Automatic Driving Operation Strategy of Urban Rail Train Based on Improved DQN Algorithm

To realize a better automatic train driving operation control strategy for urban rail trains,an automatic train driving method with improved DQN algorithm(classical deep reinforcement learning algorithm)is proposed as a research object.Firstly,the train control model is established by considering the train operation requirements.Secondly,the dueling network and DDQN ideas are introduced to prevent the value function overestimation problem.Finally,the priority experience playback and“restricted speed arrival time”are used to reduce the useless experience utilization.The experiments are carried out to verify the train operation strategy method by simulating the actual line conditions.From the experimental results,the train operation meets the ATO requirements,the energy consumption is 15.75%more energy-efficient than the actual operation,and the algorithm convergence speed is improved by about 37%.The improved DQN method not only enhances the efficiency of the algorithm but also forms a more effective operation strategy than the actual operation,thereby contributing meaningfully to the advancement of automatic train operation intelligence.

High-Speed Trains Automatic Operation with Protection Constraints: A Resilient Nonlinear Gain-based Feedback Control Approach

This paper addresses the control design for automatic train operation of high-speed trains with protection constraints.A new resilient nonlinear gain-based feedback control approach is proposed,which is capable of guaranteeing,under some proper non-restrictive initial conditions,the protection constraints control raised by the distance-to-go(moving authority)curve and automatic train protection in practice.A new hyperbolic tangent function-based model is presented to mimic the whole operation process of high-speed trains.The proposed feedback control methods are easily implementable and computationally inexpensive because the presence of only two feedback gains guarantee satisfactory tracking performance and closed-loop stability,no adaptations of unknown parameters,function approximation of unknown nonlinearities,and attenuation of external disturbances in the proposed control strategies.Finally,rigorous proofs and comparative simulation results are given to demonstrate the effectiveness of the proposed approaches.

Joint On-Demand Pruning and Online Distillation in Automatic Speech Recognition Language Model Optimization

Automatic speech recognition(ASR)systems have emerged as indispensable tools across a wide spectrum of applications,ranging from transcription services to voice-activated assistants.To enhance the performance of these systems,it is important to deploy efficient models capable of adapting to diverse deployment conditions.In recent years,on-demand pruning methods have obtained significant attention within the ASR domain due to their adaptability in various deployment scenarios.However,these methods often confront substantial trade-offs,particularly in terms of unstable accuracy when reducing the model size.To address challenges,this study introduces two crucial empirical findings.Firstly,it proposes the incorporation of an online distillation mechanism during on-demand pruning training,which holds the promise of maintaining more consistent accuracy levels.Secondly,it proposes the utilization of the Mogrifier long short-term memory(LSTM)language model(LM),an advanced iteration of the conventional LSTM LM,as an effective alternative for pruning targets within the ASR framework.Through rigorous experimentation on the ASR system,employing the Mogrifier LSTM LM and training it using the suggested joint on-demand pruning and online distillation method,this study provides compelling evidence.The results exhibit that the proposed methods significantly outperform a benchmark model trained solely with on-demand pruning methods.Impressively,the proposed strategic configuration successfully reduces the parameter count by approximately 39%,all the while minimizing trade-offs.

Tracking performance of large margin classifier in automatic modulation classification with a software radio environment

Automatic modulation classification is the process of identification of the modulation type of a signal in a general environment. This paper proposes a new method to evaluate the tracking performance of large margin classifier against signal-tonoise ratio （SNR）, and classifies all forms of primary user＇s signals in a cognitive radio environment. For achieving this objective, two structures of a large margin are developed in additive white Gaussian noise （AWGN） channels with priori unknown SNR. A combination of higher order statistics and instantaneous characteristics is selected as effective features. Simulation results show that the classification rates of the proposed structures are well robust against environmental SNR changes.

