ID 原文 译文
53067 此外，在请求的平均响应时间上，QRFS算法较加权轮询(WRR)算法提高了 12%，较加权最小连接数(WLC)算法提高了 11%; Besides, QRFS improves the average response time by 12% compared to WRR, and 11% compared to WLC.
53068 在服务器负载的均衡性上，QRFS算法取得了更好的效果，节点间的负载方差更小。 In terms of load balance of the server, QRFS algorithm achieves smaller load variance.
53069 针对传统调度算法在高危多变环境下实现多目标在线调度所面临的困境，提出基于深度强化学习的调度优化算法， Aiming at the predicament faced by traditional scheduling algorithms to achieve multi-objective online scheduling in high-risk and variable environments, a scheduling algorithm based on deep reinforcement learning is proposed.
53070 并应用于大规模舰载机出动回收多目标在线调度问题中。 This method applies to the multi-objective online scheduling problem of large-scale carrier aircraft dispatch and recovery.
53071 该方法以减小舰面位移、减少会遇次数、均衡设备利用率和稳定调度周期作为调度决策目标， This approach takes the reduction of deck displacement and the number of meets, the stability of the scheduling cycle, the balance of equipment utilization as the scheduling multi-objective.
53072 依照马尔可夫决策过程(MDP)，构造以舰载机和各设备状态作为输入，调度行为动作作为输出，带权特征向量作为奖赏的在线调度即时决策模型。 According to the Markov decision process(MDP), the model is built, which uses the aircraft and equipment's states as inputs, the behavior value function as the outputs and weighted feature vectors as the reward.
53073 搭建用于训练的优化深度强化学习网络，改进动作选择策略和网络结构以提升性能，从而实现在线调度决策优化。 Then, an optimized deep reinforcement learning network is built for training, improving action selection strategies and neural network's training performance.
53074 实验结果表明，利用该方法得到的决策模型能够在线解决突发状况，在静态和动态调度方面，相对于启发式算法和调度规则本文方法在安全性和高效性方面具有优势。 Experiment results show that the decision model obtained can be processed immediately with the emergency, while compared with the scheduling rules and heuristic algorithm for scheduling multi-objective decision-making problems, the method has significant advantages on security and flexibility.
53075 保密容量最大化是物理层安全(PLS)的主要目标。 Maximizing the secrecy capacity is a key goal of physical layer security(PLS).
53076 本文首先利用到达角(AOA)估计窃听者位置， In this paper, the location of an eavesdropper is estimated based on angle of arrival(AOA) at first.