An Adaptive Spectral Conjugate Gradient Method with Restart Strategy

As a generalization of the two-term conjugate gradient method(CGM),the spectral CGM is one of the effective methods for solving unconstrained optimization.In this paper,we enhance the JJSL conjugate parameter,initially proposed by Jiang et al.(Computational and Applied Mathematics,2021,40:174),through the utilization of a convex combination technique.And this improvement allows for an adaptive search direction by integrating a newly constructed spectral gradient-type restart strategy.Then,we develop a new spectral CGM by employing an inexact line search to determine the step size.With the application of the weak Wolfe line search,we establish the sufficient descent property of the proposed search direction.Moreover,under general assumptions,including the employment of the strong Wolfe line search for step size calculation,we demonstrate the global convergence of our new algorithm.Finally,the given unconstrained optimization test results show that the new algorithm is effective.

A Shifting Strategy for Electric Commercial Vehicles Considering Mass and Gradient Estimation

The extended Kalman filter (EKF) algorithm and acceleration sensor measurements were used to identify vehiclemass and road gradient in the work. Four different states of fixed mass, variable mass, fixed slope and variableslope were set to simulate real-time working conditions, respectively. A comprehensive electric commercial vehicleshifting strategy was formulated according to the identification results. The co-simulation results showed that,compared with the recursive least square (RLS) algorithm, the proposed algorithm could identify the real-timevehicle mass and road gradient quickly and accurately. The comprehensive shifting strategy formulated had thefollowing advantages, e.g., avoiding frequent shifting of vehicles up the hill, making full use ofmotor braking downthe hill, and improving the overall performance of vehicles.

Enhanced Deep Reinforcement Learning Strategy for Energy Management in Plug-in Hybrid Electric Vehicles with Entropy Regularization and Prioritized Experience Replay

Plug-in Hybrid Electric Vehicles(PHEVs)represent an innovative breed of transportation,harnessing diverse power sources for enhanced performance.Energy management strategies(EMSs)that coordinate and control different energy sources is a critical component of PHEV control technology,directly impacting overall vehicle performance.This study proposes an improved deep reinforcement learning(DRL)-based EMSthat optimizes realtime energy allocation and coordinates the operation of multiple power sources.Conventional DRL algorithms struggle to effectively explore all possible state-action combinations within high-dimensional state and action spaces.They often fail to strike an optimal balance between exploration and exploitation,and their assumption of a static environment limits their ability to adapt to changing conditions.Moreover,these algorithms suffer from low sample efficiency.Collectively,these factors contribute to convergence difficulties,low learning efficiency,and instability.To address these challenges,the Deep Deterministic Policy Gradient(DDPG)algorithm is enhanced using entropy regularization and a summation tree-based Prioritized Experience Replay(PER)method,aiming to improve exploration performance and learning efficiency from experience samples.Additionally,the correspondingMarkovDecision Process(MDP)is established.Finally,an EMSbased on the improvedDRLmodel is presented.Comparative simulation experiments are conducted against rule-based,optimization-based,andDRL-based EMSs.The proposed strategy exhibitsminimal deviation fromthe optimal solution obtained by the dynamic programming(DP)strategy that requires global information.In the typical driving scenarios based onWorld Light Vehicle Test Cycle(WLTC)and New European Driving Cycle(NEDC),the proposed method achieved a fuel consumption of 2698.65 g and an Equivalent Fuel Consumption(EFC)of 2696.77 g.Compared to the DP strategy baseline,the proposed method improved the fuel efficiency variances(FEV)by 18.13%,15.1%,and 8.37%over the Deep QNetwork(DQN),Double DRL(DDRL),and original DDPG methods,respectively.The observational outcomes demonstrate that the proposed EMS based on improved DRL framework possesses good real-time performance,stability,and reliability,effectively optimizing vehicle economy and fuel consumption.

