【作者】 叶秀芬；

【导师】 李国斌；

【作者基本信息】 哈尔滨工程大学 ， 控制理论与控制工程， 2003， 博士

【摘要】 基于互联网的遥操作系统具有广阔的应用前景。但是，互联网的传输时延、时延抖动及带宽限制和带宽变化都对遥操作系统不利。目前，大多数都在开发新型先进的远程补偿控制算法及开发创新的遥操作接口界面，很少探究互联网数据传输的影响和特性。为此，本文研究基于互联网的遥操作系统中操作人员与远程设备之间数据传输的主要问题。 首先，根据开发新协议的约束要求及遥操作系统对数据传输的要求，提出了一种三项式拥塞控制机制。基于该机制，开发了基于速率的点对点三项式协议。在个体性能和容量方面，三项式协议提供了比当前TCP和UDP协议更好的传输服务；在稳态，它的传输速率是平滑的，并且当可用的网络带宽增加或减小时，能迅速而自适应地进行改变。在集体行为方面，它是易响应的，协议间公平(TCP-兼容)的，协议内部公平收敛及有效收敛的。论文对在理论上给出的协议特性都进行了广泛的仿真研究，仿真结果证实了理论的正确性及所提出协议的有效性。 其次，对网络时延(RTT)特性进行了分析，利用线性和非线性方法对从互联网上采集的RTT时间序列进行统计分析。分析表明RTT时间序列与一个常数加右偏的Γ分布极其相符合；在时间序列中相邻的及相近的时延之间是线性相关而不是不相关的；也可以看到在两个采集的RTT观测值之间只有微弱的或没有非线性相关性。因此RTT时间序列能用线性模型从过去的和当前的RTT数据来预测其下一个RTT的特性。 再次，给出了一种新的基于信息理论和最大熵原理(MEP)的变参数自适应RTT和RTO估计算法，在三项式协议实现过程中用于探测丢包及速率调整。与传统的参数固定的自回归滑动平均(ARMA)方法比较，MEP算法是自适应的并能够迅速动态地跟踪RTT。哈尔滨工程大学博士学位论文 最后，开发并实现了一个新型E一SerVice可移动机器人系统标准平台。这个平台采用了机器人控制和反馈信息的一个客户一服务器结构，最大的优点在于这种客户一服务器结构对于机器人系统底层细节是独立的，所以不需要大量的编程工作就很容易在这个平台上完成和测试新的遥控算法、接口界面设计及其应用。这个平台在互联网上已通过了实验的研究测试，实验中，在一个普通的PC机上通过可运行Java的网络阅览器，用户可以成功地通过互联网远程导航可移动机器人穿过一个典型的实验室环境。更多还原

【Abstract】 There are wide application areas for Internet-based teleoperation systems. However, Internet transmission time delays, delay jitter and bandwidth limitation and variation are unfavorable to the teleoperation system. To deal with these problems, the majority of current endeavours are devoted to developing various remote compensation control algorithms and interface techniques, without actively studying the effects and characteristics of Internet data transmission. Therefore, in this thesis, we study the main problems during data transportation between the operator and the remote device in the teleoperation system.Based on the constraints on new transport protocol and requirements of data transmission of teleoperation systems, we first present a novel congestion control algorithm. Based on this algorithm, an end-to-end rate-based transport protocol named the trinomial protocol is developed, In terms of individual performance and capacities, the trinomial protocol provides much better transmission service than current TCP or UDP protocols: it minimizes transmission delays and delay jitter by omitting re-transmission; in the steady state, its transmission rate is smooth; and when available network bandwidth increases/decreases, it adapts to the change quickly. Regarding social behaviors, it is responsive, inter-protocol fair (TCP-compatible), intra-protocol fairness convergent, and efficiency convergent. All these theoretical characteristics are conducted with extensive simulation studies. The simulation results validate the theories and conclusions.Next, the thesis analysis the characterization of Roundtrip time delay (RTT). The RTT time series collected from the Internet are studied statistically by using both linear and nonlinear methods. It is revealed that the RTT distribution is best fitted by a constant plus a "skewed right" gamma distribution. The adjacent and near-adjacent delays in the time series are linearly dependent rather than random. It is also indicated that there are only weak or no nonlinear correlations among RTT observations. Based on these findings, it is concluded that RTT time series can be characterized by linear models in estimating next RTT from past and current observations.Thirdly, a novel parameter-varying adaptive algorithm for RTT and RTO estimations based on the information theory and the maximum entropy principle (MEP) is presented. It is used in the implementation of trinomial protocol to detect packet losses and to adjust the sending rate. Compared to the traditional parameter-fixed autoregressive moving average (ARMA) method, the MEP algorithm is adaptive and capable of tracking RTT dynamics rapidly.Finally, a new modular platform for e-service mobile robotic systems is developed and implemented. This platform adopts a client-server architecture for robot control and feedback information display. The greatest benefit of this client-server architecture is that the applications are insulated from the lowest level details of the robot system. As a result, it is very easy to implement and test new advanced teleoperation control algorithms, interface designs and applications on this platform without large programming work. This platform is tested via experimental studies on the Internet. In the experiments, by using any Java-enabled web browsers, the users successfully navigated a mobile robot through a typical laboratory environment on a regular PC.更多还原