Aiming to the task assignment issue of multi-AUV(autonomous underwater vehicles) system, a self-organizing map(SOM) neural network based strategy of task assignment of multi-AUV and multi-objective is presented. Targets’ locations are set as input vectors of SOM neural network. Then self-organizing competitive calculations are carried out. Its output vectors are the corresponding AUV robots’ locations, so that a group of AUVs can be controlled to complete different tasks in different locations, and the robots can reach the designated targets in optimized paths. Simulation results in two-dimensional and three-dimensional working environments show the effectiveness of the proposed method.