基于模糊推理的潜艇舱室温度控制研究

针对复杂的具有不确定数学模型的潜艇舱室温度环境，利用模糊推理的设计思想，提出了一种参数自调整双层模糊控制器，用于舱室温度控制，并通过采用确定性模糊调整规则，简化了模糊控制器的设计，具有较强的适应性和鲁棒性。仿真结果表明，该控制器明显地改善了控制系统的性能。     

基于流量矩阵估计的路由推断算法

路由信息对网络而言非常重要。现有的路由推断技术很难适用于任意网络。本文提出了一种路由推断算法,在已知网络的拓扑结构和链路流量的情况下,先通过期望最大化算法对一个包含了多种路由可能性的流量矩阵进行估计,然后根据对应的可能需求的估计值的差异来推断实际路由。仿真结果证明了方法的有效性。     

船舶核动力装置运行管理专家系统

本文通过对船舶核动力装置二回路凝给水系统运行规则的理解与分析，对运行经验、实行实例的搜集、整理，建立起了多元分层知识体系结构，根据不同知识的特点建立多种知识表示方式，采用多路推理机制，使系统很好地克服了传统专家系统所面临的某些不足，在一定程度上实现子模拟人类思维方式进行决策和解决问题，从而达到故障及时诊断和系统运行管理的目的。     

基于知识库的船舶管系定额工时测算研究

为了解决船舶制造企业管系定额工时测算效率低、误差大、难以指导生产计划编排及成本控制等问题,依据“设计、生产、管理、信息一体化”的思想,在深入分析管系设计数据特征的基础上,构建设计物量自动化抽取模型,并以管系定额工时标准体系为依据建立工时测算规则,运用知识库推理逻辑进行定额工时精确测算,最后通过系统研发与企业实际应用,验证抽取模型及测算方法的有效性.     

基于多层事件融合的场景事件实时分析

为解决智能监控场景中场景事件实时分析问题,提出了一种基于多层事件融合的场景事件分析模型以及对应的RBPF实时推理方法.利用事件具有层次结构特性以及各层事件之间的关系,将场景事件分解为不同层次的子事件,利用多层子事件融合进行场景事件分析,并用多层动态贝叶斯网络模型对其建模.构建模型对应的RBPF推理方法,以实现对复杂场景事件进行实时分析.仿真实验证明了该方法能够对动态场景中的场景事件进行实时推理,比PF方法具有更高的精度及较少的时间代价.     

基于长下坡路段车速与制动器温度实时监测的模糊推理预警算法

提出一种实时的车辆长下坡路段车速与制动器温度预警算法.建立长下坡路段的整车纵向力平衡方程和能量方程,分析制动器耗散能量占总能量的比例,研究车速对制动器耗散能量大小的影响,结合制动器吸收能量占制动器耗散能量的比例经验公式,建立制动器温升计算模型;基于试验数据,采用最小二乘法确定模型中的待定系数,比较模型计算的温升与试验数据,最大的均方根误差为12.1℃,对应车速为38 km/h,最小均方根误差为3.7℃,对应车速为50 km/h.安全车速根据安全制动距离和路面纵坡计算得出.预警算法依据车速和制动器温度变化,构造模糊推理系统计算车辆危险指数,综合评价车辆下长坡的危险程度.     

基于贝叶斯网络的出行方式选择模型研究

论文以居民出行方式选择为研究对象,分析了城市居民出行方式的影响因素,建立了基于贝叶斯网络的居民出行方式选择模型,并以苏州市为例,结合居民出行调查数据采用极大似然法对模型进行了参数估计,并采用贝叶斯网络推理方法验证模型精度。结果表明,该模型能较全面地考虑居民出行选择的影响因素,模型精度较高。     

Learning User Intention in Networked Mobile Robot Control

Teleoperated networked robot often has unpredictable behaviors due to uncertain time delay from data transmission over Internet. The robot cannot accomplish the desired actions of the remote operator in time, which severely impairs reliability and efficiency of the robot system. This paper investigated a novel approach, learning user intention, to compensate the uncertain time delay with the autonomy of a mobile robot. The user intention to control and operate the robot was modeled and incrementally inferred based on Bayesian techniques so that the desired actions could be recognized and completed by the robot autonomously. Thus the networked robot is able to fulfill the task assigned without frequent interaction with the user, which decreases data transmission and improves the efficiency of the whole system. Experimental results show the validity and feasibility of the proposed method.     

基于理性推理的观点演化模型

针对现实生活中个体观点形成中的理性推理过程,提出了一种基于DS（Dempster-Shafer）证据理论的观点更新规则,并在连续观点离散决策的框架下,对个体如何利用其他个体的意见和相关知识来形成自身意见的过程进行建模.实验仿真结果表明,群体中能够出现观点一致、分散等常见的舆论现象,观点演化过程中伴随着知识由确定性个体向不确定性个体的扩散,同时发现在规则网络中,意见领袖的观点的影响力在扩散过程中是逐渐减弱的,其影响范围与个体对知识不确定性的接受程度密切相关.     

An integrated solution for lane level irregular driving detection on highways

Global Navigation Satellite Systems (GNSS) has been widely used in the provision of Intelligent Transportation System (ITS) services. Current meter level system availability can fulfill the road level applications, such as route guide, fleet management and traffic control. However, meter level of system performance is not sufficient for the advanced safety applications. These lane level safety applications requires centimeter/decimeter positioning accuracy, with high integrity, continuity and availability include lane control, collision avoidance and intelligent speed assistance, etc. Detecting lane level irregular driving behavior is the basic requirement for these safety related ITS applications. The two major issues involved in the lane level irregular driving identification are accessing to high accuracy positioning and vehicle dynamic parameters and extraction of erratic driving behaviour from this and other related information. This paper proposes an integrated solution for the lane level irregular driving detection. Access to high accuracy positioning is enabled by GNSS and Inertial Navigation System (INS) integration using filtering with precise vehicle motion models and lane information. The detection of different types of irregular driving behaviour is based on the application of a Fuzzy Inference System (FIS). The evaluation of the designed integrated systems in the field test shows that 0.5 m accuracy positioning source is required for lane level irregular driving detection algorithm and the designed system can detect irregular driving styles.