RI: Small: From Impact to Impulsive Manipulation

当人们工作时，他们经常利用物体之间的碰撞，例如，工人用锤子撞击钉子以将其打入。 这些撞击产生了所谓的“冲力”来做功。 人们也会利用较小的冲动力来完成许多常见的日常任务：将钥匙插入钥匙孔，打开卡住的窗户，以及打碎鸡蛋。 虽然盘子和桌子之间的碰撞比钉子轻得多，但它同样有用。今天的机器人还没有被设计成利用冲力来完成任务。 PI进行了一些基本的和实验性的问题，周围的脉冲机器人操作的调查，并开发方法，用于机器人控制。 项目成果正在推进多体碰撞理论，并扩大机器人的能力。 该研究成果的直接应用将为民用和军用机器人带来更快、更敏捷的速度，并提高工业过程的效率。该项目的技术目标是深入了解碰撞结果的控制，更重要的是，将碰撞规划与机器人操作中的运动规划相结合。 该项目包括三个阶段。 第一阶段推进了首席研究员最近关于顺应性和多重影响的工作，以模拟n体同时碰撞，解决包括状态空间凝聚，高维空间中的脉冲增长，对物理参数的敏感性以及区域和非线性接触等问题。 在并行的分析是一个名为MultiCollide的交互式碰撞模拟的图形界面的发展。 第二阶段研究如何机械手可以通过影响所需的运动传递到一个对象，解决一个逆冲击问题，连接冲击动力学与轨迹规划和机械手动力学。 最后一个阶段的重点是两个脉冲操作任务，分别解决以下重要问题：如何利用接触的顺应性和几何形状，以及如何交错的影响规划与机器人运动规划。

When people work, they often take advantage of impacts between objects, for example, a worker impacts a nail with a hammer to drive it in. These impacts generate what are known as "impulsive forces" to do work. People take advantage of smaller impulsive forces to complete many common daily tasks too; inserting a key into a key hole, opening a stuck window, and cracking an egg. While the impacts between a plate and a table is much lighter than that applied to a nail, it is nonetheless just as useful. Today's robots have not been designed to take advantage of impulsive forces to accomplish tasks. The PI conducts an investigation into several fundamental and experimental issues surrounding impulsive robotic manipulation and develops methods for their use in robot control. Project results are advancing the theory of multi-body impacts and widening the capabilities of robots. Direct applications of the results are leading to faster and more agile robots for civil and military services and improved efficiency of industrial processes.The technical goal of this project is to gain in-depth understanding about control of collision outcomes and, more importantly, integration of impact planning with motion planning in robot manipulation. The project consists of three phases. The first phase advances the Principal Investigator's recent work on compliant and multiple impacts to modeling of n-body simultaneous collisions, tackling issues that include state space condensation, impulse growth in high-dimensional space, sensitivity to physical parameters, and area and nonlinear contacts. In parallel with the analysis is the development of a graphical interface called MultiCollide for interactive collision simulation. The second phase investigates how a manipulator can impart via impact a desired motion to an object, solving an inverse impact problem that connects impact dynamics with trajectory planning and manipulator dynamics. The final phase focuses on two impulsive manipulation tasks that respectively address the following important issues: how to leverage contact compliance and geometry, and how to interleave impact planning with robot motion planning.