Pre-Grasp Interaction as a Manipulation Strategy for Movable Objects

Publication Year:
Usage 583
Downloads 529
Abstract Views 54
Repository URL:
Chang, Lillian Y
thesis / dissertation description
Robotic systems have yet to match humans in skill for movement planningand tool manipulation. For example, humans can robustly graspand manipulate objects even under task variation. However, successfulgrasping methods for robotic manipulators are often limited to structuredenvironmental conditions. Our dual goals are to understand manipulationactions in humans and to add such skills to a robot manipulator’srepertoire. In particular, we examine strategies for object acquisition,which is a common first component in manipulation actions.Many approaches to automating robot motion for object acquisitionhave focused on reach-to-grasp tasks, where the arm motion and handconfiguration are planned for grasping an object. With these solutions,the object placement often remains fixed in the environment until theobject is carefully grasped from its presented configuration. In contrast,humans often take advantage of an object’s movability to reorient andregrasp an object during the acquisition process.This thesis investigates how such pre-grasp interaction can improvegrasping through preparatory manipulation of the object’s configuration.Specifically we studied the strategy of pre-grasp object rotation for graspacquisition prior to a transport task. First, we examined human performanceof the pre-grasp rotation strategy. A larger amount of pre-graspobject rotation correlated to a greater lifting capability, or maximumpayload, of the grasping posture used at the time of object acquisition.In addition, when the task was more difficult due to increased objectmass or increased upright orientation constraints, there was decreasedvariability in the object orientation selected for grasping. Second, wedeveloped and evaluated a method for planning pre-grasp rotation for arobot manipulator. Our results show that the pre-grasp rotation strategycan improve a robot’s manipulation capabilities by both extending theeffective workspace for a transport task and improving the quality of thetransport action.