Construction work demands a lot from a hand: holding power tools, carrying heavy materials, and gripping hard enough to resist large loads, while still being able to pick up and manipulate small items like nails and fasteners. A fully actuated robotic hand, with one motor per joint, gives complete control over every finger, but each extra actuator adds weight, cost, and complexity, and most of that dexterity isn't needed for the high-force portion of the work. My thesis started with designing a hand that meets the high grip force and payload demands of construction tasks, with a particular focus on securely holding power tools while operating them, without paying for actuators at unnecessary joints.

I conducted a literature review benchmarking against existing underactuated and prosthetic hands, tendon-driven grasping mechanisms, and other robotic legs, arms, and hands to identify the gap this design targets, analyzing trade-offs in actuator selection and kinematic design. Before coming to this conclusion, I explored an underactuated ankle-toe mechanism for bipedal robots, as well as underactuated proprioceptive touch sensing for humanoid hands.

My thesis question now is whether adding membranes between the digits, kind of like webbing, makes grasps more stable during dexterous manipulation. The idea grew out of looking at how people hold drills and the pressure point in the webbing between the thumb and index finger. I'm designing the membrane myself.

I'm trying to map out the effects of a membrane between two rigid fingers, both on the object being held and on the fingers themselves. That includes how it changes grasp quality and resistance to specific wrenches, as well as the grasp's compliance under those wrenches. I'm testing different natural tension schemes in the web so I can classify those effects and design the web for whatever positive effects are possible with it.

Alongside the mechanical design, I'm working on reinforcement learning policies to complete grasp and manipulation tasks using the underactuated hand, to evaluate how well the design handles both high-force grasps and finer manipulation. I'm responsible for the mechanical design, embedded systems, controls, and experimentation for this project.