Airborne Multi-platform Sensor Scheduling Based on Reinforcement Learning

In the modern battle, information acquisition is the key for combat success and the reconnaissance is one of the main measures. Aiming at the systematic development of combat units, the reconnaissance mission is usually achieved by multi-platform cooperation. Airborne sensors, as the essential equipment to obtain battlefield information, are coordinated effectively for reaching the operation aim. There are two types of cooperative control strategies, short-sighted and non-short-sighted ones. In the process of strategy optimization, the former only aims to maximize the current immediate return, but ignores the long-term return. In addition, active sensors continuously radiate electromagnetic waves outward when obtaining continuous measurement, which is easy to expose their own position. Therefore, how to improve their ability to survive is particularly important. To this end, considering the target threat, the airborne multi-platform collaborative detection method is proposed based on reinforcement learning, which takes into account the current immediate return as well as the future long-term return, and aims to maximize information perception under the premise of self-security. The simulation tests demonstrate the effectiveness of this method.