Contact State Space: Automatic Generation, Reasoning, and Search

在许多机器人操作任务中，特别是装配任务中，接触物体的运动是常见的，甚至是需要的，以减少不确定性。接触状态推理是处理接触运动的一个重要工具，也是本课题研究的重点。具体来说，本研究有两个主要目标：(1)研究离散接触状态和状态转换的适当表示；(2)研究根据接触状态转换自动规划接触运动的标准和技术。第一个主要研究目标回应了这样一个事实，即尽管在相关研究中提出了各种表示多面体离散接触状态的方法，但缺乏对基于某些重要标准定义/表示接触状态和状态转移的最佳方法的一般和系统研究，例如(i)描述/表示状态转移有多容易；（二）区分相邻的州有多容易。第一个准则是重要的，因为需要自动生成接触状态空间。第二个标准，处理表示的粒度，对于在不确定性存在的情况下自动识别接触状态的需要尤为重要。第二个主要研究目标与第一个目标密切相关，特别针对柔性装配任务。在追求这两个目标的过程中，理论概念将通过实验实现和执行得到验证。这将导致对最常见装配任务的物理学的基本理解，并将这些知识定量地与可在灵活装配中执行它们的规划过程联系起来。这种协作的成果将对自动装配接触运动计划的设计、评估、生成和执行产生重大影响。

In many robotic manipulation tasks, especially assembly tasks, motion of objects in contact is common and even desired in order to reduce uncertainties. An important tool in dealing with motion in contact is reasoning about contact states, which is the focus of this project. Specifically, this research has two major goals: (1) to investigate proper representation of discrete contact states and state transitions, and (2) to investigate criteria and techniques for automatic planning of contact motions in terms of contact state transitions. The first major research goal responds to the fact that although various ways of representing discrete contact states for polyhedra have been proposed in related research, there is a lack of general and systematic investigation on the best way(s) of defining/representing contact states and state transitions based on certain important criteria, such as (i) how easy it is to describe/represent state transitions, and (ii) how easy it is to distinguish adjacent states. The first criterion is important from the need of automatic generation of contact state space. The second criterion, addressing the granularity of representation, is particularly important from the need of automatic identification of contact states in the presence of uncertainties. The second major research goal is closely related to the first one and is particularly aimed at flexible assembly tasks. In pursuing both goals, theoretical concepts will be validated by experimental implementations and executions. This will result in a fundamental understanding of the physics of the most common assembly tasks and relate this knowledge quantitatively to the planning processes which can perform them in flexible assembly. This collaborative effort is expected to lead to results that will significantly impact the design, evaluation, generation and execution of conta ct-motion plans for automatic assembly in general.