The traditional ant colony algorithm is based on positive feedback search way, leading to the existence of slow convergence speed and shortcoming of easily trapped in local minima. This paper proposes a kind of extended ant colony algorithm based on the mixed feedback mechanism(MF-ACO). On the basis of the traditional ant colony algorithm, the algorithm defines a kind of extension type ants, which have strong global search ability, to help the algorithm get out of local minima. In addition, the negative feedback balance mechanism based on stimulus-response model, the convergence ability and global search ability of dynamic balance algorithm are designed by referring to the labor division behavior of ant colony. Finally, on the basis of the labor division model, the individual ants pheromone update strategy is improved to further accelerate the algorithm convergence speed. Using a multiple TSP instance as the test object to conduct simulation experiments, the experimental results show the superiority of the proposed algorithm, and then this algorithm is applied to the robot path planning problem, to verify the effectiveness of the proposed algorithm in actual application.