Collaborators: Philippe Wu | Advisors: Joaquin Palacios, Gagan Khandate, Matei Ciocarlie
With the goal of developing a low-cost testing platform for the lab, this project focued on building on the work of the ALOHA team to create a teleoperated robotic gripper. Our aim was to enable robots to complete dexterous bi-manual manipulation tasks using accessible hardware and imitation learning, negating the need for expensive sensors, vision systems, or closed-loop feedback.
The system consists of two parts: a human-operated controller and a robotic end-effector that attaches to the lab’s UR5 arm. Motion from the controller is mapped directly to the end-effector, allowing human participants to model tasks for the robot to learn.
End-Effector: The end-effector is built around a Dynamixel servo and uses a custom-designed linkage to convert rotational motion into translational grasping. The linkage is torque-optimized and ensures that the fingers can start and stop from any position.
Fingers: Designed to be modular, the “fingers” (white in the GIF to the right) are hot-swappable for different applications. During the design process, we fabricated both PLA and TPU versions, providing stiff and pliable options.
Motion Tracking: In our design, motion is tracked using a Trackstar positioning system (for absolute location) and a 1-turn rotational potentiometer (for opening/closing the gripper). The potentiometer is embedded into the base of the controller along with a torsional spring that provides tactile resistance to the user.
Control: As a preliminary method of control, we used serial communication between an Arduino board (controller) and OpenRB-150 board (end-effector). The voltage values from the potentiometer were mapped linearly to the effective range of the angular position on the end effector, allowing for the motion shown here. In the time since, the project has switched to control using ROS1.
In the time since this project, the system has been modified and used in the lab for some cool research, including two publications (one in arXiv preprint at https://doi.org/10.48550/arXiv.2503.12297 and another to appear in IROS 2025!).