Knotting task execution based a hand-rope relation

In this paper, we focus on relationship between the rope and fingers to implement a knot and estimate error occurred during execution of knotting task. In the previous work, the authors have proposed a method for making various knots from synthesis of basic operations called skills that are derived from analyzing the knotting processes. Each individual skill has an operation to handle the rope and conditions for executing and completing the task skill. By combining these skills, we can make the whole knotting process robust. In this paper, the ‘relations’, which describe the positions of the rope relative to the fingers, found from the successful cases, using the teaching data, were defined as the conditions of the task skills. Relation of rope and fingers at a specific state is presented by using propositional variables and logical connectives. Knotting skills were constructed to be executed robustly using the detected conditions. Occurrence of an error is estimated by finding out an abnormal relation. In this case, a corrective action can be inserted to modify the position of contact points, which are relative to the unsatisfied conditions to satisfy the detected conditions; therefore, skill execution is more reliable. We confirm the effectiveness of the methods through experiments.本文聚焦于绳索与手指的相对位置关系，旨在实现绳结打结操作，并对打结任务执行过程中产生的误差进行估计。 既往研究中，研究者提出了一类通过合成基础操作（skills，即技能）来制作各类绳结的方法：此类技能源自对打结流程的分析，每项独立技能均包含用于操控绳索的操作，以及执行并完成该技能任务所需的条件。通过组合此类技能，可使整体打结流程具备更强的鲁棒性。 本文中，我们借助教学数据集从成功的打结案例中提取出用于描述绳索相对于手指位置的「关系（relations）」，并将其定义为任务技能的执行条件。特定状态下绳索与手指的关系可通过命题变量（propositional variables）与逻辑连接词（logical connectives）进行表示。基于所提取的条件，可构建出可稳健执行的打结技能。通过识别异常的关系，可对误差的发生进行估计。在此场景下，可插入校正操作以调整接触点的位置——针对未满足的条件，将其调整至符合已检测到的关系要求，从而提升技能执行的可靠性。我们通过实验验证了所提方法的有效性。