An Actor-Critic-Based UAV-BSs Deployment Method for Dynamic Environments

• DOI: 10.48550/arxiv.2002.00831
• Publication Date: 2020-01-01
• Authors: Zhiwei Chen; Yi Zhong; Xiaohu Ge; Yi Ma
• Keywords: Robustness; Position (finance)

In this paper, the real-time deployment of unmanned aerial vehicles (UAVs) as flying base stations (BSs) for optimizing throughput mobile users is investigated UAV networks. This problem formulated a time-varying mixed-integer non-convex programming (MINP) problem, which challenging to find an optimal solution in short time with conventional optimization techniques. Hence, we propose actor-critic-based (AC-based) deep reinforcement learning (DRL) method near-optimal positions at every moment. proposed method, process searching iteratively particular moment modeled Markov decision (MDP). To handle infinite state and action spaces improve robustness process, two powerful neural networks (NNs) are configured evaluate position adjustments make decisions, respectively. Compared heuristic algorithm, sequential least-squares fixed UAVs methods, simulation results have shown that outperforms these three benchmarks terms