广义预测控制方法在网络遥操作机器人系统中的应用

Internet网络的时变时延及网络数据丢包严重影响了遥操作机器人系统的操作性能,甚至造成系统不稳定。为了解决这一问题,提出一种新的基于Internet的遥操作机器人系统控制结构。通过在主端对给定信息加入时间标签获得过去的系统回路时延,采用多元线性回归算法,预测下一时刻系统回路时延,然后在从端设计一个广义预测控制器控制远端机器人,从而改善时变时延对系统性能的影响。应用广义预测控制器产生的冗余控制信息,降低了网络数据丢包对系统的影响。最后根据预测控制稳定性定理,推导出系统的稳定性条件。仿真试验结果表明,该方法能有效解决时变时延以及网络数据丢包引起的性能下降问题。

基于模型差分析的旋翼飞行机器人航向与垂向预测控制方法

本文针对旋翼飞行机器人全包线机动飞行中的驱动器滞后以及动力学模型时变的问题,提出了应对不确定性动力学模型的基于模型差分析的增量平稳预测控制方法。该方法首先通过建立增量平稳预测过程模型来应对驱动器输出滞后与稳态模型以及系统工作点的不确定性,并提升控制系统鲁棒性。然后通过自适应集员滤波器在线估计系统瞬态动力学与名义模型的偏差来补偿全包线飞行中时变模型对于名义控制器跟踪性能的影响。最后,通过实际的飞行试验验证了此方法能够有效的解决全包线飞行中航向与垂向的驱动器滞后与动力学时变问题,并且可以实用于旋翼机器人航向与垂向的全包线自主飞行控制。

基于主动视觉和超声的机器人目标跟踪系统

运动目标跟踪技术是未知环境下移动机器人研究领域的一个重要研究方向。该文提出了一种基于主动视觉和超声信息的移动机器人运动目标跟踪设计方法,利用一台SONY EV-D31彩色摄像机、自主研制的摄像机控制模块、图像采集与处理单元等构建了主动视觉系统。移动机器人采用了基于行为的分布式控制体系结构,利用主动视觉锁定运动目标,通过超声系统感知外部环境信息,能在未知的、动态的、非结构化复杂环境中可靠地跟踪运动目标。实验表明机器人具有较高的鲁棒性,运动目标跟踪系统运行可靠。

基于多传感器的可变形机器人自主控制方法研究

将GPS、电子罗盘、倾角仪、码盘传感器等应用到可变形机器人自主运动控制中.针对可变形机器人自身结构特点,提出了一种基于多传感器信息融合的可变形机器人在野外环境中自主控制的方法.该方法主要实现了在非结构环境中机器人的自主变形、自主避障和自主导航定位等功能.实验验证了该方法的有效性.

基于视觉和超声信息的机器人行为控制

随着机器人应用范围的不断扩展,机器人所面临的工作环境也越来越复杂,多数是未知的、动态的和非结构化的。通过对基于行为的机器人控制技术的研究,设计了一种用于完成多目标任务的移动机器人行为控制系统。将基于行为的控制技术融合进模糊控制的思想中,使移动机器人的行为通过运用模糊控制和基于优先度的行为决策来实现,并且通过视觉信息使机器人能够完成面向目标的任务。

具有实时性保障的INTERNET机器人控制系统设计方法

本文提出了一种具有实时性、可靠性保障的INTERNET网络机器人控制系统的设计方法。基于该方法设计的网络实时控制系统能够满足机器人实时、高效、灵活的技术特点。该方法的核心为基于UDP传输协议的网络数据补偿算法,通过对网络传输过程中丢失的数据进行实时在线补偿预测,降低了网络数据的丢失对系统的影响。实验结果证明该方法的有效性、合理性。

基于加速度反馈的柔性关节机械臂接触力控制

针对柔性关节机械臂从自由空间运动控制过渡到约束空间力控制的过程中,存在冲击、震荡甚至不稳定等问题,利用加速度传感器反馈控制,为柔性关节机械臂的接触力控制在较宽的带宽内提供阻尼,克服了利用单纯速度反馈控制带宽窄的局限。对柔性关节机械臂的接触力控制进行建模和基于加速度反馈的控制策略分析,并在柔性关节机械臂上进行了接触力控制的试验研究。结果表明,这种方法有效。

