My overarching research vision is to create computational material robots -- robots with advanced functionality through the algorithmic design of their material properties.
Robotics is traditionally the combination of three fields: mechanical engineering, computer science and electrical engineering. Recently, there has been more of a push to add a fourth field to the mix, material science, creating new material robots that use the embedded intelligence within their material properties to create new techniques. My goal is to create customizable computational material robotics so that given a task, robots can be autonomously designed, optimized and fabricated for any application.
In this vein, I have been primarily focused on soft robotic manipulators, creating new actuators and sensor integration with the potential for computational optimization and mass-customizable production.
This work introduces a new kind of metamaterial called handed shearing auxetics (HSAs). Auxetic materials are those with a negative Poisson's ratio, meaning that they expand perpendicularly when pulled in tension (unlike most materials which contract when pulled). Shearing auxetics expand with a bias rather than just isotropically expanding. By adding chirality to a shearing auxetic pattern, we can create new structures that have a strong coupling between twisting and extension, as well as structures with selective stiffness based on the pairing of chirality. We have demonstrated the potential of HSAs by using it to create a compliant linear actuator and as a soft robotic manipulator without the need for pneumatic actuation.
Abstract: In this paper, we explore a new class of electric motor-driven compliant actuators based on handed shearing auxetic cylinders. This technique combines the benefits of compliant bodies from soft robotic actuators with the simplicity of direct coupling to electric motors. We demonstrate the effectiveness of this technique by creating linear actuators, a four degree-of-freedom robotic platform, and a soft robotic gripper. We compare the soft robotic gripper against a state of the art pneumatic soft gripper, finding similar grasping performance in a significantly smaller and more energy-efficient package.
Published in 2019 IEEE-RAS International Conference on Robotics and Automation (ICRA), 2019.
Abstract: Compliant robotic grippers are more robust to uncertainties in grasping and manipulation tasks, especially when paired with tactile and proprioceptive feedback. Although considerable progress has been made towards achieving proprioceptive soft robotic grippers, current efforts require complex driving hardware or fabrication techniques. In this paper, we present a simple scalable soft robotic gripper integrated with high-deformation strain and pressure sensors. The gripper is composed of structurally-compliant handed shearing auxetic structures actuated by electric motors. Coupling deformable sensors with the compliant grippers enables gripper proprioception and object classification. With this sensorized system, we are able to identify objects' size to within 33% of actual radius and sort objects as hard / soft with 78% accuracy.
Published in 2018 IEEE-RAS International Conference on Soft Robotics (RoboSoft), 2019.
Abstract: Single-stream recycling is currently an extremely labor intensive process due to the need for manual object sorting. Soft robotics offers a natural solution as compliant robots require less computation to plan paths and grasp objects in a cluttered environment. However, most soft robots are not robust enough to handle the many sharp objects present in a recycling facility. In this work, we present a soft sensorized robotic gripper which is fully electrically driven and can detect the difference between paper, metal and plastic. By combining handed shearing auxetics with high deformation capacitive pressure and strain sensors, we present a new puncture resistant soft robotic gripper. Our materials classifier has 85% accuracy with a stationary gripper and 63% accuracy in a simulated recycling pipeline. This classifier works over a variety of objects, including those that would fool a purely vision-based system.
I am currently working on building on this prior work about programmable viscoelastic materials for further applications.