检索结果-南通市图书馆

Improved Double Deep Q Network Algorithm Based on Average Q-Value Estimation and Reward Redistribution for Robot Path Planning

在线全文

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2024年第11期81卷 2769-2790页

作者： yameng yin Lieping Zhang Xiaoxu Shi Yilin Wang Jiansheng Peng Jianchu Zou Key Laboratory of Advanced Manufacturing and Automation Technology Guilin University of TechnologyEducation Department of Guangxi Zhuang Autonomous RegionGuilin541006China Guangxi Key Laboratory of Special Engineering Equipment and Control Guilin University of Aerospace TechnologyGuilin541004China Guilin Mingfu Robot Technology Company Limited Guilin541199China Key Laboratory of AI and Information Processing Education Department of Guangxi Zhuang Autonomous RegionHechi UniversityYizhou546300China

By integrating deep neural networks with reinforcement learning,the Double Deep Q Network(DDQN)algorithm overcomes the limitations of Q-learning in handling continuous spaces and is widely applied in the path planning of mobile ***,the traditional DDQN algorithm suffers from sparse rewards and inefficient utilization of high-quality *** those problems,an improved DDQN algorithm based on average Q-value estimation and reward redistribution was ***,to enhance the precision of the target Q-value,the average of multiple previously learned Q-values from the target Q network is used to replace the single Q-value from the current target Q ***,a reward redistribution mechanism is designed to overcome the sparse reward problem by adjusting the final reward of each action using the round reward from trajectory ***,a reward-prioritized experience selection method is introduced,which ranks experience samples according to reward values to ensure frequent utilization of high-quality ***,simulation experiments are conducted to verify the effectiveness of the proposed algorithm in fixed-position scenario and random *** experimental results show that compared to the traditional DDQN algorithm,the proposed algorithm achieves shorter average running time,higher average return and fewer average *** performance of the proposed algorithm is improved by 11.43%in the fixed scenario and 8.33%in random *** not only plans economic and safe paths but also significantly improves efficiency and generalization in path planning,making it suitable for widespread application in autonomous navigation and industrial automation.

关键词： Double Deep Q Network path planning average Q-value estimation reward redistribution mechanism reward-prioritized experience selection method

Structural properties and electrochemical performance of different polymorphs of Nb_(2)O_(5) in magnesium-based batteries

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Energy Chemistry 2021年第7期30卷 586-592页

作者： Cunyuan Pei yameng yin Xiaobin Liao Fangyu Xiong Qinyou An Mengda Jin Yan Zhao Liqiang Mai State Key Laboratory of Advanced Technology for Materials Synthesis and Processing Wuhan University of TechnologyWuhan 430070HubeiChina State Key Laboratory of Silicate Materials for Architectures International School of Materials Science and EngineeringWuhan University of TechnologyWuhan 430070HubeiChina Foshan Xianhu Laboratory of the Advanced Energy Science and Technology Guangdong Laboratory Xianhu Hydrogen ValleyFoshan 528200GuangdongChina

The selection of the most suitable crystal structure for ions storage and the investigation of the corresponding reaction mechanism is still an ongoing challenge for the development of Mg-based *** this article,high flexible graphene network supporting different crystal structures of Nb2 O5(TTNb_(2)O_(5)@rGO and T-Nb_(2)O_(5)@rGO) are successfully synthesized by a spray-drying-assisted *** three-dimensional graphene framework provides high conductivity and avoids the aggregation of Nb2 O5 *** employed as electrode materials for energy storage applications,TT-Nb_(2)O_(5) delivers a higher discharge capacity of 129.5 mAh g^(-1), about twice that of T-Nb_(2)O_(5) for Mg-storage,whereas,T-Nb_(2)O_(5) delivers a much higher capacity(162 mAh g^(-1)) compared with TT-Nb_(2)O_(5)(129 mAh g^(-1)) for *** investigations reveal the Mg intercalation mechanism and lower Mg^(2+) migration barriers,faster Mg^(2+) diffusion kinetics of TT-Nb_(2)O_(5) as cathode material for Mg-storage,and the faster Li+ diffusion kinetics,shorter diffusion distance of T-Nb_(2)O_(5) as cathode material for *** work demonstrates that exploring the proper crystal structure of Nb2 O5 for different ions storage is necessary.

关键词： Crystal structure Nb_(2)O_(5) Mg-ion storage Li-ion storage Diffusion kinetics

Fast and stable Mg^2+ intercalation in a high voltage NaV2O2(PO4)2F/rGO cathode material for magnesium-ion batteries

在线全文

学校读者我要写书评

暂无评论

Science China Materials 2020年第9期63卷 1651-1662页

作者： Junjun Wang Shuangshuang Tan Guobin Zhang Yalong Jiang yameng yin Fangyu Xiong Qidong Li Dan Huang Qinghua Zhang Lin Gu Qinyou An Liqiang Mai State Key Laboratory of Advanced Technology for Materials Synthesis and Processing Wuhan University of TechnologyWuhan 430070China Institute of Physics Chinese Academy of SciencesBeijing 100190China

