A Two-Stage Geometric Approach to Planning Robotic Tasks Involving Sliding and Rolling Contacts in Uncertain Environments

点击翻译按钮获取中文摘要

9304734 Trinkle This is the first year of a three year continuing award. The research addresses automatic planning for robotic systems performing contact tasks with movable objects in uncertain environments. Tasks in this class include mechanical assembly operations, dexterous manipulation, and fine motion tasks. The objective of this research project is to develop a practical, reliable planner for robotic contact tasks. The planner will operate in two stages. In the first stage, parameter uncertainty will be ignored to reduce the dimension of the space to be searched for a manipulation plan. The geometry of that space will be determined and regions in which the multibody contact model admits multiple solutions will be identified and avoided. In the second stage, the plan found in the first stage will be modified to make it robust with respect to variations in the uncertain parameters. The second stage will not involve a search, so its computational complexity will be relatively low. It will however involve understanding the geometric variation in the space as a function of the uncertain parameters. To demonstrate the results of the proposed research, a planner for planar manipulation systems will be implemented in a distributed computing environment, with parallel computers used where appropriate. A prototype manipulator system with position, torque, and tactile sensors will be designed and built to test generated plans. ***