基于强化学习的连续型机械臂自适应跟踪控制

针对空间主动碎片清除操作中连续型三臂节机器人系统跟踪问题,提出一种基于强化学习的自适应滑模控制算法。基于数据驱动的建模方法,采用BP神经网络对三臂节连续型机械臂进行建模,并作为预测模型指导强化学习实时调节所提出滑模控制器的控制参数,从而实现连续型机器人运动的实时跟踪控制。仿真结果表明：提出的数据驱动的预测模型对随机轨迹预测的相对误差保持在±1%以内,能够高精度地反映系统动态特性。对比固定参数的滑模控制器,提出的自适应控制器在保证系统达到控制目标的同时具有更低的超调量和更短的调节时间,表现出更好的控制效果。     

低能小型医用回旋加速器谐振腔的有限元分析

应用有限元方法分析了一台小型医用回旋加速器的谐振腔.以PC机为平台,采用ANSYS软件进行建模,对腔体的谐振频率、空载品质因数进行了研究,探讨了微调电容间距、网格剖分精度以及建模方法对计算结果的影响.回旋加速器谐振腔不是一个严格的封闭腔体,结构复杂.为减少计算量,根据其物理特性对腔体做了封闭处理,研究发现,这一简化的模型对计算精度的影响不大.计算结果和实际测量结果的误差在工程允许的范围之内,表明基于有限元的软件能够很好地计算类似回旋加速器高频谐振腔这样具有复杂边界的腔体结构,可用于加速器的物理设计及工程设计.     

Toxin bioportides: exploring toxin biological activity and multifunctionality

Toxins have been shown to have many biological functions and to constitute a rich source of drugs and biotechnological tools. We focus on toxins that not only have a specific activity, but also contain residues responsible for transmembrane penetration, which can be considered bioportides—a class of cell-penetrating peptides that are also intrinsically bioactive. Bioportides are potential tools in pharmacology and biotechnology as they help deliver substances and nanoparticles to intracellular targets. Bioportides characterized so far are peptides derived from human proteins, such as cytochrome c (CYCS), calcitonin receptor (camptide), and endothelial nitric oxide synthase (nosangiotide). However, toxins are usually disregarded as potential bioportides. In this review, we discuss the inclusion of some toxins and molecules derived thereof as a new class of bioportides based on structure activity relationship, minimization, and biological activity studies. The comparative analysis of the amino acid residue composition of toxin-derived bioportides and their short molecular variants is an innovative analytical strategy which allows us to understand natural toxin multifunctionality in vivo and plan novel pharmacological and biotechnological products. Furthermore, we discuss how many bioportide toxins have a rigid structure with amphiphilic properties important for both cell penetration and bioactivity.     

Retinoic acid regulates avian lung branching through a molecular network

Retinoic acid (RA) is of major importance during vertebrate embryonic development and its levels need to be strictly regulated otherwise congenital malformations will develop. Through the action of specific nuclear receptors, named RAR/RXR, RA regulates the expression of genes that eventually influence proliferation and tissue patterning. RA has been described as crucial for different stages of mammalian lung morphogenesis, and as part of a complex molecular network that contributes to precise organogenesis; nonetheless, nothing is known about its role in avian lung development. The current report characterizes, for the first time, the expression pattern of RA signaling members (stra6, raldh2, raldh3, cyp26a1, rarα, and rarβ) and potential RA downstream targets (sox2, sox9, meis1, meis2, tgfβ2, and id2) by in situ hybridization. In the attempt of unveiling the role of RA in chick lung branching, in vitro lung explants were performed. Supplementation studies revealed that RA stimulates lung branching in a dose-dependent manner. Moreover, the expression levels of cyp26a1, sox2, sox9, rarβ, meis2, hoxb5, tgfβ2, id2, fgf10, fgfr2, and shh were evaluated after RA treatment to disclose a putative molecular network underlying RA effect. In situ hybridization analysis showed that RA is able to alter cyp26a1, sox9, tgfβ2, and id2 spatial distribution; to increase rarβ, meis2, and hoxb5 expression levels; and has a very modest effect on sox2, fgf10, fgfr2, and shh expression levels. Overall, these findings support a role for RA in the proximal–distal patterning and branching morphogenesis of the avian lung and reveal intricate molecular interactions that ultimately orchestrate branching morphogenesis.     