Configuration Synthesis and Performance Analysis of 9‑Speed Automatic Transmissions

Current research of automatic transmission(AT)mainly focuses on the improvement of driving performance,and configuration innovation is one of the main research directions.However,finding new configurations of ATs is one of the main limitations of configuration innovation.In the present study,epicyclic gear trains(EGTs)are applied to investigate mechanisms of 9-speed ATs.Then four kinematic configurations are proposed for automatic transitions.In order to evaluate the performance of proposed mechanisms,the lever analogy method is applied to conduct kinematic and mechanical analyses.The power flow analysis is conducted,and then transmission efficiencies are calculated based on the torque method.The comparative analysis between the proposed and existing mechanisms is carried out where obtained results show that proposed mechanisms have reasonable performance and can be used in ATs.The prototype of an AT is manufactured and the speed test is conducted,which proves the accuracy of analysis and the feasibility of proposed mechanisms.

The influence of AAR coupler features on estimation of in-train forces

Inadequate management of large in-train forces transferred through coupler systems of a railway train leads to running and structural failures of vehicles.Understanding these phenomena and their mitigation requires accurate estimation of relative motions and in-train forces between vehicle bodies.Previous numerical studies have ignored inertia of coupling elements and the impacts between couplers.Thus,existing models underestimate the additional dynamic variations in in-train forces.Detailed multi-body dynamic models of two AAR(Association of American Railroads)coupler systems used in passenger and freight trains are developed,incorporating coupler inertia and various slacks.Due to the modeling and simulation com-plexities involved in a full train model,with such details of coupler system,actual longitudinal train dynamics is not studied.A system comprising only two coupling units,inter-connecting two consecutive vehicles,is modeled.Considered system has been fixed at one end and an excitation force is applied at the other end,to mimic a relative force transmission through combined coupler system.Simulation results obtained from this representative system show that,noticeable influence in in-train forces are expected due to the combined effect of inertia of couplers and intermittent impacts between couplers in the slack regime.Maximum amplitude of longitudinal reaction force,transferred from draft gear housing to vehicle body,is expected to be significantly higher than that predicted using existing models of coupler system.It is also observed that the couplers and knuckles are subjected to significant longitudinal and lateral contact forces,due to the intermittent impacts between couplers.Thus,accurate estimation of draft gear reaction force and impact forces between couplers are essential to design vehicle and coupler components,respectively.

机车充风能力对重载列车缓解性能及车钩力影响研究

机车充气与排气性能是列车制动系统重要指标,目前机车制动系统验收指标仅对排气性能有明确要求,对充气性能无要求。机车供风能力不仅影响重载列车缓解特性和车钩力,而且严重制约重载列车在循环制动中的操纵方法。了解机车充气能力对缓解特性和车钩力的影响规律,以及重载列车安全运行和制动系统设计具有重要意义。通过建立列车空气制动系统仿真模型,分析机车充风能力对重载列车缓解特性和车钩力的影响。分析发现,机车充风能力对列车再充风时间、缓解波传播和车钩力都有明显影响;充风能力越弱,则缓解波传递越慢,车钩力越大。在本文研究范围内,合适的充气能力将比弱充风能力首尾车缓解时间差缩短3 s,最大车钩力降低16.2%。建议机车验收时增加机车充风能力检测,并给出了建议检测标准。针对重载列车充风能力,建议多部门联合系统性开展实验与仿真研究,制订重载列车制动系统检测标准与方法。

具有未知时变参数城轨列车的精准停车鲁棒控制

列车运行情况复杂多变,使得列车质量、基本运行阻力及线路状况等参数难以精确获得且随着列车运行存在明显的时变特性。列车模型参数的不确定性和时变特性会对列车的精准停车性能产生重要的影响。在动车组多质点动力学模型的基础上,根据Lyapunov再设计方法设计了一种非线性列车鲁棒控制器。该控制器将系统中的不确定性因素视为等价干扰,并设计相应的鲁棒补偿项。根据参数的先验区间,设计的鲁棒补偿项能够抑制模型估计误差,使得列车的位置和速度跟踪误差快速收敛,最终实现精准停车的控制目标;同时,也证明了系统的最终一致有界稳定性。最后,选取城市轨道交通场景进行仿真验证,仿真结果表明:本文提出的控制方法在列车模型参数未知且时变情景下也能实现对期望停车曲线的快速、精确追踪,表现出较强的鲁棒特性。

