针对自动化立体仓库出库作业过程中剩余货物退库问题,以堆垛机作业总能耗最小化为目标,以退库货位分配为决策变量,建立了自动化立体仓库退库货位优化模型,提出了基于深度强化学习的自动化立体仓库退库货位优化框架。在该框架内,以立体...针对自动化立体仓库出库作业过程中剩余货物退库问题,以堆垛机作业总能耗最小化为目标,以退库货位分配为决策变量,建立了自动化立体仓库退库货位优化模型,提出了基于深度强化学习的自动化立体仓库退库货位优化框架。在该框架内,以立体仓库实时存储信息和出库作业信息构建多维状态,以退库货位选择构建动作,建立自动化立体仓库退库货位优化的马尔科夫决策过程模型;将立体仓库多维状态特征输入双层决斗网络,采用决斗双重深度Q网络(dueling double deep Q-network,D3QN)算法训练网络模型并预测退库动作目标价值,以确定智能体的最优行为策略。实验结果表明D3QN算法在求解大规模退库货位优化问题上具有较好的稳定性。展开更多
如何在客户规定的时间内合理安排车辆运输路线,一直是物流领域亟待解决的问题。基于此,文章提出使用基于软更新策略的决斗双重深度Q网络(Dueling Double Deep Q-network,D3QN),设计动作空间、状态空间与奖励函数,对带时间窗的绿色车辆...如何在客户规定的时间内合理安排车辆运输路线,一直是物流领域亟待解决的问题。基于此,文章提出使用基于软更新策略的决斗双重深度Q网络(Dueling Double Deep Q-network,D3QN),设计动作空间、状态空间与奖励函数,对带时间窗的绿色车辆路径问题进行建模与求解。选择了小、中、大规模的总计18个算例,将三种算法的实验结果在平均奖励、平均调度车辆数、平均里程和运算时间四个维度进行比较。实验结果表明:在大多数算例中,与Double DQN和Dueling DQN相比,D3QN能在可接受的增加时间范围内,获得更高的奖励函数,调度更少的车辆数,运输更短的里程,实现绿色调度的目标。展开更多
可再生能源出力的波动性、间歇性、用户电力负荷的随机不确定性,使微电网的能量调度极具挑战性.为此,该文提出激励竞争双深度Q网络(motivation dueling double deep Q-network,简称MD3QN)算法,对微电网能量进行协调优化.仿真分析结果表...可再生能源出力的波动性、间歇性、用户电力负荷的随机不确定性,使微电网的能量调度极具挑战性.为此,该文提出激励竞争双深度Q网络(motivation dueling double deep Q-network,简称MD3QN)算法,对微电网能量进行协调优化.仿真分析结果表明:采用MD3QN算法对微电网进行能量调度,能实现削峰填谷,使微电网的经济效益最大化;相对于其他4种算法,MD3QN算法具有更高的综合性能.因此,MD3QN算法具有有效性.展开更多
文中提出了一种基于深度强化学习(deep reinforcement learning,DRL)的船舶智能避碰方法.该方法利用D3QN(double deep q-learning network with dueling architecture)算法与船舶领域模型,结合《国际海上避碰规则》(COLREGs)的避碰操作...文中提出了一种基于深度强化学习(deep reinforcement learning,DRL)的船舶智能避碰方法.该方法利用D3QN(double deep q-learning network with dueling architecture)算法与船舶领域模型,结合《国际海上避碰规则》(COLREGs)的避碰操作规范设计奖励函数,通过时序差分法实现优先经验回放,构建自主避碰的智能体.通过ROS-gazebo搭建仿真环境,构建神经网络处理环境中的视觉与雷达数据,快速有效地获取环境特征信息.结果表明:对比传统DQN算法,该方法具有更好的决策能力,训练时间更短;在避碰过程中可以对会遇局面做出正确的判断,选择符合COLREGs规范的避碰动作,最终可以准确并及时的避让目标船.展开更多
Physiological indices related to the efficiency (F-v/F-m) of light energy conversion in PS II and the peroxidation of membrane lipid were measured in leaves of Oryza sativa L. sp. indica rice cv. 'Shanyou 63' ...Physiological indices related to the efficiency (F-v/F-m) of light energy conversion in PS II and the peroxidation of membrane lipid were measured in leaves of Oryza sativa L. sp. indica rice cv. 'Shanyou 63' and sp. japonica rice cv. '9516'' under different temperatures and fight intensities for 4 days. No changes in F-v/F-m and membrane lipid peroxidation product (MDA) were observed, so neither photoinhibition nor photooxidation happened in both rice cultivars under moderate temperature and medium light intensity. However, F-v/F-m dropped obviously with no change in MDA contents, and photoinhibition appeared in indica rice cv. 