Unmanned ground weapon target assignment based on deep Q-learning network with an improved multi-objective artificial bee colony algorithm. (January 2023)

• Record Type: Journal Article
• Authors: Wang, Tong; Fu, Liyue; Wei, Zhengxian; Zhou, Yuhu; Gao, Shan
• Is Part Of: Engineering applications of artificial intelligence. Volume 117:Part B(2023)
• Publication Date: 2023-01
• Subjects: Multi-objective optimization -- Deep reinforcement learning -- Weapon target assignment -- Adaptive operator selection mechanism; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2022.105612
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Various objective functions in the operation process of unmanned ground combat vehicles (UGVs) have an important impact on the equilibrium of the system. Unbalanced scheduling of unmanned ground combat vehicles and poor target strikes exist in complex urban battlefields. A new multi-weapon target assignment architecture and a multi-objective artificial bee colony (MOABC) algorithm with an elite strategy are proposed to solve these problems. Considering the influence of mutation operator on multi-objective assignment, by introducing the action mechanism of the self-adaptive variation operator and combining the state representation of the nectar source with the overall allocation scheme, the deep Q-learning network with improved multi-objective artificial bee colony (MOADQN) algorithm is proposed. Through comparative analysis with multi-objective artificial bee colony algorithm, non-dominated sorting genetic algorithm-II (NSGA-II), multi-objective particle swarm optimization (MOPSO), the multi-objective evolutionary algorithm based on decomposition with electronic countermeasure (ECM-MOEA/D) and the deep Q-learning network with multi-objective artificial bee colony (MOAIQL) algorithm, the proposed MOADQN algorithm can solve the problems such as poor allocation effectiveness and low gain of traditional algorithms. The proposed MOADQN algorithm has significant advantages in solving multi-objective optimization problems and strong expansion performance in the complex urban environment. Highlights: A new deep reinforcement learning framework for unmanned ground weapon target assignment is proposed. The adaptive action operator is proposed for the overall allocation scheme. The proposed MOADQN algorithm has significant advantages in large-scale complex scenarios.