Multi Drones Target Allocation Auction Algorithm Based on DQN Bidding Strategy
CHEN Zihao
HU Chunhe
Abstract：To maximize the revenue of task matching and success rate for multi drone monitoring target al-location,as well as minimize the costs of path length and obstacle collision risks,this paper proposes an autonomous evolutionary auction algorithm based on a data-driven reinforcement learning method that in-tegrates deep Q-network(DQN)bidding strategy.This algorithm first constructs a Markov decision mod-el for multi drone task target auction,and constructs a new type of DQN bidding and auctioning decision neural network model with taking the remaining bidding capacity of bidders as the environment,output-ting the price increase factor as the action,and the increase in auction revenue before and after the two rounds as the return.This model constructs a reinforcement learning training sample library that includes elements such as auction environment,markup factors,and returns.During the auction process,the DQN neural network is continuously trained in an offline learning mode,allowing it to output markup factors based on the auction environment according to the DQN strategy during the auction process,achieving op-timization of auction results and revenue.Finally,the simulation of multi drone and multi monitoring tar-get allocation verifies the effectiveness of the target allocation method based on the DQN auction mecha-nism proposed in this paper.Compared with the results of traditional auction algorithms,the auction reve-nue obtained by this method is increased by 21.4%.
Keywords：multi dronesauction algorithmDQNmulti-objective allocationauction proceeds
Publication Date：2024-08-28
Online Publishing Date：2026-08-28(First online date of this platform, not the publication date of the document)
Pages：10( 23-32 )