自动飞行员复诵指令生成方法研究

为了提高效率,降低培训成本并推广使用计算机来取代管制模拟机中的飞行员席位,采用集成学习的策略来生成飞行员复诵指令。选用5个大规模预训练语言模型进行微调,并使用K折交叉验证来筛选出性能较好的4个模型作为基础模型来构建集成学习模型。所构建的集成学习模型在管制指令数据集上取得在本领域中的最优效果。在通用的ROUGE(recall-oriented understudy for gisting evaluation)评价标准中,取得R_(OUGE-1)=0.998,R_(OUGE-2)=0.995,R_(OUGE-L)=0.998的最新效果。其中,R_(OUGE-1)关注参考文本与生成文本之间单个单词的匹配度,R_(OUGE-2)则关注两个连续单词的匹配度,R_(OUGE-L)则关注最长公共子序列的匹配度。为了克服通用指标在本领域的局限性,更准确地评估模型性能,针对生成的复诵指令提出一套基于关键词的评价标准。该评价指标准基于管制文本分词后的结果计算各个关键词指标来评估模型的效果。在基于关键词的评价标准下,所构建模型取得整体准确率为0.987的最优效果,对航空器呼号的复诵准确率达到0.998。

融合GPT和知识图谱的洪涝应急决策智能问答系统研究

为提高生成式预训练语言大模型(generative pre-trained transformer, GPT)的应急管理信息分析能力,以实现洪涝灾害应急处置过程中的在线辅助决策,提出融合GPT和知识图谱的应急决策智能问答系统(KG-GPT)。改进GPT架构以识别问题中的关键信息,利用知识图谱推理应急领域知识并生成具有逻辑性的回答;结合洪涝灾害的实际应急决策问答数据集并编制演练脚本,使用自动评估和专家评估方法将本系统与GPT进行对比实验。研究结果表明:该系统成功融合应急领域知识图谱和GPT模型,能够深刻理解问题的背景信息并生成流畅回答;与GPT相比,该系统可为决策者提供更快速准确的在线辅助决策工具。研究结果可提升洪涝灾害应急信息分析和决策效率。

工程训练模式下的火灾自动报警控制系统设计实践与探索

以工程训练中心的基础实践课程为主线,探索OBE模式下的火灾报警控制系统设计,即在基础实践技能篇的基础上,提出一种能够体现学生对电子元件、工具使用具有导向价值的实践项目,同时保持一定的可塑性和拓展性,设计一种基于AT89C51单片机的新型火灾自动报警装置,仿真实现了声光电一体化报警、烟雾浓度测量和温度测量显示等检测功能,进而培养学生的创新实践能力。

基于ASP-SAC算法的列车自动驾驶速度控制

随着经济建设的绿色转型以及人工智能的快速发展,城市轨道交通已成为居民日常出行的重要方式,在保障安全性、高效性和准点性的前提下,列车运行的节能性和舒适性需求也越来越被关注。合理的运行策略能够有效实现多种目标需求下的列车自动驾驶速度控制,强化学习作为一种智能决策方法,能够有效解决这一控制问题。首先,通过综合分析技术、安全性和乘客体验等方面的因素,基于专家经验动作划分和状态信息熵将软演员-评论家(SAC)改进为动作状态经验优先软演员-评论家(ASP-SAC)方法,用于研究列车自动驾驶速度控制问题。其次,将问题马尔可夫形式化,搭建了列车运行环境,确定了状态空间、动作空间以及基于目标控制的奖励函数。最后,以北京地铁亦庄线的一段区间数据为例进行试验,对ASP-SAC方法进行验证并与其他一些算法在相同环境下进行性能优劣比较。研究结果表明:该方法对于多目标控制需求下的列车自动驾驶速度控制问题具有可行性,与未改进前相比算法效率提高22.73%,与PPO算法相比提高29.17%,改进效果良好。同时,列车运行时在安全性、舒适性无误的情况下,准时性、精确性和节能性都强于SAC、DQN、PPO以及PID算法,其中能耗分别减少3.64%、5.62%、4.38%、7.35%,控制效果良好。此外,该方法亦具备鲁棒性,在列车自动驾驶速度控制方面具有一定的优越性和可参考性。