Sluggish kinetics of Mg^2+intercalation and low working potential seriously hinder the development of highenergy-density magnesium-ion batteries(MIBs).Hence developing cathode materials with fast Mg^2+diffusion and high working voltage is a key to overcome the obstacles in ***,a tetragonal NaV2O2(PO4)2 F/reduced graphene oxide(r GO)is proposed as an effective Mg^2+host for the first *** exhibits the highest average discharge voltage(3.3 V ***^2+/Mg),fast diffusion kinetics of Mg^2+with the average diffusivity of 2.99×10^-10 cm^2s^-1,and ultralong cycling stability(up to 9500 cycles).The Mg^2+storage mechanism of NaV2O2(PO4)2 F/r GO is demonstrated as a single-phase(de)intercalation reaction by in situ X-ray diffraction(XRD)*** functional theory(DFT)computations further reveal that Mg^2+ions tend to migrate along the a direction.X-ray absorption near edge structure(XANES)demonstrates a decrease in the average valence of vanadium,and the local coordination environment around vanadium site is highly conserved after ***,the assembled NaV2O2(PO4)2 F//Mg0.79NaTi2(PO4)3 Mg-ion full cell exhibits high power and energy densities,which indicates that NaV2O2(PO4)2 F/r GO owns potential for practical *** work achieves a breakthrough in the working voltage of cathode materials for MIBs and provides a new opportunity for high-energy-density MIBs.

关键词： high-voltage fast kinetics MIBs fluorophosphate full cell

cKBET:assessing goodness of batch effect correction for single-cell RNA-seq

在线全文

学校读者我要写书评

暂无评论

Frontiers of Computer Science 2024年第1期18卷 215-228页

作者： yameng ZHAO yin GUO Limin LI School of Mathematics and Statistics Xi’an Jiaotong UniversityXi’an 710049China

Single-cell RNA sequencing reveals the gene structure and gene expression status of a single cell,which can reflect the heterogeneity between ***,batch effects caused by non-biological factors may hinder data integration and downstream *** the batch effect can be evaluated by visualizing the data,which actually is subjective and *** this work,we propose a quantitative method cKBET,which considers the batch and cell type information *** cKBET method accesses batch effects by comparing the global and local fraction of cells of different batches in different cell *** verify the performance of our cKBET method on simulated and real biological data *** experimental results show that our cKBET method is superior to existing methods in most *** general,our cKBET method can detect batch effect with either balanced or unbalanced cell types,and thus evaluate batch correction methods.

关键词： single-cell RNA-seq dataset b batch effect assessment cKBET method

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

基于多分支卷积的单目深度估计方法研究

计算机与数字工程 2023年第12期51卷 2966-2970页

作者：印雅萌周嘉麒王指辉南京航空航天大学南京210016

单目深度估计目前存在精度欠佳和物体边界深度预测模糊等挑战,针对上述问题,论文提出了一种基于多分支卷积的单目深度估计算法,利用复杂的卷积结构提取场景中更丰富的语义信息。模型在训练阶段使用四分支卷积替代原有单分支卷积,测试部... 详细信息

单目深度估计目前存在精度欠佳和物体边界深度预测模糊等挑战,针对上述问题,论文提出了一种基于多分支卷积的单目深度估计算法,利用复杂的卷积结构提取场景中更丰富的语义信息。模型在训练阶段使用四分支卷积替代原有单分支卷积,测试部署时可将多支路卷积的权重参数移植到原单支路网络上,从而网络模型在测试及使用阶段不会增加额外的推理时间。在公开数据集的测试对比中,论文提出的方法预测的深度图结果更加清晰,能有效地应对图片中物体边界等区域,实验结论证明论文提出的方法具备一定的有效性。

关键词：卷积神经网络自监督学习单目深度估计多分支卷积

Grow defect-rich bamboo-like carbon nanotubes on carbon black for enhanced microwave absorption properties in X band

在线全文

学校读者我要写书评

暂无评论

Journal of Materials Science & Technology 2022年第24期119卷 200-208页

作者： yameng Jiao Qiang Song Xuemin yin Liyuan Han Wei Li Hejun Li State Key Laboratory of Solidification Processing Shannxi Province Key Laboratory of Fiber Reinforced Light Composite MaterialsCarbon/Carbon Composites Research CenterNorthwestern Polytechnical UniversityXi’an 710072China

Due to the limited electromagnetic wave(EMW)loss capacity and agglomeration,carbon black(CB)gradually fails to meet the increasingly harsh demanding ***,defect-rich bamboo-like carbon nanotubes(CNTs)were grown on CB by the process of chemical vapor *** prepared in situ on CB can assist it to build a developed multilevel conductive network and introduce plentiful CB/CNTs ***’s more,the defects that accompany the growth of CNTs endow CNTs with a moderate conductivity and good impedance matching,thereby causing an effective microwave absorption(MA).Meanwhile,the high-density defects on CNTs can induce dipole polarization to further strengthen the EMW loss *** influence of CNTs with different growth time on MA performance has been *** from the structural merits,the synthesized CB-CNT with CNTs growth time of 40 min exhibits the optimal absorbing property,which has the minimum reflection loss of-53.6 d B and maximum effective absorption bandwidth of 4.1 GHz with the thickness of 2.7 mm,covering almost the entire X *** introduction of defect-rich CNTs significantly enhances the EMW loss ability of CB,which provides a rational strategy for the design of high-efficient microwave absorption materials.

