摘要
为了解决传统通信−感知融合网络模式对地面基础设施的依赖,针对复杂场景下通感融合网络系统功耗较大、信号阻塞、覆盖盲区等问题,提出了一种无人机搭载边缘计算服务器与雷达收发器辅助通感融合网络。首先,在满足用户传输功率、雷达估计信息率、任务卸载比例限制的条件下,通过联合优化无人机雷达波束成形、计算资源分配问题、任务卸载量划分、终端用户发射功率和无人机飞行轨迹,建立系统总能耗最小化问题;其次,将该非凸优化问题重新构建为一个马尔可夫决策过程,使用深度强化学习中的近端策略优化算法实现系统的优化决策。仿真结果表明,所提算法训练速度较快,能够在保证应用的感知与计算时延需求的同时有效降低系统能耗。
To address the dependence of traditional integrated sensing and communication network mode on ground infrastructure,the unmanned aerial vehicle(UAV)with edge computing server and radar transceiver was proposed to solve the problems of high-power consumption,signal blocking,and coverage blind spots in complex scenarios.Firstly,under the conditions of satisfying the user’s transmission power,radar estimation information rate and task offloading proportion limit,the system energy consumption was minimized by jointly optimizing UAV radar beamforming,computing resource allocation,task offloading,user transmission power,and UAV flight trajectory.Secondly,the non-convex optimization problem was reformulated as a Markov decision process,and the proximal policy optimization method based deep reinforcement learning was used to achieve the optimal solution.Simulation results show that the proposed algorithm has a faster training speed and can reduce the system energy consumption effectively while satisfying the sensing and computing delay requirements.
作者
李斌
彭思聪
费泽松
LI Bin;PENG Sicong;FEI Zesong(School of Computer Science,Nanjing University of Information Science and Technology,Nanjing 210044,China;School of Information and Electronics,Beijing Institute of Technology,Beijing 100081,China)
出处
《通信学报》
EI
CSCD
北大核心
2023年第9期228-237,共10页
Journal on Communications
基金
国家重点研发计划基金资助项目(No.2021YFB2900200)
国家自然科学基金资助项目(No.62101277)
江苏省自然科学基金资助项目(No.BK20200822)。
关键词
感知−通信−计算融合网络
无人机
深度强化学习
资源分配与优化
integrated sensing-communication-computation network
UAV
deep reinforcement learning
resource allocation and optimization