模糊回归参数估计方法及应用于景气对策信号分析

模糊回归采用样本模糊数 (Xi,Yi)来对模糊回归参数进行估计 ,其中 Yi为观测模糊数 .本文认为一般模糊参数 A的估计方式不能真实地表达出样本所蕴含的信息 ,因而另行设立了一套模糊参数估计方法 ,此法对样本的解释方式将更为合理 ,且估计的过程也比线性规划简便 .本文还举出了该方法应用于景气对策信号分析的一个实例.     

随机生产中拖后需求的变质产品最优生产策略

针对部分短缺量拖后的变质产品,建立随机生产状态下的生产-库存模型.所谓随机生产状态,是指生产过程会随机地从一种可控状态向另一种不可控状态转移,当生产处于不可控状态时,会生产出一定比例的次品.其中,生产状态的转移服从均匀分布.通过生产-库存模型得到了随机生产状态下总成本函数,分析得到,该总成本函数是一个关于生产次数的凹函数,并由此提出了最优生产次数算法.最后给出了算例.算例结果表明,最优生产次数算法能够有效地将一个4n 1维的最优问题转化为一个一维问题,为变质商品的生产-库存决策提供了简便有效的方法.     

DMGSP:一种快速分布式全局序列模式挖掘算法

为了解决分布式环境下挖掘全局序列模式常产生过多候选序列,加大网络通信代价问题,提出了一种基于分布式环境下的快速挖掘全局序列模式算法--DMGSP.该算法将分布式环境下的各站点得到的局部序列模式压缩到一种语法序列树上, 避免了重复的序列前缀传输. 采用合并树中结点序列规则和项序扩展策略,对非频繁序列进行剪枝,有效地约简了候选序列,减少了网络传输量,从而快速生成全局序列模式.算法分析和实验结果表明,在大数据集环境下的DMGSP算法性能优越,能够有效地挖掘全局序列模式.     

一种高效挖掘高维数据的频繁闭合模式算法

为了克服传统高维数据挖掘频繁闭合模式算法迭代产生子表,引起算法执行时间长和存储开销大等问题,提出了一种高效挖掘高维数据的频繁闭合模式的算法EMHCP. EMHCP算法采用一种新型结构位图表来压缩存储数据,在仅扫描数据库一次后,建立位图转换表.根据位图转换表来构建混合树结构,采用深度优先的方式和有效的剪枝策略高效挖掘出所有的闭合模式.从而有效地缩小了搜索空间,加快了处理速度.通过在生物数据库应用的实验结果表明, EMHCP算法比已有的CARPENTER和TD-close等算法更为有效.     

项目投资时机及投资概率分析

针对实物期权理论中存在的不足,运用鞅理论分析研究了市场环境下的确定性和不确定性因素对实物资产项目投资决策的影响.首先,通过鞅方法建立了反映投资时机和投资回报临界值之间关系的投资时机模型;然后,使用测度变换评价了项目投资能够被实施的可能性.研究结果表明确定性因素对投资时机和投资回报临界值具有相反作用的影响,不确定性因素则均具有正向作用的影响,且在确定性因素取值较小情况下,不确定性对投资却具有促进作用.     

利他、公平和信任等因素对合作的影响

分析了利他、公平、信任等因素对博弈过程中合作的影响,以典型的"gift-exchange博弈模型"为例,假设博弈方行为受到一些因素影响并且可以用启发式规则简单描述,通过对成员Agent内部的利他、公平、信任等属性和交互过程中启发式规则的设计和编程实现,使得博弈双方在博弈过程中不断学习和调整策略,形成最终达成一致的合作机制.利用多Agent系统平台--JADE对所提出的模型进行仿真,通过对实验数据的分析,揭示了在多Agent系统中利他、公平以及信任等因素对合作的影响,指出了这些因素在经济生活中的实际指导意义.