关键词： Microwave absorption Carbon nanotubes Defect-rich Conductive network Interface polarization

在线全文

学校读者我要写书评

暂无评论

基于几何控制的四旋翼轨迹跟踪方法

计算机与数字工程 2024年第4期52卷 1055-1060页

作者：周嘉麒王指辉王民泰印雅萌翁祥瑞南京航空航天大学南京211106

针对欠驱动四旋翼无人机在大角度和复杂机动飞行中的控制问题,提出了一种基于几何控制的轨迹跟踪方法,该控制器具有全局稳定性,系统可以从任何初始状态恢复。首先,对四旋翼系统构型进行建模,解决了非线性流形上平移动力学和旋转动力学... 详细信息

针对欠驱动四旋翼无人机在大角度和复杂机动飞行中的控制问题,提出了一种基于几何控制的轨迹跟踪方法,该控制器具有全局稳定性,系统可以从任何初始状态恢复。首先,对四旋翼系统构型进行建模,解决了非线性流形上平移动力学和旋转动力学之间的耦合效应;其次,针对动态系统的基本几何特征,采用微分几何方法设计了跟踪控制器;最后,对该系统进行仿真验证了轨迹跟踪的闭环稳定性。仿真结果表明,当初始姿态误差小于90°时,系统表现出指数稳定性;当初始姿态误差大于90°小于180°时,系统具有全局指数吸引性。

关键词：四旋翼轨迹跟踪非线性流形几何控制

在线全文

学校读者我要写书评

暂无评论

基于分散注意力机制的无人机目标跟踪算法

计算机与数字工程 2024年第5期52卷 1287-1292页

作者：王指辉周嘉麒廖万斌印雅萌翁祥瑞南京航空航天大学南京211106

针对无人机平台的目标跟踪场景中目标小、形变大、背景复杂、易出现目标遮挡等问题,提出了一种基于分散注意力机制的无人机目标跟踪算法。首先,针对背景复杂、目标遮挡问题,在特征提取残差网络中引入分散注意力机制,增加卷积通道间的相... 详细信息

针对无人机平台的目标跟踪场景中目标小、形变大、背景复杂、易出现目标遮挡等问题,提出了一种基于分散注意力机制的无人机目标跟踪算法。首先,针对背景复杂、目标遮挡问题,在特征提取残差网络中引入分散注意力机制,增加卷积通道间的相互作用,赋予通道间自注意力权重。其次针对小目标、尺度变化大,添加特征金字塔结构,增加高语义信息在特征图中的占比,提高无人机场景下多尺度目标的跟踪能力,提高跟踪精度。在无人机公开数据集UAV123上采用一次通过评估模式进行测试,所提算法精度达到了82.7%,成功率达到了61.6%。实验结果证明论文所提算法的测试结果优于目前主流的目标跟踪算法,有效地提升了无人机在复杂场景下指定目标跟踪的精度和鲁棒性。

关键词：无人机目标跟踪分散注意力机制特征金字塔

Templated synthesis of spinel cobaltite MCo_(2)O_(4) (M¼Ni, Co and Mn) hierarchical nanofibers for high performance supercapacitors

在线全文

学校读者我要写书评

暂无评论

Journal of Materiomics 2021年第4期7卷 858-868页

作者： Xuemin yin Hejun Li Ruimei Yuan yameng Jiao Jinhua Lu State Key Laboratory of Solidification Processing Shaanxi Province Key Laboratory of Fiber Reinforced Light Composite MaterialsCarbon/Carbon Composites Research CenterNorthwestern Polytechnical UniversityXi’an710072China

Rational construction of transitional metal oxides electrode materials with suitable structure and composition is an effective strategy of improving their electrochemical ***,novel MCo_(2)O_(4) hierarchical nanofibers(H-MCo_(2)O_(4)NFs,M¼Ni,Co and Mn)were fabricated by a multi-step selftemplating method using electrospun nanofibers as *** from the unique structure,such as numerous of vertically interlinked nanosheets on the surface and 1D interwoven nanofibers networks,the obtained HeNiCo_(2)O_(4)NFs electrode exhibits a high specific capacitance of 1750 F g1(At a current density of 0.5 A g1),good rate capability(Capacitance retention of 70%at 20 A g1),and outstanding cycling stability(Capacitance retention of 92%after 6000 cycles).Moreover,the solid-state hybrid supercapacitor assembled by HeNiCo_(2)O_(4)NFs and activated carbon(AC),delivers a high energy density of 38.4 Wh kg1 at a power density of 800 W kg1,and excellent cycling ***,the HeNiCo_(2)O_(4)NFs is a promising candidate material for supercapacitors electrode and this self-templating method in this work also provides a new path for the preparation of one-dimensional hierarchical metallic oxides.

关键词： Electrospinning Supercapacitor NiCo_(2)O_(4) Hierarchical nanofibers Self-templating method