Competitive learning parallel agents: application to travelling salesperson problem