Studies on cooperative systems of multiple robots by use of knowledge concerned with task-context and task-circumstance

多机器人协作的研究分为以下三个层次：(1)任务规划层，(2)识别与建模层，(3)机械手控制层。(1)开发了一种利用时间逻辑获得执行子目标的任务序列的方法，并在此基础上开发了一种新的高效的多机器人规划系统。其次，针对基于AND/OR图的机器人装配规划，提出了利用零件知识和装配解析过程自动生成规划的方法以及高效的树搜索方法。(2)提出了利用曲面信息分离物体表面的方法和三维曲面物体的识别方法。提出了一种基于目标显著性自动提取的三维目标建模方法。介绍了利用分层六边形图像结构集成原始图像处理及其应用。(3)利用基于基本模糊规则的力控制算法，建立了多机械臂协作任务的任务级控制规则，并将这些规则应用于一些协作任务，结果表明了该方法的有效性。这种基于规则的多机械臂控制方法是由不同层次的规则分层构建的，具有很强的灵活性，适用于复杂的任务。

Researches on cooperation of multiple robots have been done corresponding to following three levels;(1) task planning level,(2) recognition and modeling level and (3) manipulator control level.(1) Developing a method of obtaining task-sequences for performing subgools by use of temporal logic,we developed a new and efficient planning system for multiple robots on the basis of this method. Scondly,as for as robot assembly planning based on AND/OR graph is concerned, automatic generation of planning by use of the knowledge of parts and assembly-resolution processes and efficient search method of the tree were developed.(2) A separation method of surfaces of an object by use of surfaces in formation and a recognition method of 3-D curved objects were proposed.A modeling method of 3-D objects wasalso proposed on the basis of extracting saliency of objects automatically. Integration of primitive image processings and its applications by use of hierarchical hexagonal image structure were also performed.(3) Control rules at a task level for cooperative tasks with multiple manipulators were developed by use of force control algorithms based on fundamental fuzzy rules, and resultsof applying these rules to some cooperative tasks revealed the usefulness of this approach. This rule based control method of multiple manipulators is hierarchically constructed byrules belongs to some levels, and is seemed to be very flexible and applicable to complicated tasks.