Based on task allocation of multi-AUV (Autonomous Underwater Vehicle) system, the innovative auction algorithm is applied to hunting task assignments of multi-AUV. The two most decisive factor in the auction algorithm: the election of the auctioneer and the calculation of bidding value. In the traditional auction algorithm, the auctioneer's choice follows the principle of "Who finds the target, who is the auctioneer", but this principle seriously threatens the security and robustness of auction process of the multi-AUV. We propose a new method to determine auctioneer and the method of determining the bidding value of multiple AUV in the auction algorithm, and we have made an essential optimization of the auction algorithm. The final experimental results show that the optimized auction algorithm is compared with the traditional auction algorithm and other hunting tasks allocation algorithms, the security and robustness of the innovative auction has a better effect.