​I make {soft, gentle, safe} robotic hands.

In-Hand Manipulation using Soft Hands!

Harvard Microrobotics Lab (w/ Prof. Robert Wood)
The last few years, I have been working to understand how to design soft robotics hands with enough dexterity to perform in-hand manipulation.  I am looking at all aspects of hand design including finger design (and relevant workspace), finger arrangement, and palm design. Working with an excellent Masters student, we kicked this work off in 2019, and since then I've worked with several stellar undergrads to triage and explore several aspects of hand design.

To facilitate design studies, we've also been developing performance metrics that are relevant for in-hand manipulation, but take softness into account.

I've also been working with a fellow PhD student on a fast, accurate simulation platform for soft robots (SoMo) that we can use for design explorations as well as machine learning for control.
Relevant Publications ⮞

Grasping and Manipulation Using Soft Robotic Hands

Harvard Microrobotics Lab (w/ Prof. Robert Wood)
My recent research directions attempt to answer the following question:

How can we design soft hands with high dexterity using minimal added complexity?

This has pushed me to begin studying grasping and in-hand manipulation at a more-basic level. My approach involves a semi-rigorous design process:
  1. Define task-level grasping and manipulation goals.
  2. Use these goals to derive how a hand should be structured (i.e. how should the soft fingers should move? how should fingers be arranged? How should a palm be utilized, etc.)
  3. Design and fabricate soft robotic hardware capable of achieving these motions.
  4. Validate that the hardware can perform the task-level  goals reliably.

Using this process, we can explore how the fundamental design of soft fingers and other hand components affect the resulting grasping or manipulation performance. 

My recent work  uses this design process to explore the problem of precision grasping. Drawing from traditional grasping literature, we developed an understanding of the desired shape and compliance of fingers at contact points that enable more-stable precision grasps. This suggests that for the best precision grasping, we need soft fingers designed with two separate bending segments. Read more about it in the publication in IJRR.
Relevant Publications ⮞

Soft Sensors for Underwater Soft Robotic Grippers

Harvard Microrobotics Lab (w/ Prof. Robert Wood)
Existing sampling equipment used on deep sea ROV's involves rigid, overpowered arms which are very difficult to operate, and put unnecessary amounts of stress on organisms during the collection process . "Squishy" soft robotic fingers can provide a gentle grasp to avoid applying too much force to organisms. The concept has been demonstrated by our lab several times on actual dives.

In this project, we outfitted our existing soft fingers with proprioception and contact senses. Building upon existing technology, I developed optical waveguide-based soft sensors that operate reliably in low-temperature, high-pressure salt water environments. We plan to implement these on our gripper, and move forward to higher level tasks such as optimizing sensor placement, sensor fusion, and various grasping tasks with machine-learning, and designing effective user-interface devices. 
​
Relevant Publications ⮞

Simple Legged Microrobots: ​Locomotion Based on Contact Dynamics and Vibration

Vibration and Acoustics Laboratory: Microsystems
​(w/ Prof. Kenn Oldham)
We were interested in a family of simple, small-scale robots with multiple sets of high-frequency actuated compliant elastic legs and a rigid body. We developed a dynamic model (derived from beam theory) to describe the motion of these simple small-scale robots in the presence of variable properties of the underlying terrain.  The motion of the small-scale robots results from dual-direction motion of cantilever-beam legs, with impact dynamics increasing the robot's locomotion complexity. This model is modal in nature, meaning it is invariant to the specific robot design.

We verified the model using two different centimeter-scale robot prototypes having an analogous actuation scheme to millimeter-scale microrobots. In accounting for the interaction between the robot and ground, a dynamic model using the first two modes of each leg shows good agreement with experimental results for the centimeter-scale prototypes, in terms of both magnitude and the trends in robot locomotion with respect to actuation conditions.

Relevant Publications ⮞