Haptically Guided Teleoperation for Learning Manipulation Tasks
Howard, Ayanna M.
Park, Chung Hyuk
MetadataShow full item record
In this paper, we present a methodology that uses control signals provided through guided teleoperation to assist in the learning of new manipulation tasks. The approach incorporates haptic feedback that guides human behavior in performing a manipulation task using guidance forces derived from visual input data. The control signals provided by the user are then utilized by the robotic system to learn the control sequences necessary for task execution. A neural network learning method that incorporates historical information is utilized for the learning process. The primary focus of our approach is to develop a method to enable the robotic system to improve its ability to learn manipulation tasks, whether or not the instruction is provided by an expert or general user. The methodology is explained in detail, and results of a manipulation system learning an object-centering task is presented.