基于深度强化学习的AGV智能导航系统设计

针对现有的AGV在大规模未知复杂环境中进行自主导航配送的问题,基于深度强化学习完成了AGV智能导航系统设计。首先,结合传感器对周围的障碍物进行探测感知,利用DDPG（deep deterministic policy gradient）算法实现AGV小车从环境的感知输入到动作的直接输出控制,帮助AGV完成自主导航和避障任务。此外,针对训练样本易受环境干扰的问题,提出了一种新颖的DL（disturb learning）- DDPG算法,通过对学习样本中相关数据进行高斯噪声预处理,帮助智能体适应噪声状态下的训练环境,提升了AGV在真实环境中的鲁棒性。仿真实验表明,经改进后的DL-DDPG 算法能够为AGV导航系统提供更高效的在线决策能力,使AGV小车完成自主导航与智能控制。     

无线传感器网络一种改进的随机密钥预分配方案

为解决基本的随机密钥预分配方案连通率较低的问题,提出了一种无线传感器网络的密钥管理改进方案.通过执行一个"邻居"节点扩展协议,使安全孤立节点与一跳以外的节点进行共享密钥发现,从而找到至少一条安全路径来保证网络的安全连通.对改进方案和原有方案在全局连通率、安全性和能量消耗等方面性能的分析与仿真结果比较,表明密钥管理的改进方案以有限的额外密钥协商开销为代价,能换来网络全局安全连通率、可用节点总能量和安全性能的提高.     

Action range communication (ARC): A digital architecture for user and device interaction

We propose a new networking interface (ARC) for electronic devices. This interface will unify display, user input, and data transmission into one information broadcasting structure. The concept is based on the dependence of action on vision. Because action and vision are consecutive steps, they should be considered together as one information transmission process. We propose a model to describe this digital information transmission. In this model, a new information carrier containing both image and data is broadcasted to a receiver through the user. The user acts like a network switch that selects information to receiver. On‐screen user input and short‐range data transmission represent such transmission but in different environments. The display function is also included by incorporating the “null” data concept. The display matrix can be converted into an information broadcaster, and the panel becomes a transceiver for ARC. This reduces the cost and leads to a simpler device architecture with only networking interfaces. The information carrier is a generalization of data carrier. A general information transmission is proposed to include both ARC and data communication. We also discuss new interactions based on ARC such as cyborg input.     

1-(4-硝基苯基)-3-(5,6-二甲基-1,2,4三氮唑)-三氮烯质子迁移的理论研究

采用密度泛函B3LYP在6-31G*基组下,对有机显色剂1-(4-硝基苯基)-3-(5,6-二甲基-1,2,4三氮唑)-三氮烯(NPDMTT)的各种可能结构进行质子迁移的3种可能途径:(a)分子内质子迁移,(b)水助质子迁移,(c)甲醇助质子转移的计算,得到了各种途径异构体的相对能,获得了它们的互变异构过程的活化能、活化吉布斯自由能和质子转移反应的速率常数等性质。计算结果表明,分子内质子转移形成的各种异构体相对能量较大,当水分子或甲醇分子参与反应时,异构体的相对能量明显减小,但无论是孤立分子、一水合物还是一甲醇合物,其最稳定的异构体都相同,均为A2。溶剂化效应对异构化能垒的影响较大。最稳定的异构体分子内质子转移在N11和N13间转移的速控步骤的活化能为130.9 kJ.mol-1,反应速度常数为2.172×10-11s-1;当水分子参与反应以双质子转移机理异构化时,活化能显著降低,有利于三氮异构化,其中异构体质子在N11和N13间转移的速控步骤的活化能为22.55 kJ.mol-1,反应速度常数为3.617×107s-1;当醇分子参与反应以双质子转移机理异构时,活化能减小得更多,其中异构体质子在N11和N13间转移的速控步骤的活化能为2.384 kJ.mol-1,反应速度常数为9.032×1011s-1。计算结果还表明,氢键作用在增大NPDMTT一水合物和NPDMTT一甲醇合物相对稳定性、降低质子转移异构化反应活化能等方面起着重要作用。     

掺锆SnO_2陶瓷的乙醇气敏特性研究

在研究SnO_2乙醇气体敏感陶瓷中,比较添加物,发现ZrO_2对乙醇气体有较高的灵敏度。进而从工艺上研究了添加ZrO_2时的最佳烧结温度和预烧温度。分析了烧结温度与测试初始电阻的关系。总结出不同条件下乙醇的分解物不同。提出ZrO_2的催化机理是活化表面羟基,释放占有的电子,释放表面的吸附位,从而提高氧离子的效率。     

精子运动图象的多目标检测与分割

采用差图象动目标检测,窗口自适应分割技术实现了精子的快速检测定位和分割,克服了因目标亮暗不均造成的分割误差和目标丢失。利用形态特征鉴别静止精子大小差不多的杂质,为高效实用的ＣＡＳＡ系统奠定了基础。     

地震图中地震波形的自动提取方法研究

讨论了地震图中地震波形的自动提取方法,提出了一种适用性较强的水平校正和水平消除算法,以及一种利用峰值与零点进行波形校正的算法,归纳了波形提取中的各种波形模式并给出了智能化的波形搜索规则。     

A new analytical IGBT model with improved electricalcharacteristics

A new analytical insulated gate bipolar transistor (IGBT) model for improved on-state characteristics prediction is proposed. Two-dimensional (2-D) effects in the forward conduction of an IGBT are studied analytically. These effects significantly in8uence the on-state characteristics of the device and must be accounted for when IGBT modeling. Dynamic characteristics and temperature effects are included in the model. Both three-dimensional (3-D) numerical simulation and experimental results support the theoretical analysis     

Experiments on Vertical Turbulent Plane Jets in Water of Finite Depth

Detailed experiments on vertical turbulent plane jets in water of finite depth were carried out in a two-dimensional water tank. The jet velocities were measured with a laser Doppler velocimeter (LDV). The LDV measurement covers the entire flow regime: the zone of flow establishment (ZFE), the zone of established flow (ZEF), the zone of surface impingement (ZSI), and the zone of horizontal jets (ZHJ). From the experimental results, the following conclusions are reached. First, the jet flow is independent of the Reynolds number if the Reynolds number is sufficiently large to produce a turbulent jet. Second, in the initial ZFE, the jet flow is nonsimilar and is characterized by the two free shear layers along the two edges of the jet orifice. Third, the jet flow in ZEF is self-similar. Both mean and fluctuation velocities are scaled with the mean jet centerline velocity. The turbulent shear stress is predictable by Prandtl's third eddy viscosity model. The spreading of the confined vertical jets is larger than that of a free jet, so is the faster decay of jet centerline velocity. Fourth, in ZSI the jet flow is nonsimilar and high turbulent intensities were found. The vertical turbulent jet transforms into two opposite horizontal surface jets after the impingement. And finally, the maximum velocity of the horizontal surface jet in ZHJ decays according to a power law.