The Multiple Vehicle Pickup and Delivery Problem with Time Windows (MV_PDPTW) which constitutes an important variant of the vehicle routing problems, deals with goods that have to be transported from origin to the destination points. In this problem, routes are designed in order to satisfy capacity, time windows, coupling and precedence constraints with the aim of minimization of total costs (which can be total distance, number of vehicles or both of them). Although many real life operations in logistics and transportation management can be modeled as MV_PDPTW, it has relatively less attention among vehicle routing literature because of it's difficulty. In this paper we propose a real valued genetic algorithm approach to solve MV_PDPTW. Problem variables are presented by real valued chromosomes. By the this way we assume to use less genes which improve search process. Proposed genetic algorithm approach has been tested on available benchmark problem sets and has compared with three previous GA results.