TY - GEN
T1 - A deep reinforcement learning based intelligent decision method for UCAV air combat
AU - Liu, Pin
AU - Ma, Yaofei
N1 - Publisher Copyright:
© Springer Nature Singapore Pte Ltd. 2017.
PY - 2017
Y1 - 2017
N2 - Based on deep reinforcement learning, an intelligent tactical decision method is proposed to solve the problem of Unmanned Combat Aerial Vehicle (UCAV) air combat decision-making. The increasing complexity of the air combat environment leads to a curse of dimensionality when using reinforcement learning to solve the air combat problem. In this paper, we employed the deep neural network as the function approximator, and combined it with Q-learning to achieve the accurate fitting of action-value function, which is a good way to reduce the curse of dimensionality brought by traditional reinforcement learning. In order to verify the validity of the algorithm, simulation of our deep Q-learning network (DQN) is carried out on the air combat platform. The simulation results show that the DQN algorithm has a good performance in both the reward and action-value utility. The proposed algorithm provides a new idea for the research of UCAV intelligent decision.
AB - Based on deep reinforcement learning, an intelligent tactical decision method is proposed to solve the problem of Unmanned Combat Aerial Vehicle (UCAV) air combat decision-making. The increasing complexity of the air combat environment leads to a curse of dimensionality when using reinforcement learning to solve the air combat problem. In this paper, we employed the deep neural network as the function approximator, and combined it with Q-learning to achieve the accurate fitting of action-value function, which is a good way to reduce the curse of dimensionality brought by traditional reinforcement learning. In order to verify the validity of the algorithm, simulation of our deep Q-learning network (DQN) is carried out on the air combat platform. The simulation results show that the DQN algorithm has a good performance in both the reward and action-value utility. The proposed algorithm provides a new idea for the research of UCAV intelligent decision.
KW - Combat simulation
KW - Deep reinforcement learning
KW - Intelligent decision
KW - UCAV
UR - https://www.scopus.com/pages/publications/85028985681
U2 - 10.1007/978-981-10-6463-0_24
DO - 10.1007/978-981-10-6463-0_24
M3 - 会议稿件
AN - SCOPUS:85028985681
SN - 9789811064623
T3 - Communications in Computer and Information Science
SP - 274
EP - 286
BT - Modeling, Design and Simulation of Systems - 17th Asia Simulation Conference, AsiaSim 2017, Proceedings
A2 - Mohamed Ali, Mohamed Sultan
A2 - Wahid, Herman
A2 - Mohd Subha, Nurul Adilla
A2 - Sahlan, Shafishuhaza
A2 - Md. Yunus, Mohd Amri
A2 - Wahap, Ahmad Ridhwan
PB - Springer Verlag
T2 - 17th International Conference on Asia Simulation, AsiaSim 2017
Y2 - 27 August 2017 through 29 August 2017
ER -