梯度人工骨修复支架调控骨骼系统组织的修复与再生

背景:梯度人工骨修复支架模拟了骨骼系统中的独特特征,在骨骼系统再生中具有巨大的应用潜力。目的:综述梯度人工骨修复支架在骨骼系统组织工程中的最新研究进展,并阐述了其优势与制造策略。方法:由第一作者检索Web of Science和PubMed数据库2000-2023年发表的文献,英文检索词为“gradient,bone regeneration,scaffold”,最终筛选后对76篇文献进行分析总结。结果与结论:①作为骨骼系统组织高效、高质量修复的重要手段,梯度人工骨修复支架目前针对骨组织、骨-软骨、肌腱-骨组织的天然梯度特征进行了仿生设计,这些支架能够一定程度地从结构、成分上模拟原生组织的细胞外基质,从而促进细胞黏附、迁移、增殖和分化,促进受损组织向原生状态再生恢复。②先进制造技术为梯度人工骨修复支架制备提供了更多可能;目前已经开发了通过空间差异化纤维排布和生物活性物质加载构建的梯度电纺纤维支架;分层叠加、分级孔隙率与生物3D打印技术制造的梯度3D打印支架;原位分层注射、简单逐层叠加、冷冻干燥法制造的梯度水凝胶支架;另外还包括其他方式或多方法联用的支架;这些支架在体外实验中展示了良好的生物相容性,在小型动物实验中能够加速组织再生并且观察到组织学结构明显改善。③目前开发的梯度人工骨修复支架仍需进一步优化,提高在梯度尺度上的匹配性,进一步明确材料与组织相互作用,避免降解产物导致的副反应等问题,未来需要结合相关学科优势与临床需求进一步优化。

基于改进的DISK算法的农林航拍图配准研究

针对无人机的农林航拍图像配准算法存在特征点识别较少、特征点匹配不精准等问题,提出了一种改进的离散关键点(Discrete Key point,DISK)算法。该算法首先采用梯度策略DISK算法对特征点进行有效描述;其次采用基于深度学习局部特征的匹配方法进行预匹配;最后利用RANSAC算法去除离群点,筛选匹配结果。为验证算法的有效性,选取了几组农业航拍图像的数据集进行实验比较。实验结果显示,与经典的SIFT、Dark feat算法及原始的DISK算法相比,改进的DISK算法显著提高了匹配精度,匹配精度由41.7%提升至98.9%,充分满足农林航拍图的匹配需求。

Adaptations of the floral characteristics and biomass allocation patterns of Gentiana hexaphylla to the altitudinal gradient of the eastern Qinghai-Tibet Plateau

This study addressed the floral component traits and biomass allocation patterns of Gentiana hexaphylla as well as the relationships of these parameters along an elevation gradient(approximately 3700 m, 3800 m, 3900 m, and 4000 m) on the eastern Qinghai-Tibet Plateau. The plant height, floral characteristics, and biomass allocation of G. hexaphylla were measured at different altitudes after field sampling, sorting, and drying. Plant height was significantly greater at 3700 m than that at other elevations. Flower length was significantly greater at 4000 m than that at other elevations, whereas the flower length at low elevations showed no significant differences. Corolla diameter increased with altitude, although the difference was not significant between 3800 m and 3900 m. Variations in biomass accumulation, including the aboveground, photosynthetic organ, flower and belowground biomasses, showed non-linear responses to changes in altitude. The aboveground and photosynthetic organ biomasses reached their lowest values at 4000 m, whereas the belowground and flower biomassreached minimum values at 3700 m. The sexual reproductive allocation of G. hexaphylla also increased with altitude, with a maximum observed at 4000 m. These results suggest that external environmental factors and altitudinal gradients as well as the biomass accumulation and allocation of G. hexaphylla play crucial roles in plant traits and significantly affect the ability of this species to adapt to harsh environments. The decreased number of flowers observed at higher altitudes may indicate a compensatory response for the lack of pollinators at high elevations, which is also suggested by the deformed flower shapes at high altitudes. In addition, the individual plant biomass(i.e., plant size) had significantly effect on flower length and corolla diameter. Based on the organ biomass results, the optimal altitude for G. hexaphylla in the eastern Qinghai-Tibet Plateau is 3800 m, where the plant exhibits minimum propagule biomass and asexual reproductive allocation.

Hybrid gradient vector fields for path-following guidance

Guidance path-planning and following are two core technologies used for controlling un-manned aerial vehicles(UAVs)in both military and civilian applications.However,only a few approaches treat both the technologies simultaneously.In this study,an innovative hybrid gradient vector fields for path-following guidance(HGVFs-PFG)algorithm is proposed to control fixed-wing UAVs to follow a generated guidance path and oriented target curves in three-dimensional space,which can be any combination of straight lines,arcs,and helixes as motion primitives.The algorithm aids the creation of vector fields(VFs)for these motion primitives as well as the design of an effective switching strategy to ensure that only one VF is activated at any time to ensure that the complex paths are followed completely.The strategies designed in earlier studies have flaws that prevent the UAV from following arcs that make its turning angle too large.The proposed switching strategy solves this problem by introducing the concept of the virtual way-points.Finally,the performance of the HGVFs-PFG algorithm is verified using a reducedorder autopilot and four representative simulation scenarios.The simulation considers the constraints of the aircraft,and its results indicate that the algorithm performs well in following both lateral and longitudinal control,particularly for curved paths.In general,the proposed technical method is practical and competitive.

