Replication of Human Two-Fingered Grasping Motions in Robotic Manipulation of Household Objects

In our project, we consider the problem of replicating human two-fingered grasping motions on the AX12 Smart Robotic Arm. Two-fingered grasping motions are among the most common motions used in everyday life. Examples include picking up a plate from a dishwasher, moving a chess piece, and as we will demonstrate, opening a drawer. In fact, many more complicated motions, such as holding a pen/pencil and writing involve a two-fingered grasping component. To accomplish this goal, we designed a sensor mount positioned behind the robotic arm that accommodates a variety of sensors including a 720p highdefinition webcam. Then, using SIFT features, we locate (in the webcam image) the feature associated with the drawer that we want to open and close defined by a point in the image. Using previous calibration and bilinear interpolation, we convert this coordinate into global Cartesian space. This coordinate is subsequently fed into the analytical inverse kinematics service which replies with the joint coordinates. Subsequently, these are given to the controller which moves the arm to a desired location. Then, using a series of movements, we can perform our task. In particular, this project takes advantage of ROS’s service/client and publisher/subscriber model. These paradigms form the basis for communication between different components of the project.