Tunning parameters of evolutionary algorithm in Travelling Salesman Problem with profits and returns

A huge number of papers studies Travelling Salesman Problem (TSP) in classical version. In standard TSP all cities must be visited and graph is completed. While this is indeed the case in many practical problems, there are many other practical problems where these assumptions are not valid. This paper presents a new evolutionary algorithm (EA) which solves TSP with profits and returns (TSPwPR). This version of TSP is often applied in Intelligent Transport Systems, especially in Vehicle Routing Problem (VRP). TSPwPR consists in finding a cycle which maximizes collected profit but does not exceed a given cost constraint. A graph which is considered in this problem can be not completed, salesman doesn't have to visit all cities and he can repeat (with zero profit) cities in his tour. The method was implemented and tested on real network which consists of 160 cities in eastern and central voivodeships of Poland. The main parameter which has the highest influence on quality of obtaining results is the size of population and our experiments are directed to determine an optimal value of this parameter.