基于加速度传感器的丝杠传动系统建模及控制研究

提出了基于加速度传感器的丝杠传动系统建模及控制策略 ,可以通过实验测量存在结构非刚性因素时丝杠系统的模型。基于该模型可以进一步设计基于加速度传感器的控制策略 ,加速度反馈可以抑制系统的扰动力对丝杠传动系统控制精度的影响 ,实验结果表明提出的方法可以有效地提高丝杠系统的性能。

基于循环抑制CPG的蛇形机器人运动控制方法

蛇具有细长无肢的身体、独特的半球形关节，使其可在神经系统控制下完成与环境相适应的多种节律运动。模仿蛇的运动机理和行为方式而设计的蛇形机器人克服了轮腿式机器人的缺点，增加了机器人的运动方式，扩大了机器人的应用范围。但应用传统的控制策略实现蛇形机器人运动控制遇到了很难克服的问题。随着社会经济与科技的发展，研究人员把从蛇运动神经系统研究中得到的启示应用到蛇形机器人上，希望不仅可以解决其运动控制问题，更能在构型、步态及控制机制上皆可展示蛇的特征。 生物学家已经证明动物的节律运动是其低级神经中枢的自激行为，是由中枢模式发生器(Central Pattern Generator，CPG)控制的。中枢模式发生器是一种能够在缺乏有规律的感知和中枢控制输入的情况下，产生有节奏模式输出的神经网络。 本文以国家自然科学基金课题《基于CPG的蛇形机器人控制方法研究》和国家“863”高技术计划资助项目《具有环境适应能力的蛇形机器人的研究》为依托，突破以相互抑制机理研究CPG的传统观点，首次创新性地提出应用循环抑制(Cyclic Inhibition, CI)机理来研究蛇形机器人的CPG建模与实现问题。本研究涵概了神经元模型的特性分析、蛇形机器人关节循环抑制CPG建模理论、蛇形机器人循环抑制CPG神经网络稳定性分析以及典型步态的生成方法、循环抑制CPG神经网络控制蛇形机器人蜿蜒运动参数设定策略、应用动力学仿真和实验对该CPG控制方法有效性的验证。 首先，本文介绍了两个用于CPG建模研究的蛇形机器人“勘查者”和“勘查者-I”。给出各自机械系统、控制系统的构成和动力学仿真平台。 其次，详细分析了神经元以及传统的相互抑制(Mutual Inhibition, MI)CPG的特性。从工程角度首次创新性地应用循环抑制建模理论构建了蛇形机器人CPG模型，并对其稳定性进行了深入的分析。首次证明持续型神经元构成的单向循环抑制(Unilateral Cyclic Inhibition, UCI) CPG是能产生振荡输出CPG中微分方程数量最少的，而且其产生振荡输出的机理完全不同于传统的相互抑制CPG。其不需要具备调整功能，只需要神经元之间强的单向循环抑制连接。 第三，首次应用单向激励连接循环抑制CPG构成蛇形机器人神经网络系统。分析了其稳定性，给出其产生振荡输出的条件。通过仿真和实验验证了循环抑制CPG神经网络实现典型步态(蜿蜒运动、伸缩运动和侧向运动)的有效性。首次应用双向循环抑制(Bidirectional Cyclic Inhibition, BCI)CPG神经网络在不同高级控制神经元命令激活下的输出实现蛇形机器人典型运动步态之间的转换。为蛇节律运动生成机制建模提供了新方法。 最后，从实时性、控制方便性等工程应用的角度，对单向循环抑制CPG神经网络实现蛇形机器人蜿蜒运动控制进行了深入的分析。给出了S-波形、幅值、运动速度和运动轨迹曲率的参数设定策略。该系统应用首CPG自激励权重调解成功解决了传统CPG控制系统中CPG的个数比蛇形机器人关节数多一个的问题，并用其实现了一种独特的转弯控制策略。 综上，为蛇形机器人运动控制提供了全新的方法。

加速度反馈控制抑制谐波传动系统振动的研究

在分析谐波传动系统特性的基础上，建立了谐波传动系统的基于加速度传感器反馈控制的数学模型，通过仿真分析和实验研究结果表明，加速度反馈控制能有效地换制谐波传动系统负载端的振动。

