【作者】 李焱；

【导师】 贺汉根；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2002， 博士

【摘要】 大时延遥操作技术是空间机器人领域的一项关键技术。大时延遥操作的一个核心问题是如何在大时延和有限带宽的条件下，保证遥操作的平稳性，同时提高遥操作的透明性。为了解决这一问题，本文的基本思想是利用机器人的局部自主能力，把时间延迟排除在底层控制回路之外，保证遥操作的平稳性；同时，利用基于虚拟现实的人机接口技术提高遥操作的透明性。碰撞检测技术是虚拟现实应用中的一项关键技术，对于复杂虚拟环境，碰撞检测往往成为系统的计算瓶颈。本文针对遥科学实验应用，对多体快速碰撞检测技术进行了研究。本文的主要研究内容分为三个部分：大时延遥操作系统框架研究、遥编程技术研究，以及快速碰撞检测技术研究。 为了建立灵活的人机协调机制，本文基于监控思想，提出了一个基于Petri网的大时延遥操作系统框架。该框架涵盖了任务建模、任务协调，以及任务监控等几个方面；本文研究了基于Petri网的的任务建模方法。针对自由运动，提出了一个基于路径关键点提取的动作级任务建模方法；针对复杂任务，提出了一个根据子任务的Petri网系统模型，通过变迁细化方法建立任务的分层Petri网模型的方法；本文提出了一个基于Petri网的任务协调方法，该方法根据任务空间的谓词/变迁网描述，通过状态识别和动作识别，实现了从操作员的连续动作到任务的动作序列的转化。 本文对遥编程所涉及的一些关键技术，包括：虚拟预测仿真、程序自动生成，以及误差校正，进行了研究。本文实现了一个遥编程系统，并基于遥编程建立了一个大时延遥操作试验系统。进行了典型的大时延遥操作仿真试验，成功完成了高精度遥操作任务。 本文研究了基于空间分割的快速碰撞检测方法，其中包括：基于均匀空间分割的快速碰撞检测和基于非均匀空间分割的快速碰撞检测。提出了一种基于均匀空间分割的快速碰撞检测算法USSCD，与经典的ICOLLIDE算法进行了对比试验，试验结果表明，在物体均匀分布的情况下，随物体数量的增多，USSCD表现出明显的优势；提出了一类非均匀空间分割方法，在该方法中，空间分割由基于密度的非均匀空间分割和对每个子空间进行均匀分割两个步骤实现；提出了一种基于投影的非均匀空间分割方法。在该方法中，通过物体的轴向投影直方图表征空间分布密度，通过区域生长实现空间分割；提出了一种基于聚类的自适应空间分割方法。该方法由两个部分构成，首先，通过一个基于局部密度的聚类算法对物体进行聚类；然后，根据聚类结果，通过简化分割方法实现空间分割。更多还原

【Abstract】 Teleoperation through large time delay (T3D) is one of the key techniques in the field of space robot. One of the core problems in T3D is how to maintain the stability and to promote the transparency of the teleoperation system,on the condition of large time delay and limited bandwidth in the communication link between the operator control station on the earth and the robot workstation in space. This paper presents two basic points to cope with this problem. First,the impact of large time delay in the servo-control loop is avoided through the local automation of the telerobot,therefore,the stability is maintained. Second,the transparency is promoted by developing the advanced human-computer interface based on Virtual Reality (VR). Collision detection is one of the key techniques in the application of VR,which usually presents as the bottleneck of the computing performance in complex virtual environment. With the background of telescience experiment application,this paper concentrates on the research of accelerated algorithm for N-Body collision detection. The main contents in this paper include three aspects:framework for T3D system,teleprogramming,and accelerated collision detection methods.To develop a flexible coordination mechanism between the human operator and the machine,a Petri Nets based unified system framework for T3D is proposed,which,follows the concept of supervisory control,and covers the main aspects of the system,such as task modeling,task coordination,and task supervision. Under this framework,the task modeling methods based on Petri Nets are studied. An action-level task modeling method for free move is proposed,which is based on a method of key point extraction from the trajectory. A hierarchical task modeling method is proposed,which constructs the hierarchical model through the transition-refinement according to the Petri Nets system model of the subtasks. A task coordination method is proposed,which,according to the Predicates Transition Nets representation of the task space,through state recognization and action recognization,transacts the continuous actions of the operator to a series of instructions for the telerobot.In this paper,several key techniques relative to teleprogramming are studied,including:predective display through virtual environment,automatic program generation,and error corrtection of the virtual environment. A teleprogramming based experimental T3D system is developed. The typical peg-in-hole experiment is conducted,and T3D is simulated by artificial delay setup. The system is succeeded in the part-mating task,which proves the validity of the teleprogramming system for high-precision task.In this paper,a set of accelerated collision detection algorithms based on space subdivision are developed. One accelerated N-Body collision detection algorithm based on uniform spacial subdivision,called USSCD,is proposed. The performance of USSCD is compared with the well known algorithm,ICOLLIDE,through a series of experiments. Theresults show that,when the objects are evenly distributed,and the number of objects islarge,the performance of USSCD is better than that of ICOLLIDE. In this paper,a class of non-uniform spacial subdivision method is proposed,which is composed of two steps:first,the space is coarsely subdivided non-uniformly,based on the distribution density of the objects,then,every subspace is uniformly subdivided. A projection based non-uniform spacial subdivision method is proposed,in which the distribution density is measured by the histgram of the objects projection on three coordination axis,and the space is subdivided through a ’region-grown’ algorithm. An adaptive space subdivision method based on clustering is proposed,which completes space subdivision through two steps:first the objects are classified through a new local-density based clustering algorithm,then the space is subdivided according to the clustering.更多还原