CAREER: Towards General-Purpose Manipulation of Deformable Objects through Control and Motion Planning with Distance Constraints

该项目旨在提高自主机器人在工厂、医院和家庭等环境中的能力。为了在这样的环境中发挥作用，机器人需要操纵多种可变形的物体，比如布、绳子或电线，以及植物或动物组织。赋予机器人操纵这些物体的能力有可能彻底改变机器人在几个应用领域的使用：在制造业中，允许机器人包装盒子并处理食品和织物；在医学领域，允许机器人执行繁琐的手术任务和铺床；在服务中，允许机器人处理衣服和准备食物。然而，这样的任务超出了目前的技术水平，主要是因为可变形物体难以建模和控制。该项目将发展新的理论，并通过广泛的实验验证，旨在以可变形物体的方式建模，以适应规划和控制。该项目还将使用可变形物体的操作，作为学生在本科和研究生阶段以及K-12课程中学习机器人的激励问题。这个项目研究了可变形物体的建模、控制和运动规划的基本原理。这个问题是具有挑战性的，因为可变形对象具有很大的状态空间，是极度欠驱动的，并且这些对象的运动规划需要数值模拟，这可能是计算昂贵和不准确的。这个项目的主要假设是，这些挑战可以通过用距离约束来表示对象和任务来克服，距离约束表示可变形对象部分之间以及对象与环境部分之间的允许距离，并根据这种表示来制定控制和规划方法。本研究的预期结果是增强机器人执行有用任务的能力，基于对可变形对象操作的更好理解和一组广泛适用的实验验证控制器，规划构建块和规划算法。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project seeks to advance the capabilities of autonomous robots in environments such as factories, hospitals, and homes. To be effective in such settings, robots need to manipulate many kinds of deformable objects, such as cloth, string or wires, and plant or animal tissue. Giving robots the ability to manipulate such objects has the potential to revolutionize the use of robots in several application domains: In manufacturing, allowing robots to pack boxes and handle food and fabric; in medicine, allowing robots to perform tedious tasks in surgery and make hospital beds; and in service, allowing robots to handle clothes and prepare food. However, such tasks are beyond the current state-of-the-art, largely because deformable objects are difficult to model and control. This project will develop new theory, validated with extensive experimentation, that aims to model deformable objects in ways that are amenable to both planning and control. This project will also use manipulation of deformable objects as a motivating problem for students studying robotics at the undergraduate and graduate levels, and in K-12 outreach.This project investigates the fundamentals of modeling, control, and motion planning for deformable objects. This problem is challenging because deformable objects have large state spaces, are extremely-underactuated, and motion planning for these objects requires numerical simulation, which can be computationally expensive and inaccurate. The main hypothesis of this project is that these challenges can be overcome by representing the object and task in terms of distance constraints, which represent allowable distances between parts of the deformable object as well as between the object and parts of the environment, and formulating control and planning methods based on this representation. The expected outcome of this research is an enhanced ability of robots to perform useful tasks, based on a better understanding of deformable object manipulation and a set of broadly-applicable experimentally-verified controllers, planning building-blocks, and planning algorithms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.