The beginning of an ongoing project in human-robot physical collaboration, we begin to explore and characterize the information communicated between participants in a joint physical task, like moving a table or other large object. To start, we characterize static forces at the gripper using only the joint torque and angle information provided from the JACO arm's ROS node, and basic kinematic information such as the mass and length of the arm links. The next step of this project will be to characterize dynamic forces in a human-human collaboration that make up haptic communication.
Once we have the human side of the haptic communcation and a method for identifying the transmitted forces in the robot, we can extend our model to the human-robot scenario. There, we can use the model we built for determining forces exerted at the gripper with arm states to identify the forces that come in from a human collaborator, and design a controller to respond appropriately to those forces, or to transmit haptic communication as the "motion leader", and to hand off leader-follower roles in a fluid and dynamic scenario.
Spring 2022 Slides from a class project presentation: initial results promising but in need of an improved algorithm. Current work is focused on combining the kinematic approach with the machine learning approach to handle nonlinearities.
Documentation (files or external links):