This paper considers a cognitive communication network, which consists of a flying base station deployed by an unmanned aerial vehicle (UAV) to serve its multiple downlink ground terminals (GTs), and multiple underlaid device-to-device (D2D) users. To support the GTs’ throughput while guaranteeing the quality-of-service for the D2D users, the paper proposes the joint design of D2D assignment, bandwidth, and power allocation. This design task poses a computationally challenging mixed-binary optimization problem, for which a new computational method for its solution is developed. Multiple binary (discrete) constraints for the D2D assignment are equivalently expressed by continuous constraints to leverage systematic processes of continuous optimization. As a result, this problem of mixed-binary optimization is reformulated by an exactly penalized continuous optimization problem, for which an alternating descent algorithm is proposed. Each round of the algorithm invokes two simple convex optimization problems of low computational complexity. The theoretical convergence of the algorithm can be easily proved and the provided numerical results demonstrate its rapid convergence to an optimal solution. Such a cognitive network is even more desirable as it outperforms a non-cognitive network, which uses a partial bandwidth for D2D users only.

中文翻译：

---

认知无人机网络中的联合 D2D 分配、带宽和功率分配

本文考虑了一个认知通信网络，该网络由无人机（UAV）部署的飞行基站为其多个下行链路地面终端（GT）和多个底层设备到设备（D2D）用户组成。为了支持GTs的吞吐量，同时保证D2D用户的服务质量，本文提出了D2D分配、带宽和功率分配的联合设计。这个设计任务提出了一个计算上具有挑战性的混合二元优化问题，为此开发了一种新的计算方法来解决这个问题。D2D 分配的多个二元（离散）约束由连续约束等效表示，以利用连续优化的系统过程。因此，这个混合二元优化问题由一个精确惩罚的连续优化问题重新表述，为此提出了一种交替下降算法。算法的每一轮都会调用两个计算复杂度低的简单凸优化问题。该算法的理论收敛性很容易证明，所提供的数值结果表明其快速收敛到最优解。这种认知网络甚至更受欢迎，因为它优于仅对 D2D 用户使用部分带宽的非认知网络。该算法的理论收敛性很容易证明，所提供的数值结果表明其快速收敛到最优解。这种认知网络甚至更受欢迎，因为它优于仅对 D2D 用户使用部分带宽的非认知网络。该算法的理论收敛性很容易证明，所提供的数值结果表明其快速收敛到最优解。这种认知网络甚至更受欢迎，因为它优于仅对 D2D 用户使用部分带宽的非认知网络。