels of the city, and more ants to the higher ranks, while the lower ranks are opposite; And through the introduction of bionics scout ants, is that during ant routing process, if there is the difficulty of choosing cause that it is difficult to choose what way to go, dispatched as appropriate scout ants help to explore, at the end contrast the length of scout ants and ant path to decide which path a is better. In each routing is completed, the new pheromone update strategy appear, which is obtain when the time iteration of the shortest path, and ratio with the shortest path before when the time of seeking, if the ratio is greater, pheromone update amplitude is bigger also; When the algorithm falls into local optimal, in order not to let the pheromone completely cleared, use a inherited strategy, so that not only can jump out of local optimum, but also can improve the efficiency of the algorithm.The principle of traditional ant colony algorithm to solve TSP were in detail, the analysis of ant colony algorithm in solving traveling salesman problem (TSP) produces the root cause of the three defects, starting from the root causes, design improved ant colony algorithm. The traditional ACA and DLACA were used to solve TSP are simulated by Matlab. Simulation results show that the improved algorithm in solving traveling salesman problem (TSP) to greatly optimize the solution quality, that the solving results and become stable, and can effectively jump out of local optimal solution, and can greatly accelerate the convergence speed. All aspects show that its performance is better than the traditional ant colony algorithm.