Semi-Autonomous Manipulator Control Using 2D and 3D Scene Recognition

Autonomous Solutions, Inc. (ASI) has teamed with Johns Hopkins Universitys Laboratory for Computational Sensing and Robotics (LCSR) to continue its Phase I development of autonomous manipulation and grasping behaviors for robotic manipulators. The team will continue the Phase I effort to engineer a system which will: Rely primarily on low-cost and ubiquitous monocular vision to allow for positioning of a robotic end effector to a point of interest identified by a user, Use computationally light grasp planning algorithms to grasp objects of interest, Use structure-from-motion algorithms to enable path planning and obstacle avoidance as well as to generate 3D points on the object of interest for more robust grasping and manipulation, Use monocular vision-based techniques for scene parsing and understanding which will enable a user to: o Easily select an object of interest using verbal commands, o Preview and easily adjust the robots intended grasp plan, o Invoke manipulation behaviors (such as pick-and-place and use of tools) based on a database of recognized objects.