Week 1
My first week in the lab was mostly focused on getting oriented and learning more about the project. I got an intro to the lab, toured the SRC, and started learning more about the plan to adapt Sharpa hands for underwater manipulation. I also spent time thinking through how to design and prototype a waterproof glove system for the hands, what early testing might look like, and created a shared google drive and drafted a rough 10-week plan for the project.
The rest of the week was a mix of getting settled into lab life and meeting the team. On Wednesday, I helped with lab cleanup, and on Thursday I finished my lab safety training. Friday was the lab retreat to Angel Island, which was a fun way to get to know everyone better outside of the lab as a start to the summer.
Week 2
This week I started working more directly on the waterproof glove system for the Sharpa hands. The goal of the gloves is to keep the hands dry while allowing them to manipulate objects underwater, so I spent time researching different fabrication methods, including dip molding, spray molding, and brush molding. I started putting together a glove fabrication research document, looking at materials like silicone, urethane, and latex, and thinking through how each process could work with 3D-printed mandrels.
I also scanned the Sharpa hands in the SRC using Polycam, an app that uses the built-in LiDAR on an iPhone to create 3D scans. The results were surprisingly good and definitely usable for my current goal of making a mandrel for glove molding. I brought the scan into Fusion, where I remeshed it and converted it from a mesh into a surface body. One interesting thing that happens during this process is that the original scan, which is made of triangular mesh faces, gets rebuilt into a more editable surface made from quadrilateral patches. Triangles are useful for capturing irregular scanned geometry, but quads are generally easier to edit, smooth, and use in CAD workflows. Here an interesting article I found about the difference between quads and triangles in 3D modeling.
Finally, I designed and fabricated a small rotisserie fixture that will eventually help with glove fabrication. The idea is to slowly rotate a mandrel while silicone or urethane is applied, so the coating can build up more evenly instead of sagging to one side. This could be useful for dip, spray, or brush molding, but I am mainly thinking about it for brush molding, which is the process we are currently leaning toward. The mandrel can be attached with a set screw, and the motor is controlled with a simple on/off switch and plugs directly into the wall through a 12V adapter.
Week 3
This week I started prototyping the glove fabrication process more directly. The goal of the gloves is to keep the Sharpa hands dry while still allowing them to manipulate objects underwater, and we are exploring dip, spray, and brush molding techniques to fabricate them. I used the single-finger mandrels I designed and brushed silicone, specifically Dragon Skin 20, onto the mandrels while they rotated on the rotisserie fixture I made last week. The rotisserie speed can be adjusted by changing the supplied voltage; right now I have it running at about 5V, which gives roughly 8 rpm.
I fabricated several prototypes this week. On 7/7, I tested a single layer of Dragon Skin 20 on a fairly angular, unsmoothed finger mandrel. The silicone was thin at sharp edges, thicker on flat areas, and pooled around the smaller-radius joint regions. It was removable from the mandrel even without release agent, but the thin areas tore during removal and stretched into larger tears afterward. On 7/8, I tried two layers and added dye to make it easier to see the silicone while brushing. This version did not tear, but it ended up too thick overall and was still thin near the fingertip, where the rotation seemed less effective. On 7/9, I tested one layer again on a smoother mandrel with 5% silicone thinner. The result was very similar to the 7/7 prototype, with uneven texture near the tip, tearing at the tip, and pooling at the joints. On 7/10, I made another prototype with three layers and 10% silicone thinner, which is still curing over the weekend.
I also spent time working with Soy on adding another degree of freedom to a gripper design from YUBI. The goal is to create a more dexterous gripper that can better pick up small objects, like screws. This connects back to the glove work because waterproofing could make it possible to use robot hands with more degrees of freedom underwater. More dexterity usually means putting motors, sensors, and electronics closer to the fingers and hand instead of keeping everything sealed farther back in the arm, so a reliable glove or waterproofing method could open up more hand design options.
The gripper I designed tries to preserve the function of the original single-DoF gripper while adding another degree of freedom at the fingertip. This could allow the gripper to roll an object along the finger, grasp from different angles, pick up smaller objects that are flat on a surface, and generally manipulate objects more precisely. I integrated a Dynamixel XC330-M288-T motor into the design because it is small, and I tried to minimize the space around the motor so the finger does not become too bulky. There is also space for a Paxini tactile sensor at the fingertip. I have started printing test pieces, and the next step is to get the new fingertip mounted on a gripper so we can compare it to the original design.
Week 4
This week I continued working on both glove fabrication and hand design. For the glove fabrication work, I dip molded the tip of the 7/10 finger prototype to add more silicone to the fingertip, which has consistently been the thinnest area in my brush molded samples. This prototype seems like the most effective version so far, but the wall thickness is still uneven. Because of that, I am starting to explore other fabrication methods, including molding silicone over existing gloves and casting gloves using a mold and core. I also still need to experiment more with different mandrel geometries for brush molding on the rotisserie, including corrugations and other shapes that could help the silicone build up more evenly around the joints.
The NEO hands from 1X are an interesting reference point for this work because they combine many degrees of freedom with tactile skin and waterproofing, showing how a sealed outer glove/skin can support dexterous manipulation while protecting the hardware underneath.
I also spent a lot of time this week working on hand design prototyping. I created a physical model of the gripper I designed, which has two motors in the fingers to actuate the fingertips and two motors at the base of the fingers to rotate the fingers. I started coding a few basic hand movements so I could test how the gripper moves outside of CAD. The goal of this design is to eventually perform fine manipulation tasks, hopefully underwater, where waterproofing could allow the motors and sensors to be placed closer to the fingers.
The next steps for the hand are to design the gripper to be human operated so we can gather training data (much like UMI), hard code example grasps for several objects, and perhaps code the gripper to be controlled from a computer with more intuitive controls. These tests should help show whether the added fingertip degree of freedom actually improves manipulation and whether this type of gripper could be a useful platform for underwater dexterous manipulation.