Curiosity Driven Collaborative Reconnaissance of Multiple Unmanned Aerial Vehicles

In the process of multi-UAV reconnaissance and exploration, the effective rewards given to intelligent agents by the environment are too sparse, while standard reinforcement learning algorithms perform poorly in environments with sparse feedback to intelligent agents, specifically manifested as not actively exploring the environment. A curiosity driven reinforcement learning algorithm (ICM-IDQN) combining intrinsic motivation learning is proposed to address the problem of sparse environmental rewards. After experimental verification, this method can obtain more rewards in sparse environments, accelerate convergence, and increase exploration performance.