A novel distributed cooperative search strategy for multiple underwater robots is proposed based on the initial target information, so as to reduce the average time of searching target and improve the search target probability in statistical sense. The cooperative search strategy divides the target search process into two phases. In the first search stage, the underwater robot predicts the possible existence range of the target based on the speed and the elapsed time of the target. When the underwater robot is not in the target prediction range, it directly moves towards the initial position of the target to reduce the time of the blind search. In the second search stage, the underwater robot enters the target prediction range, then the target existence probability is updated in real time according to the sensor detection results, and the predictive control thought is used to make optimization decisions according to the target existence probability. Finally, the Monte Carlo simulation experiments are carried out for 1000 times. The proposed search strategy in this paper is compared with the cooperative search strategy with no target initial information. The comparison results show the effectiveness and feasibility of searching a moving target using the proposed search strategy.(1)