An Improved Ant System using Least Mean Square Algorithm

In this paper, we propose a modified model of pheromone updation for Ant-System (AS), entitled as Improved Ant System (IAS), and develop a new modeling framework for the above mentioned AS using the properties of basic Adaptive Filters. Here, we have exploited the properties of Least Mean Square (LMS) algorithm for the pheromone updation to find out the best minimum tour length for the Travelling Salesman Problem (TSP) and to resolve the basic shortcoming of easily falling into local optima and slow convergence speed. The desired length is updated in every iteration, which is the global minimum length and LMS algorithm is used to calculate the cost function (i.e., pheromone, which depends on the tour length). Hence, the pheromone is updated for the best minimum tour path. This improved algorithm has better search ability and good convergence speed. TSP library has been used for selection of a benchmark problem and the proposed IAS determines the minimum tour length for the problems containing large number of cities. Our algorithm shows effective results and gives least tour length in most of the cases as compared to other existing approaches.