The use of electrochemical-metallization-based volatile threshold switching selectors in cross-point arrays has been widely explored owing to their high on-off ratios and simple structure.However,these devices are uns...The use of electrochemical-metallization-based volatile threshold switching selectors in cross-point arrays has been widely explored owing to their high on-off ratios and simple structure.However,these devices are unsuitable for cross-point architectures because of the difficulty in controlling the random filament formation that results in large fluctuations in the threshold voltage during operation.In this study,we investigated the unidirectional threshold transition characteristics associated with an Ag/GST/HfO_(x)/Pt-based bilayer selector and demonstrated the occurrence of a low leakage current(<1×10^(-11) A) and low distribution of the threshold voltage(Δ0.11 V).The bilayer structure could control the filament formation in the intermediate state through the insertion of an HfO_(x) tunneling barrier.By stacking a bilayer selector with NiO_(x)based resistive random-access memory,the leakage and programming currents of the device could be significantly decreased.For the crossbar array configuration,we performed equivalent circuit analysis of a one-selector oneresistor(1S1R) devices and estimated the optimal array size to demonstrate the applicability of the proposed structure.The maximum acceptable crossbar array size of the 1S1R device with the Ag/GST/HfO_(x)/Pt/Ti/NiO_(x)/Pt structure was 5.29×10^(14)(N^(2),N=2.3×10^(7)).展开更多
针对卫星拒止环境下无人车在未知区域自主定位难题,提出一种从航空图像到地面点云的跨模态地点识别方法,并设计相应的网络架构(Aerial-to-Ground Position Recognition Network, AG-PRNet)。该方法通过数据预处理将点云投影到鸟瞰视图(B...针对卫星拒止环境下无人车在未知区域自主定位难题,提出一种从航空图像到地面点云的跨模态地点识别方法,并设计相应的网络架构(Aerial-to-Ground Position Recognition Network, AG-PRNet)。该方法通过数据预处理将点云投影到鸟瞰视图(Bird's Eye View, BEV)空间,减小其与航空图像的模态差异;设计旋转平移不变特征编码模块(Rotation And Translation Invariant CNN,RATI-CNN),提取跨模态数据的旋转平移不变特征;利用交叉注意力模块融合学习跨模态数据的共享特征,提升特征匹配的鲁棒性。在自建跨网域地点识别(Cross-Domain Place Recognition, CDPR)数据集上的实验表明,所提方法的Top-1和Top-5召回率分别达60.08%和76%,显著优于基线方法,验证了其在跨模态地点识别中的有效性。展开更多
基金financially supported by the National Research Foundation of Korea (NRF)(No.2016R1A3B1908249)。
文摘The use of electrochemical-metallization-based volatile threshold switching selectors in cross-point arrays has been widely explored owing to their high on-off ratios and simple structure.However,these devices are unsuitable for cross-point architectures because of the difficulty in controlling the random filament formation that results in large fluctuations in the threshold voltage during operation.In this study,we investigated the unidirectional threshold transition characteristics associated with an Ag/GST/HfO_(x)/Pt-based bilayer selector and demonstrated the occurrence of a low leakage current(<1×10^(-11) A) and low distribution of the threshold voltage(Δ0.11 V).The bilayer structure could control the filament formation in the intermediate state through the insertion of an HfO_(x) tunneling barrier.By stacking a bilayer selector with NiO_(x)based resistive random-access memory,the leakage and programming currents of the device could be significantly decreased.For the crossbar array configuration,we performed equivalent circuit analysis of a one-selector oneresistor(1S1R) devices and estimated the optimal array size to demonstrate the applicability of the proposed structure.The maximum acceptable crossbar array size of the 1S1R device with the Ag/GST/HfO_(x)/Pt/Ti/NiO_(x)/Pt structure was 5.29×10^(14)(N^(2),N=2.3×10^(7)).
文摘针对卫星拒止环境下无人车在未知区域自主定位难题,提出一种从航空图像到地面点云的跨模态地点识别方法,并设计相应的网络架构(Aerial-to-Ground Position Recognition Network, AG-PRNet)。该方法通过数据预处理将点云投影到鸟瞰视图(Bird's Eye View, BEV)空间,减小其与航空图像的模态差异;设计旋转平移不变特征编码模块(Rotation And Translation Invariant CNN,RATI-CNN),提取跨模态数据的旋转平移不变特征;利用交叉注意力模块融合学习跨模态数据的共享特征,提升特征匹配的鲁棒性。在自建跨网域地点识别(Cross-Domain Place Recognition, CDPR)数据集上的实验表明,所提方法的Top-1和Top-5召回率分别达60.08%和76%,显著优于基线方法,验证了其在跨模态地点识别中的有效性。
文摘在陆地与水域共存的复杂环境中,水陆两栖无人车(amphibious unmanned ground vehicle,A-UGV)跨域(即在水域与陆地之间的路径转换)三维路径规划是一项具有挑战性的任务。为应对这一挑战,提出一种基于地形信息优化启发函数的改进A^(*)算法,并结合最佳下水上岸点检测进行全局路径规划的方法(improved A^(*)path planning with optimal launch and ashore point detection,IA^(*)OLAPD)。对水陆环境进行地图构建,通过动态体素网格对环境点云数据进行分割和评估,将水域和陆地进行区分,并根据地形信息生成2D占用栅格地图、2.5D数字高程图及通行性地图。在路径规划阶段,将2.5D地图的多层地形信息转化为动态权重因子,优化A^(*)算法的启发函数,以增强复杂地形的适应性。在A-UGV跨越陆地和水域的过程中,算法结合路径长度、路径粗糙度、坡度、高程差和下水上岸点处的地形信息等因素,确定最佳的跨域过渡点,从而最小化整体路径代价和风险系数,实现陆地和水域之间的安全高效跨域过渡。仿真实验结果表明,IA^(*)OLAPD算法在水陆两栖跨域路径规划的安全性、稳定性和路径选择合理性方面具有显著优势。