广西猫儿山优势木本植物叶功能性状关联性沿海拔梯度的变化规律

为揭示植物对海拔差异导致的环境变化的适应策略,以猫儿山不同海拔梯度上的5个植物群落为研究对象,测定了群落优势树种的叶含水率(LWC)、叶干物质含量(LDMC)、比叶面积(SLA)、叶体积(LV)、叶厚度(LT)、叶组织密度(LTD)以及比叶重(SLW)等7个功能性状,运用单因素方差分析、Pearson相关分析以及标准化主轴法分析了猫儿山植物叶功能性状的海拔分异规律及性状组合关系。结果显示7个叶功能性状变异系数的大小顺序为SLA>LV>SLW>LTD>LT>LDMC>LWC。叶功能性状在海拔梯度上的差异均达极显著水平(P<0.01),不同性状随海拔的变化规律不一致,如SLA随海拔升高先升后降、LV则呈下降趋势。性状之间存在较为普遍的组合关系,如SLA与LDMC、LT、LTD及SLW间呈极显著负相关,LDMC与LTD及SLW间呈极显著正相关,多数性状组合间的相关关系只在部分海拔达显著水平。研究结果表明,猫儿山植物叶功能性状对海拔梯度引起的环境差异产生响应,叶功能性状组合的协同变异或权衡在海拔梯度上存在非同步现象。

小型无人有缆遥控水下机器人智能控制方法

针对深度确定性策略梯度(DDPG)算法应用于无人有缆遥控水下机器人(ROV)运动控制时存在的坏样本影响学习稳定性、缺少环境探索能力以及学习时间长难收敛等问题,从神经网络结构、噪声引入和融合监督学习3个方面对DDPG算法进行改进,并提出了基于混合神经网络结构和参数噪声的监督式DDPG算法。仿真结果表明,监督式DDPG算法比常规DDPG算法和传统比例-积分-微分(PID)算法更加有效。

梯度风场滑翔仿生飞行器设计与仿真

将信天翁从梯度风场中获取能量滑翔数千英里的飞行策略应用在飞行器上是突破飞行器续航瓶颈的重要方法之一。本文结合信天翁身体特性和飞行策略,设计了一种具有信天翁仿生外形、气动布局和控制方式的仿信天翁飞行器,巡航时升阻比为梯形翼飞行器的1.13倍,最大可达梯形翼飞行器升阻比的1.31倍,通过翼尖变形阻力最高可减小12%,翼根载荷减小35%。根据信天翁在滑翔时不时扑动翅膀的行为提出了间歇动力梯度风滑翔方法,对仿信天翁飞行器进行了风梯度滑翔操控性能仿真。结果表明实际飞行路径与风梯度滑翔路径的误差小于5%,能量消耗为匀速飞行的51.5%,验证了间歇动力梯度风滑翔策略的合理性。

基于二阶价值梯度模型强化学习的工业过程控制方法

为了实现对高延时、非线性和强耦合的复杂工业过程稳定准确的连续控制,提出了一种基于二阶价值梯度模型强化学习的控制方法。首先,该方法在模型训练过程中加入了状态价值函数的二阶梯度信息,具备更精确的函数逼近能力和更高的鲁棒性,学习迭代效率更高;其次,通过采用新的状态采样策略,可以更高效地利用模型进行策略学习。最后,通过在OpenAI的Gym公共实验环境和两个工业场景的仿真环境的实验表明:基于二阶价值梯度模型对比传统的基于最大似然估计模型,环境模型预测误差显著降低;基于二阶价值梯度模型的强化学习方法学习效率优于现有的基于模型的策略优化方法,具备更好的控制性能,并减小了控制过程中的振荡现象。可见该方法能有效地提升训练效率,同时提高工业过程控制的稳定性和准确性。

一种基于TVM的算子生成加速策略

随着人工智能(AI)的飞速发展,新算子和底层硬件层出不穷,这给算子库的开发和维护带来了巨大的工作量。单纯基于手工优化来解决AI模型的性能和效率很容易出现瓶颈。TVM深度学习编译器通过代码的自动化生成减轻了手工优化的负担,但同时也存在搜索时间长的问题。为此,针对TVM的自动化代码生成框架Ansor,提出基于梯度提升算法的新代价模型和基于预定义规则的调度空间剪枝优化2种优化策略,旨在加速TVM的自动化代码生成过程,实现模型快速落地与部署,并进一步为人工智能技术的应用提供更高效的解决方案。实验结果表明,通过应用优化后代价模型可以在不损失推理时间的前提下,使得在x86 CPU平台上模型的调优时间减少30%~35%,同时优化后算子性能最高可提升22%,使得在深度计算单元(DCU)平台上模型的调优时间减少20%左右,同时优化后算子平均性能提升5.7%,此外,基于预定义规则的剪枝策略可以有效提升代价模型的收敛速度,并且在原有最佳迭代次数下,模型推理时间可提高7.4%。

