Robotic articulated hand for modeling and dexterous manipulation of deformable objects

目前的机器人系统被编程为处理刚性物体或考虑为虚拟现实应用而设计的可变形物体，在虚拟现实应用中，弹性的实际物理参数被硬编码。然而，大量的日常生活物体实际上是非刚性的，因此不适合标准机器人的自动操作。这项研究的主要目标是开发一种综合的方法来检测、测量、建模和与特性未知的可变形对象进行交互。这是通过开发基于非刚性对象的属性的物理测量的用于可变形对象的传感策略以及表示和操纵方案来实现的。设计了一个自主机器人操作系统，包括视觉传感器定位、自动选择用于触觉探测的候选点、力/变形特性测量、将体积和变形特性合并到可变形物体的扩展表示中，以及力反馈控制方案的调整。其目标是用一只装有力和触觉传感器的灵巧机器人手，安全稳定地操纵可变形的物体。在探索性机器人、危险材料搬运、自动化温室和辅助手术中的应用也被考虑。

Current robotic systems are programmed to handle either rigid objects or consider deformable objects designed for virtual reality applications where the actual physical parameters of elasticity are hardcoded. However, a very large number of everyday life objects are actually non-rigid, and are, therefore, not appropriate for automated manipulation with a standard robot. The main objective of this research is the development of integrated approaches to detect, measure, model and interact with deformable objects whose characteristics are a priori unknown. This is achieved through the development of sensing strategies as well as representation and manipulation schemes for deformable objects that are based on physical measurements of the properties of the non-rigid objects. An autonomous robotic manipulation system is designed following a sequential approach comprised of localization with vision sensors, automated selection of candidate points for tactile probing, force/deformation characteristics measurement, merging of volumetric and deformation properties into an extended representation of deformable objects, and tuning of a force feedback control scheme. The goal is to produce safe and stable manipulation of deformable objects with a dexterous robotic hand equipped with force and tactile sensors. Applications in exploratory robotics, hazardous material handling, automated greenhouses, and assisted surgery are considered.