Swarm Intelligence which emerges from interactions of simple individuals can be used to solve many problems. The foraging task in ant system is often considered as the prototype of cooperative behavior in Swarm Intelligence. The foraging model in swarm robots which considers the random feature of individual robot is built using the mean field method. Then the conflict between robots which influences the performance is observed. To solve this problem, a modified foraging strategy based on pheromone is proposed. From the simulations in Starlogo platform, it is shown that the modified method can reduces the conflict of robots and increase the performance of the system.