OceanOneK Hand

The motivation comes from the harsh environment underwater and the robot's need for sensing. Currently, OceanOneK is equipped with a tendon-driven underactuated hand to perform grasps and is teleoperated by human operators on the boat. When performing grasps the human operator can only gain information from the RGB cameras mounted on the head of the robot. Two challenges during this process:

1. The human operators cannot precisely localize the object based on vision (have a sense of whether the hand is in contact with the object)
2. The human operators often need to deploy different forces to different objects but missing force feedback.

My plan consists of:

1. Design and fabricate the tactile sensor (analyze and iterate using finite element methods)
2. Develop machine learning methods to interpret the raw sensor signal to forces
3. Provide force feedback to human operators

7.1 to

Meeting Notes and Tasks:

• Need to check the optical fiber in the arm
• Keep Vytaflex 20 and use the UV-resistant pigment (Clearflex is hysteresis and toxic)
• Find possible way to 3d print the endoskeleton
• Better skin cast
• Test out fiber location in finger
• Start experiment

Purchase list:

• Steel rod x2
• Pmc 780
• Vytaflex 20
• Smooth on universal mold release

Questions:

Material and making: Possible 3D printing material for the endoskeleton

Flexible 80A vs PMC780

• TPU in Bambu: Ninjaflex 83A or Filaflex 82A.
  • Rachel: Tested. Hard to eliminate the air inside and water is not absorbed inside
• Flexible 80A in Formlab
  • There might be hysteresis, need to make sure it will go back to the original position after leaving closed. Adding nitinol back might help.

Design: How to better connect the endoskeleton with the shell if 3D printed?

1. Use insert instead of nuts
2. Drill all the way through the front and connect the nut to the shell

Current Progress:

• Improved Casting:
  • Skin Cast: much better than the first try.
  • Several places are too thin to demold well. Need more mold release for the side this part of the mold is not as glossy as others.
    
    Skin Cast with original skin mold
• Improved Skin Design and Mold: Thicken the skin for 2 mm to better protect FBG under
  • Proximal 20 -> 22 mm R10->R12 Distal 12->14mm
  • New skin mold: one finger
    
    New Skin Cast
    
    New Skin Mold
  • New skin cast
• 3D printed endoskeleton
  
  3D printed endoskeleton
  • No visible bubble.
  • Around 3 hours of printing time for 1 finger. Too big to fit 2 fingers together in Form 3.
  • Reserved holes for bolt shrink or clogged after post-cure. Solved: plug the bolt in and clean the holes before post-cure.
  • Hysteresis. Revert to the original place slower than PMC 780. Experiment: leave closed for 2 days. There is a little deformation of the endoskeleton, but if connected to the shell with magnets, it functions the same.
• FEA analysis

6.19 to 7.1

Questions

Material: Better skin material under the sun and in the Ocean. -> Requested suggestions from Smooth-on and Full Line Distributor DOUGLAS AND STURGESS

• Smooth-on: Clear Flex 30 - UV resistant. So it should hold up under the sun.
• Douglas: Clear Flex series no softer than a 30 (essentially 3 products 30 50 and 95 Shore A). You could go to a silicone but there we don’t carry anything harder than about 50. We do have samples of most of these at our stores. Clear Flex series

Design: Where should we put the fiber in?

1. In the Endoskeleton: the Fbg will be infected when tendon pulling
   
   In the Endoskeleton
2. On top of the Endoskeleton: reduce the effect of tendon pulling
   1. Cast on top while casting skin
   2. Make it a ring with grooves for FBGs
   
3. Cast in the skin: PMC780 might not be hard enough as a base. Might make it easier to replace?

How Fbgs should be located on the fiber?

1. Only Fbgs on fingertips - the rest of fiber moves along the tunnel like a bicycle cable
2. Fbgs on all parts of fingers - the same tunnel as a) only distal Fbgs are fixed proximal Fbgs slide in the groove using learning to predict sliding or pressing
3. Wrap fiber around the finger and leave extra space on the back for bending fiber thread outside of the shell
4. Attach fiber along the side and leave extra space at joints

Palm sense - advice from Rachel

Add a sensor in the palm to avoid pushing things forward constantly

• Contact with palm -> binary signal contact/no contact -> start motor to pull tendon
• Contact with proximal parts -> binary signal/rough force?
• Contact with distal part 3 FBGs -> Force/Haptic feedback for human operators

Current Progress

Samples:

Left to Right: PMC 780, Vytaflex 30, Vytaflex 20

• PMC 780
• Vytaflex 30
• Vytaflex 20

Latest finger design mold & cast:

• Endoskeleton Mold

• Skin Mold

• Cast - Remake the latest version with endoskeleton and skin separated
  • Need to improve in reducing bubbles - potential pressure cast all parts
  • The skin layer is too thin and fragile to separate with Vytaflex 20.

Useful files

• O1K Finger Fab Instructions
• Onshape latest design
• FBG Principle
• Tower Onshape
• A Multi-Axis FBG-Based Tactile Sensor for Gripping in Space
• Smooth-on Comparison tool
• Functional mimicry of Ruffini receptors
• Presentation