课证融通的web前端课程智能实训系统研究开发

《国家职业教育改革实施方案》提出全面实施1+X证书制度试点。专业课程要融入X证书内容,构建“课证”融通教学。传统web前端类课程实训教学存在实训内容未能融入考证内容、实训结果依赖人工评价等问题。为改进实训教学,引入浏览器内核技术,研究设计了一种web前端课程云实训系统,融入考证知识点,应用人工智能技术对上机实训操作进行实时智能结果评价。使用结果表明,系统发挥作用良好,有效提高了实训教学效率。

深度学习编译器模型训练负载均衡优化方法

对于计算密集型的人工智能(AI)训练应用,其计算图网络结构更加复杂,数据加载、计算图的任务划分以及任务调度的负载均衡性都会成为影响计算性能的关键因素。为了使深度学习编译器中模型训练应用的任务调度达到负载均衡的状态,提出了三种计算图负载均衡优化方法:第一,通过自动建立数据加载与模型训练的高效流水实现中央处理器和后端计算设备的负载均衡,提高了系统整体能效;第二,通过计算图的分层优化技术,实现计算图在后端设备执行调度时的负载均衡;最后,通过自动建立层间的高效流水提高后端设备的资源利用率。实验结果表明,计算图负载均衡优化方法实现了训练任务到底层硬件设备自动映射过程中系统的负载均衡,与Tensorflow、nGraph等传统的深度学习框架和编译器相比,在不同模型训练中通过任务调度负载均衡优化技术分别获得了2%~10%的性能提升,同时能够使系统整体的能耗降低10%以上。

基于语义特征融合的作文自动评分方法

作文自动评分技术是一种利用机器学习进行自然语言处理的技术。目前,基于深度学习的端到端模型在作文自动评分领域已经广泛使用。然而,由于端到端模型难以获取不同特征之间的相关性,因此提出一种基于语义特征融合的作文自动评分方法(TSEF)。该方法主要分为特征提取和特征融合2个阶段。特征提取阶段,使用Bert模型对输入文本进行预训练,并使用多头注意力机制对输入文本进行自训练,以补充预训练的不足;特征融合阶段,使用交叉融合方法将获取的不同特征融合,以此获得更好性能的模型。在实验中,将TSEF与许多强基线进行比较,结果表明了本文方法的有效性和稳健性。

城轨列车深度强化学习节能优化控制方法

为提高城轨列车自动驾驶(Automatic Train Operation,ATO)的控制性能,针对城轨ATO目标速度曲线追踪控制方法中工况切换频繁、牵引能耗高等问题,以列车准点、精准停车和能耗为优化目标,设计了一种以时间冗余(Time Redundancy,TR)规划参考系统为主动约束的列车深度强化学习DQN控制方法。建立了城轨列车动力学模型和多目标优化适应度函数;定义了TR规划参考系统约束下的DQN列车控制器,并对控制器中的动作空间和奖励函数进行设置;最后规定了列车控制器神经网络更新方法,利用随机梯度下降算法更新Q网络参数。结果表明:加入以TR时间规划参考系统为约束的TR-DQN算法提高了DQN迭代收敛速度以及迭代训练过程中的稳定性;TR-DQN动态调整列车运行策略的控制方法相比传统目标速度曲线追踪PID算法能耗降低12.32%,列车站间工况切换频率较低;针对设置的3种不同站间规划时间,列车牵引能耗依次降低7.5%和6.4%,列车站间工况动态切换频率和牵引能耗随行程规划时间增大而降低。

