Genetic Algorithms are finding increasing number of applications in a variety of problems in a whole spectrum of disciplines. This paper describes a genetic algorithm with novel features against premature convergence successfully applied to solution of the Traveling Salesman Problem (TSP). Authors practiced modified version of Greedy Crossover, and operators brushing the population helping to escape from local minima. As the results, the frequency of finding the optimal solution was improved with keeping good convergence. Developed application was tested on a number of benchmarks: Oliver's 30, Eilon's 50 and Eilon's 75 towns TSPs. In 30 towns problem the algorithm seems to be always reaching an optimal solution. In 50 and 75 towns new better tours were found in comparison to all ones available to authors via papers on similar investigations. Proposed approach in TSP solution is generalized as the social disasters technique.

This paper describes some new features against premature convergence applied to genetic solution of the Traveling Salesman Problem (TSP). Authors practiced modified version of Greedy Crossover [1], which is less 'greedy' than a standard one, and operators brushing the population, which are helpful in getting out of local minima. As the result, the frequency of finding an optimal solution was improved with keeping good convergence. Developed application was tested on a number of benchmarks: Oliver's 30 [2], Eilon's 50 and Eilon's 75 [3] towns TSPs. In 30 towns problem the program seems to be always reaching an optimal solution. In 50 and 75 towns new better tours were found in comparison to all ones available to authors via papers on similar researches. Proposed approach in TSP solution is generalized as the social disasters technique. INTRODUCTION The TSP is defined as a task of finding of the shortest Hamiltonian cycle or path in complete graph of N nodes. It is a classic example of...