Wu Zhaoxiang1，Ouyang Quan1，*，Wang Zhisheng1，Ma Rui1，Cong Yuhua1，2
1. Nanjing University of Aeronautics and Astronautics，Nanjing 210016，China
2. Nanjing University of Science and Technology，Nanjing 210023，China
Abstract: Regional reconnaissance is an important branch of unmanned aerial vehicle(UAV) research. Due to the complexity of the actual mission and environment, the control method of regional reconnaissance must be provided with fast calculation speed, strong autonomy and intelligence. Artificial intelligence has been used in regional reconnaissance because of its strong learning ability, high efficiency, and high degree of integration. This paper systematically introduces the background of the regional reconnaissance problem and summarizes the methods based on artificial intelligence to solve this problem, which are mainly divided into two categories: heuristic algorithms for constructing and optimizing the objective function and deep reinforcement learning methods for solving the optimal value or strategy. Given by a comprehensive comparison of the above methods, it is found that deep reinforcement learning performs self-learning and online learning well, which can adapt to complex and unknown environments,and further it can quickly and accurately solve regional reconnaissance problems. In addition, this paper also discusses the development trend of regional reconnaissance technology and the challenges faced by deep reinforcement learning.
Key Words: artificial intelligence; regional reconnaissance; deep reinforcement learning; heuristic algorithm; autonomous intelligence