In order to ensure Multiple Autonomous Underwater Vehicle (MAUV) could carry out detection under the multi-targets conflict conditions, a Double Layer Bio-inspired Self-Organism Map (DLBSOM) algorithm is proposed to complete the adaptive forward-reverse initial task assignment. Because of invalid task assignments are prone to occur in the iterative process of detection affected by ocean currents and AUV individual energy consumption, a task reassignment strategy including energy activation function is proposed to optimize task assignment. A Task Urgency Bio-inspired Neural Network (TUBNN) model is built to describe the underwater environment under the influence of ocean currents, in which distance and energy-supply intensity factor based on fuzzy-complementary judgment matrix is introduced to illustrate the mutual influence of ocean currents and trajectory distance of task reassignment. When the sub-individual AUV moves into the warning range centered on the target, the speed of the AUV is adjusted to achieve fine detection. Combining the turning database and nonlinear kinematic equation of the AUV, the path is smoothed to conform to the kinematic constraints of the AUV. The simulation tests are carried out to verify the feasibility and effectiveness of the proposed algorithm.