Object Oriented Robot Programming System Based on Knowledge of Handling Parts.

本研究有两个目的。一个是开发一个基于模型的视觉系统，使用的信息，如倾斜，距离和边缘的表面。二是利用装配零件知识和任务内容实现面向对象的机器人编程系统，在后一项研究中，通过应用实例阐明了编程系统的实用性;在视觉系统方面，提出了一种基于物体表面特征的物体识别和定位方法。此外，我们还介绍了一种适用于多种图像处理技术的分层可视化数据结构，并通过实验证明了这种结构的有效性。对于机器人程序设计系统，我们开发了一种并发面向对象语言THERA，并阐明了这种语言在描述多机器人任务程序方面的优势。使用这种语言，我们展示了一种方法来自动生成一个零件的装配树，并决定每个零件的抓取计划。此外，我们还介绍了一种机器人问题的解决方法，它决定了一个动作序列的基础上的初始状态，目标状态和知识的部分和任务。这个机器人问题解决系统利用时序逻辑和Meta知识的零件和任务的推理。

This research has two purposes. The one is to develop a model based vision system using information such as inclines, distances and edges of surfaces. The other is to implement an object oriented robot programming system utilizing the knowledge about assembly parts and contents of tasks.In the latter study, it is included to clarify the utility of the programming system by applying to some examples.As for as the vision system is concerned, we propose a method to recognize and locate a specified object for a pile of objects on the basis of surface characterization of the object. Moreover, we introduce a hierarchical visual data structure suitable for utilizing multiple picture processing techniques and experimentally show the usefulness of this structure.For the robot programming system, we develop a concurrent object oriented language 'THERA' and clarify this language good in describing task programs for multiple robots. Using this language, we show a method to automatically generate an assembly tree for parts and to decide a grasp plan for each part. We, moreover, introduce a method for robot problem solving which decides an action sequence on the basis of an initial state, a goal state and knowledge about parts and the task. This robot problem solving system utilizes a reasoning by temporal logic and a meta knowledge of parts and tasks.