【摘要】 随着物联网的普及,对物联网终端设备可使用能量的要求也在提高。能量收集技术拥有广阔前景,其能通过产生可再生能量来解决设备能量短缺问题。考虑到未知环境中可再生能量的不确定性,物联网终端设备需要合理有效的能量分配策略来保证系统持续稳定工作。文中提出了一种基于DQN的深度强化学习能量分配策略,该策略通过DQN算法直接与未知环境交互来逼近目标最优能量分配策略,而不依赖于环境的先验知识。在此基础上,还基于强化学习的特点和系统的非时变系统特征,提出了一种预训练算法来优化该策略的初始化状态和学习速率。在不同的信道数据条件下进行仿真对比实验,结果显示提出的能量分配策略在不同信道条件下均有好于现有策略的性能,且兼具很强的变场景学习能力。更多还原

【Abstract】 Due to the increasing popularization of the Internet of Things(IoT),the requirements for the power that can be used by the terminal equipment of the IoT are also constantly improving.Energy harvesting technology is a promising solution to overcome equipment energy shortages by generating renewable energy.Considering the uncertainty of renewable energy in the unknown environment, the terminal equipment of the IoT needs a reasonable and effective energy allocation strategy to ensure the continuous and stable operation of the system.In this paper, a DQN-based deep reinforcement learning energy allocation strategy is proposed, which uses DQN algorithm to directly interact with the unknown environment to approach the optimal energy allocation strategy without relying on the prior knowledge of the environment.Moreover, a pre-training algorithm is proposed to optimize the initialization state and learning rate of the strategy based on the characteristics of reinforcement learning and time-inva-riant system.The simulation results under different channel data conditions show that the energy allocation strategy proposed in this paper has better performance than the existing strategy under different channel conditions, and has strong variable scene learning ability.更多还原