基于强化学习的飞行器自主规避决策方法

考虑飞行器在执行任务过程中存在诸多不可预知的威胁或障碍,为保障飞行器的安全性,本文进行飞行器面向威胁目标的自主规避决策方法研究。首先综合考虑飞行器与威胁目标行为之间的相互影响,提出了基于深度长短期记忆(LSTM)神经网络的轨迹预测算法,实现对威胁目标未来轨迹的预测;然后结合预测信息构建拦截场景下规避机动的马尔可夫决策过程,设计了基于改进双延迟深度确定性策略梯度(P-TD3)的飞行器规避决策方法,以最大化规避过程的总收益为优化目标,实现飞行器自主规避决策。最后通过在虚拟仿真交互平台的试验验证,本文的决策方法提升了网络的收敛速度,具有84%的规避成功率,提高了飞行器对潜在威胁的成功规避概率,有利于增强飞行器的自主性与安全性。

结合局部纹理特征滤波的海天线检测方法

针对视频监测下海天线检测图像易出现的海天对比度低、海上物体、海雾、海浪及云层等干扰条件引起的图像模糊和海天线遮挡问题,提出一种基于偏心邻域的灰度共生矩阵对比度滤波的梯度域下海天线检测方法。方法设计使用上下两方向偏心盒状滤波器获取图像的对比度梯度值图,利用滑动窗对分块图像搜索梯度极大值候选位置,根据设计策略通过构建海天线参数的概率统计模型提取拟合直线作为检测结果。仿真实验表明,本文方法能提升视觉图像下含多景物等干扰的复杂场景海天线检测的准确率,与同类方法相比提升7.69%。

MEC架构下基于DDPG的车联网任务卸载和资源分配

为了缓解车联网中个体车辆计算资源配置过低而导致的任务处理时延较大的问题,提出了一种移动边缘计算(mobile edge computing,MEC)架构下的动态任务卸载策略和资源分配方案。以最小化全网任务处理时延为目标,将车联网中的任务卸载和资源分配问题建模为马尔可夫决策过程(Markov decision process,MDP),并利用深度确定性策略梯度(deep deterministic policy gradient,DDPG)算法进行了问题求解。仿真结果表明,与执行者-评价者(actor-critic,AC)和深度Q网络(deep Q-network,DQN)这2种算法相比,DDPG算法以最快的算法收敛特性获得最小的全网任务处理时延。

基于叶片功能性状的贺兰山西坡植物生态策略分析

植物叶片功能性状能够反映其对不同环境的适应策略,对不同环境梯度产生适应性改变。以贺兰山西坡沿海拔分布的4种植被类型(荒漠草原、浅山灌丛、亚高山针叶林、亚高山灌丛草甸)优势植物为对象,在物种-群落水平,基于叶片功能特征CSR(竞争-耐胁迫-杂草型植物)生态策略模型,通过叶片比叶面积(SLA)、叶干物质含量(LDMC)、单位面积碳(LC)、氮(LN)、磷(LP)含量的比较,探究4种植被类型优势植物S-R(耐胁迫-杂草)轴的经济性状权衡策略及该区优势种的生态策略组成;并且研究环境因素对CSR生态策略的作用机理,确定能够影响植物CSR生态策略的主要环境因子。结果显示,1)在物种水平,贺兰山西坡46种优势植物生态策略可分为5类,其中R(杂草型)和SR(耐胁迫-杂草型)策略为主导策略,所占比例分别为52.1%和23.9%。2)在群落水平,贺兰山西坡30个样地生态策略划分为5类,主导策略为R(杂草型策略),占比39.2%。3)S策略与比叶面积显著性负相关关系;R策略与全磷、全碳、叶干物质含量显著性负相关关系,和比叶面积显著性正相关关系。4)海拔、年均温、年均降雨量、太阳辐射、水蒸气压对C策略有显著影响,随海拔上升C策略降低。土壤全磷与S策略正相关,与R策略负相关。该研究表明植物能够改变功能性状(如比叶面积、叶片碳、氮、磷含量),改变资源获得能力和物质分配,应对不同环境,从而形成不同的生态策略。