阿片药物依赖的神经机制---内侧前额叶与药物的心理依赖

Mental dependence, characterized by craving and impulsive seeking behavior, is the matter of intensive study in the field of drug addiction. The mesolimbic dopamine system has been suggested to play an important role in rewarding of drugs and relapse. Although chronic drug use can induce neuroadaptations of the mesolimbic system and changes of drug reinforcement, these mechanisms cannot fully account for the craving and the compulsive drug-using behavior of addicts. Acknowledging the reinforcement effects of drugs, most previous studies have studied the impact of environmental cues and conditioned learning on addiction behavior, often using established classical or operant conditioning model. These studies, however, paid little attention to the role of cognitive control and emotion in addiction. These mental factors that are believed to have an important influence on conditioned learning. The medial prefrontal cortex (mPFC) has close anatomic and functional connections with the mesolimbic dopamine system. A number of the cognitive neurological studies demonstrate that mPFC is involved in motivation, emotional regulation, monitoring of responses and other executive functions. Thus we speculated that the function of abnormality in mPFC following chronic drug use would cause related to the abnormal behavior in addicts including impulse and emotional changes. In the present study of a series of experiments, we used functional magnetic resonance imaging to examine the hemodynamic response of the mPFC and related circuits to various cognitive and emotional stimuli in heroin addicts and to explore the underlying dopamine neuromechnism by microinjection of tool drugs into the mPFC in laboratory animals. In the first experiment, we found that heroin patients, relative to the normal controls, took a much shorter time and committed more errors in completing the more demanding of cognitive regulation in the reverse condition of the task, while the neural activity in anterior cingulate cortex (ACC) was attenuated. In the second experiment, the scores of the heroin patients in self-rating depression scale (SDS) and Self-rating anxiety scale (SAS) were significantly higher than the normal controls and they rated the negative pictures more aversive than the normal controls. Being congruent with the behavioral results, hemodynamic response to negative pictures showed significant difference between the two groups in bilateral ventral mPFC (VMPFC), amygdala, and right thalamus. The VMPFC of patients showed increased activation than normal controls, whereas activation in the amygdala of patients was weaker than that in normal subjects. Our third experiment showed that microinjection of D1 receptor agonist SKF38393 into the mPFC of rats decreased hyperactivity, which was induced by morphine injection, in contrast, D1 receptor antagonist SCH23390 increased the hyperactivity, These findings suggest: (1) The behavior and neural activity in ACC of addicts changed in chronic drug users. Their impulsive behavior might result from the abnormal neural activity in the mPFC especially the ACC. (2) Heroine patients were more depress and anxiety than normal controls. The dysfunction of the mPFC---amygdala circuit of heroine addicts might be related to the abnormal emotion response. (3) Dopamine in the mPFC has an inhibitory effect on morphine induced behavior. The hyperactivity induced by chronic morphine was reduced by dopamine increase with D1 receptor agonist, confirm the first experiment that the neuroadaption of mPFC system induced by chronic morphine administration appears to be the substrate the impulse behavior of drug users.

获取无意识结构知识的机制:主观熟悉感与刺激重复结构在人工语法学习中的作用

The nature of the distinction between conscious and unconscious knowledge is a core issue in the implicit learning field. Furthermore, the phenomenological experience associated with having knowledge is central to the conscious or unconscious status of that knowledge. Consistently, Dienes and Scott (2005) measured the conscious or unconscious status of structure knowledge using subjective measures. Believing that one is purely guessing when in fact one knows indicates unconscious knowledge. But unconscious structural knowledge can also be associated with feelings of intuition or familiarity. In this thesis, we explored whether phenomenological feelings, like familiarity, associated with unconscious structural knowledge could be used, paradoxically, to exert conscious control over the use of the knowledge, and whether people could obtain repetition structure knowledge. We also investigated the neural correlates of awareness of knowing, as measured phenomenologically. In study one, subjects were trained on two grammars and then asked to endorse strings from only one of the grammars. Subjects also rated how familiar each string they felt and reported whether or not they used familiarity to make their grammaticality judgment. We found subjects could endorse the strings of just one grammar and ignore the strings from the other. Importantly, when subjects said they were using familiarity, the rated familiarity for test strings consistent with their chosen grammar was greater than that for strings from the other grammar. Familiarity, subjectively defined, is sensitive to intentions and can play a key role in strategic control. In study two, we manipulated the structural characteristic of stings and explored whether participants could learn repetition structures in the grammatical strings. We measured phenomenology again and also ERPs. Deviant letters of ungrammatical strings violating the repetition structure elicited the N2 component; we took this to be an indication of knowledge, whether conscious or not. Strings which were attributed to conscious categories (rules and recollection) rather than phenomenology associated with unconscious structural knowledge (guessing, intuition and familiarity) elicited the P300 component. Different waveforms provided evidence for the neural correlates of different phenomenologies associated with knowledge of an artificial grammar.