In this letter, we consider an air-and-ground cooperative network, where several aerial base stations (ABS) help terrestrial base stations (TBS) for coverage enhancement. In this network, we first quantify the space-time coverage ratio (STCR) by fully considering the antenna models and the dynamic of the ABS, and then formulate a joint ABS deployment and TBS antenna downtilt optimization problem with the objective to maximize the STCR of the concerned area. The objective function involves many control variables and judgement operations, which make the problem very complex. To solve the problem effectively, we first adopt the genetic algorithm (GA). Using the solutions of the GA as training samples, we propose a deep neural network architecture to further reduce the computational time. Simulation results indicate that the proposed GA significantly improves the coverage ratio and the deep neural network (DNN) architecture achieves orders of magnitude acceleration in computational time with acceptable performance.

中文翻译：

---

联合 ABS 部署和 TBS 天线下倾优化以实现覆盖范围最大化


在这封信中，我们考虑了一个空地协作网络，其中多个空中基站（ABS）帮助地面基站（TBS）增强覆盖范围。在该网络中，我们首先充分考虑天线模型和 ABS 的动态来量化空时覆盖率（STCR），然后制定联合 ABS 部署和 TBS 天线下倾角优化问题，目标是最大化相关区域。目标函数涉及很多控制变量和判断操作，使得问题变得非常复杂。为了有效地解决这个问题，我们首先采用遗传算法（GA）。使用遗传算法的解决方案作为训练样本，我们提出了一种深度神经网络架构，以进一步减少计算时间。仿真结果表明，所提出的遗传算法显着提高了覆盖率，并且深度神经网络（DNN）架构在计算时间上实现了数量级的加速，并且性能可接受。