Deep reinforcement learning for unmanned aerial vehicles cluster task allocation

Real-time task allocation for Unmanned Aerial Vehicle (UAV) clusters is a complex decision problem, particularly in dynamic and uncertain environments. In practice, it can be difficult to predict enemy information, which is often obtained gradually through UAV reconnaissance. Traditional algorithms may not perform satisfactorily in such scenarios. In this study, we develop a Deep Reinforcement Learning (DRL) algorithm that combines Proximal Policy Optimization (PPO) with Long Short-Term Memory (LSTM) to address this challenge. By predicting the enemy’s positions and behavior, the algorithm performs real-time task allocation using a pre-trained model. Comparative experiments with basic DRL algorithms validate the convergence, effectiveness, and scalability of our approach.