'Shanyou 63' under medium temperature and strong light intensity. Furthermore, both photoinhibition and photooxidation were observed in two rice cultivars under chilling temperature and strong light intensity. Experiments with inhibitors under chilling temperature and strong light intensity showed that indica rice had a decrease in DI protein content and SOD activity, and the extent of inhibition of xanthophyll. cycle and nonphotochemical quenching (qN) was larger, and a higher level of MDA was observed. The photoinhibition and photooxidation in indica rice were more distinct as compared with japonica rice. The authors suggested that PS II light energy conversion efficiency (F-v/F-m) and membrane lipid peroxidation were the key indices for the detection of photooxidation.展开更多
文摘针对自动化立体仓库出库作业过程中剩余货物退库问题,以堆垛机作业总能耗最小化为目标,以退库货位分配为决策变量,建立了自动化立体仓库退库货位优化模型,提出了基于深度强化学习的自动化立体仓库退库货位优化框架。在该框架内,以立体仓库实时存储信息和出库作业信息构建多维状态,以退库货位选择构建动作,建立自动化立体仓库退库货位优化的马尔科夫决策过程模型;将立体仓库多维状态特征输入双层决斗网络,采用决斗双重深度Q网络(dueling double deep Q-network,D3QN)算法训练网络模型并预测退库动作目标价值,以确定智能体的最优行为策略。实验结果表明D3QN算法在求解大规模退库货位优化问题上具有较好的稳定性。
文摘可再生能源出力的波动性、间歇性、用户电力负荷的随机不确定性,使微电网的能量调度极具挑战性.为此,该文提出激励竞争双深度Q网络(motivation dueling double deep Q-network,简称MD3QN)算法,对微电网能量进行协调优化.仿真分析结果表明:采用MD3QN算法对微电网进行能量调度,能实现削峰填谷,使微电网的经济效益最大化;相对于其他4种算法,MD3QN算法具有更高的综合性能.因此,MD3QN算法具有有效性.
文摘文中提出了一种基于深度强化学习(deep reinforcement learning,DRL)的船舶智能避碰方法.该方法利用D3QN(double deep q-learning network with dueling architecture)算法与船舶领域模型,结合《国际海上避碰规则》(COLREGs)的避碰操作规范设计奖励函数,通过时序差分法实现优先经验回放,构建自主避碰的智能体.通过ROS-gazebo搭建仿真环境,构建神经网络处理环境中的视觉与雷达数据,快速有效地获取环境特征信息.结果表明:对比传统DQN算法,该方法具有更好的决策能力,训练时间更短;在避碰过程中可以对会遇局面做出正确的判断,选择符合COLREGs规范的避碰动作,最终可以准确并及时的避让目标船.
文摘Physiological indices related to the efficiency (F-v/F-m) of light energy conversion in PS II and the peroxidation of membrane lipid were measured in leaves of Oryza sativa L. sp. indica rice cv. 'Shanyou 63' and sp. japonica rice cv. '9516'' under different temperatures and fight intensities for 4 days. No changes in F-v/F-m and membrane lipid peroxidation product (MDA) were observed, so neither photoinhibition nor photooxidation happened in both rice cultivars under moderate temperature and medium light intensity. However, F-v/F-m dropped obviously with no change in MDA contents, and photoinhibition appeared in indica rice cv. 'Shanyou 63' under medium temperature and strong light intensity. Furthermore, both photoinhibition and photooxidation were observed in two rice cultivars under chilling temperature and strong light intensity. Experiments with inhibitors under chilling temperature and strong light intensity showed that indica rice had a decrease in DI protein content and SOD activity, and the extent of inhibition of xanthophyll. cycle and nonphotochemical quenching (qN) was larger, and a higher level of MDA was observed. The photoinhibition and photooxidation in indica rice were more distinct as compared with japonica rice. The authors suggested that PS II light energy conversion efficiency (F-v/F-m) and membrane lipid peroxidation were the key indices for the detection of photooxidation.
