Principal Investigator: Yu She
The next-generation robots will share the workspace with humans, working with them side by side to perform challenging tasks. Variable stiffness designs, ranging from low stiffness to high stiffness, will play a vital role for these robots because they can ensure safety and high performance simultaneously. The low stiffness modes guarantee safe interactions, and the high stiffness modes offer high performance (e.g., carrying heavy loads and position control). While most existing manipulators are designed with constant stiffness, we will develop a dexterous manipulator with variable stiffness that is safe for physical human-robot collaborations while providing high performance. From the perspective of design, we will develop the dexterous manipulator with 7 DOFs, including a forearm, an upper arm, a 3-DOFs shoulder, a 3-DOFs wrist, and a 1-DOF elbow, just like a human arm. The forearm and upper arm will be designed with variable stiffness, which can be achieved by layer jamming, shape morphing, effective length modulating, or Young's modulus regulating. In addition to the dexterity design, it is important to equip the dexterous manipulator with the capability of perception to understand the states of the manipulator itself and the contact environment. From the perspective of perception, we will implement the embedded vision sensors into the manipulator to achieve high-resolution and high-dimension proprioception and tactile sensing. Together with a static model, the proprioception data will be used to estimate the stiffness of the manipulator in real-time. We will apply the finite element analysis (FEA) to estimate the shear force applied to the manipulator. We will also explore the integration of the inverse FEA with depth images to reconstruct the normal force applied to the manipulator. In terms of control, we will investigate the optimal stiffness control of the dexterous manipulator according to the working environments. The manipulator will be attached with an off-the-shelf robotic hand (e.g., Robotiq Hand E) for task testing in household applications.
Keywords: Physical Human-robot Interactions, Robotic Arms, Variable Stiffness