城轨列车自动驾驶积分反步线性自抗扰控制

针对列车在外部干扰和不确定动态下的速度控制问题,设计了融合积分反步法和线性自抗扰的复合控制方案.首先,考虑列车具有强耦合性,为了更加符合列车真实的纵向动力学特性和受力情况,建立了具有时变系数的多质点模型.其次,为了降低参数调节的难度,跟踪微分器和扩张状态观测器均采用线性形式.跟踪微分器用于求取微分信号,同时具有滤波作用.利用跟踪微分器对虚拟控制量进行求导,正好可解决反步法中存在的“微分爆炸”问题.扩张状态观测器用于实时估计总和扰动.此外,利用积分反步法改进了误差反馈控制律,设计了一种积分反步线性自抗扰控制(IBS-LADRC)算法.最后,证明了观测误差的收敛性及闭环系统稳定性.结合杭州地铁6号线AH型动车组参数和实际线路数据进行仿真,并将IBSLADRC与反步法、线性自抗扰算法、PID控制进行对比,结果表明:IBS-LADRC方法下各动力单元速度误差均处在±0.04 km/h以内,加速度处在±1 m/s2以内,加速度和速度误差均变化较平稳;车钩力相对其他3类方法最小,变化也最平缓,最大车钩力仅为2 320 N;本文控制策略对列车期望速度具有较高的跟踪精度,有利于保证车钩安全,防止车钩断裂,并提高列车运行的安全性、平稳性及乘客舒适度.

基于多尺度上下文的英文作文自动评分研究

目前作文自动评分模型缺乏对不同尺度上下文语义特征的提取,未能从句子级别计算与作文主题关联程度的特征。提出基于多尺度上下文的英文作文自动评分研究方法MSC。采用XLNet英文预训练模型提取原始作文文本单词嵌入和句嵌入,避免在处理长序列文本时无法准确捕捉到符合上下文语境的向量嵌入,提升动态向量语义表征质量,解决一词多义问题,并通过一维卷积模块提取不同尺度的短语级别嵌入。多尺度上下文网络通过结合内置自注意力简单循环单元和全局注意力机制,分别捕捉单词、短语和句子级别的作文高维潜在上下文语义关联关系,利用句向量与作文主题计算语义相似度提取篇章主题层次特征,将所有特征输入融合层通过线性层得到自动评分结果。在公开的标准英文作文评分数据集ASAP上的实验结果表明,MSC模型平均二次加权的Kappa值达到了80.5%,且在多个子集上取得了最佳效果,优于实验对比的深度学习自动评分模型,证明了MSC在英文作文自动评分任务上的有效性。

人在回路深度学习垂体分割模型的建立

目的:采用深度学习方法,通过人在回路的方式进行迭代式标注-训练,建立垂体分割模型,实现垂体体积人工智能(AI)测量。方法:将1285例颅脑3D T 1WI图像按5~15岁、16~25岁、26~50岁、51~70岁年龄段分组,每个年龄组随机选择80例,分成4批次进行试验。初始每组选择3例图像进行人工预标注神经垂体和腺垂体,输入计算机进行学习,获取初始模型。应用模型对一批数据进行分割,获得分割后的神经垂体、腺垂体与垂体总体积数据,将分割结果进行人工校准,获得校准后相对应的体积数据作为金标准。用前一组校准后的分割图像进行计算机迭代式学习优化模型,再用优化后模型对新一组图像分割与校准,重复上述过程,最终将校准前后差异没有统计学意义的数据认定深度学习建模成功。数据采用配对t检验、Dice和Spearman相关性分析进行统计。结果:从第2批次开始,除5~15岁年龄段外,其它年龄段神经垂体体积在校准前后的差异没有统计学意义,腺垂体与垂体总体积的差异有统计学意义(P<0.05)。第4批次,各年龄段神经垂体、腺垂体与垂体总体积在校准前后的差异均无统计学意义(P=0.137~0.928),Dice值大于0.941,Spearman相关系数大于0.969。结论:通过迭代式训练进行深度学习建模可构建垂体分割模型,实现垂体体积AI自动测量。