东天山北坡雪岭云杉种群结构动态的海拔格局

雪岭云杉(Picea schrenkiana)是天山特有树种,也是天山森林优势种;研究其种群结构动态的海拔梯度变化有助于保护、管理这一重要的森林资源。本研究以东天山北坡江布拉克景区的雪岭云杉种群为研究对象,利用便携式背包雷达(LiBackpack)扫描与手测验证相结合的方法,在1800—2600 m的海拔梯度上设置了27个30 m×30 m(共24300 m^(2))的样方进行调查测量,基于树木胸径数据,制作了雪岭云杉的静态生命表和存活曲线,计算其种群数量动态变化指数和时间序列预测模型,分析不同海拔高度雪岭云杉种群结构动态特征。结果表明:1)整体来看,江布拉克景区雪岭云杉种群结构现状良好,但幼苗(DBH<5 cm)储备明显不足。低、高海拔雪岭云杉种群径级结构表现为“倒J型”,中海拔段为“稳定型”。雪岭云杉种群在不同海拔高度存在幼龄树死亡率高的共性。2)中海拔雪岭云杉种群动态变化指数最低;低海拔种群危险率最低;表明中海拔树木存活能力最强,雪岭云杉种群在中海拔生长状况最好。3)时间序列预测模型表明,研究区雪岭云杉种群未来动态更新趋势良好,未来2、4、6、8径级时,低海拔段Ⅱ—Ⅶ径级、高海拔段Ⅱ、Ⅳ—Ⅷ径级个体数量呈现增加的趋势;中海拔不同径级个体数量变化较小。建议对低、高海拔雪岭云杉种群个体采取抚育、间伐措施,中海拔砍伐过多的区域应移栽原生苗而促进雪岭云杉种群发育更新;同时加大管理力度,减少人为干扰。

赖草草地叶片养分、碳组分和防御性化合物对氮添加的响应

本研究依托山西右玉黄土高原赖草(Leymus secalinus)草地2017年建立的氮(Nitrogen, N)添加梯度试验平台(0~32 g·m^(-2)·a^(-1)),探究N添加对优势植物赖草叶片养分[N、磷(Phosphorus, P)、钾(Potassium, K)]、非结构性碳水化合物、结构性碳水化合物和防御性化合物含量的影响。结果表明:N素输入显著提高叶片N,K含量和N∶K和N∶P比值,降低叶片P含量和P∶K比值;N素输入显著降低叶片非结构性碳水化合物和结构性碳水化合物含量;N素输入显著增加叶片总酚、单宁和黄酮类化合物含量;PCA分析表明低N(≤8 g·m^(-2)·a^(-1))和高N(>8 g·m^(-2)·a^(-1))输入下赖草叶片性状存在显著差异,土壤无机N含量、土壤有效N∶P和N∶K是其主要影响因子。以上结果表明N素输入改变了盐渍化草地赖草叶片养分-碳组分-防御性化合物的分配策略。

自记忆的深度强化学习模型求解多维背包问题

本文针对多维背包问题维度高,约束强的特点提出了自记忆的学习优化模型(self memorized learn to improve,SML2I),通过深度强化学习的学习机制选择迭代搜索过程中的算子即模型学习当前的解以及历史搜索过程中的解,判断对当前解采用提升策略或者是扰动策略,在此基础上,进一步提出了哈希表与设计了2种有效的基于价值密度的扰动算子.使用哈希表记录历史搜索过程中的解,防止模型重复探索相同的解,基于价值密度的扰动策略生成的新解与之前的解决方案完全不同,因此针对扰动后的解再次采用提升策略同样有效,通过测试89个MKP数据集并与其他文献中先进的求解方法进行对比,实验结果验证了SML2I模型求解MKP问题的可行性与有效性.

一种用于模块化CPC均衡充电的最优化策略

随着锂电池在工业及生活中的广泛应用,锂电池组均衡充电策略的高效性及快速性也变得愈发重要,因此构建单体电池与电池模块间CPC(cell-to-pack-to-cell)的均衡充电系统,以解决锂电池组快速均衡充电问题。首先,对均衡系统进行模块化处理,建立模块内和模块间的均衡电路;然后,提出1种用于均衡充电的最优化策略,对所提策略模型进行分层求解,顶层通过二分法来计算充电时间,底层通过梯度下降法来优化充电电流;最后,通过与非模块化均衡充电策略下的充电时间、均衡时间、单体电池端电压及均衡器承担电压进行仿真对比,验证了所提策略的可行性和有效性。