Abstract: The status of material delivery for an existing automobile general assembly line is analyzed. Because the materials should be delivered to the line in a given time interval, a vehicle routing problem with time windows (VRPTW) model is developed for the problem of material delivery. Then, an improved genetic algorithm (GA) is presented to solve it. By this algorithm, new chromosome coding and crossover methods are proposed. Also, to overcome the disadvantage of the traditional roulette selection, a new roulette selection method is proposed to increase the number of random number generated. It also keeps the strategy to select the best individuals. The experimental results validate the proposed method.