The initial CAD for this project was borrowed from the research paper "An open-source anthropomorphic robot hand system: HRI hand" by Hyeonjun Park and Donghan Kim as a jumping-off point to test how the mechanism behaved before the kinematic analysis was complete. Notably, the authors of this paper chose to actuate each digit individually, while we aim to actuate them all at the same time.
| Link | Length |
|---|---|
| L1 | 6.286 mm |
| L2 | 9.5 mm |
| L3 | 27.05 mm |
| L4 | 27.3 mm |
| Loffset | 3.693 mm |
| L5 | 19.56 mm |
| L6 | 6 mm |
Table 1: Link Lengths
Figure 1: Drawing of Linkage
The CAD as given in the form of step files in the research paper was taken and altered to fit with the M2 and M3 screws we had available. The lengths of the links were altered slightly as described above in Table 1 to achieve a slightly more acute gripping angle at the end-effector. The availability of the CAD in the paper was a very useful tool as it allowed us to experiment before the analysis code was completed.
Figure 2: Initial CAD Borrowed from Research Paper and Altered
Initially, we hoped to 3D print only the fingertip and palm base of the hand (Links 1 and 6), and use laser-cut wooden pieces as the rest of the links. However, upon assembly, we realized we failed to account for the disparity between the listed thickness of the wood and the actual thickness. We found that there was no sold thickness of wood that matched, so we opted to switch to 3D printing so as to not have to manually change the thickness of each piece in CAD. Despite this failure, it was still a worthwhile incident because we were able to roughly assemble the pieces and see that the mechanism did indeed work.
Figure 3: Prototype No. 1 - Failed due to lack of link thickness tolerancing
This design looked impressive, but it had several problems. First, it had a limited range of motion. As the mechanism is mirrored on both sides for stability, there needs to be a long bolt that crosses through joint O4. This ended up interfering with the motion of the finger by blocking the four-bar from extending further once it hit the bolt. This resulted in a limited range of motion and a high flexion distance from the finger to the palm, which would mean we could only really hold large objects. Another problem we faced was the difficulty of assembly, which took at least an hour and a half because of all the small bolts and the lack of tolerancing on the 3D-printed holes. Finally, the hole sizes were not uniform which led to there being a bit of play.
Figure 4: Prototype No. 2 - Failed due to lack of ROM of end-effector
We decided that this design did not meet our needs, and went back to the drawing board. We decided to simplify the design to make manufacturing and assembly easier and also to remove the bolt at joint O4 to increase our range of motion. This process was quick since we simply followed the same 3D sketch as the original and made extrudes in the assembly to get a skeleton of fundamentally the same linkage.
This prototype had several benefits over the previous one. First, it was better toleranced and assembled more easily since we standardized all the bolt sizes to M3. Secondly, it had a greater range of motion due to the lack of a bolt going across the middle connecting Link 2 to Link 4, which was previously hitting the mechanism and stopping it. Finally, it was overall simpler to manufacture and assemble. Going forward, this is the basic design we will be using for our final product.
Figure 5: Initial Design 2, simplified linkages
Figure 6: Prototype No. 3 - Updated prototype with simpler linkage design, higher ROM
The next step for our project will be to design a more comprehensive ground link that will incorporate five of these fingers into a hand. A special case is the thumb; we will simply reduce it's DOF from that of an actual thumb to 1, and treat it as a shorter finger. We plan on using springs in parallel to connect each finger to the linear actuator located on the back of the palm. The fingers will be extended when the actuator is unextended, keeping the springs in tension. When the actuator extends, it will allow the spring to compress, thus releasing the force on the actuation joint on L2 allowing the fingers to close. We will select springs with sufficient stiffness such that there will still be force on the fingers when they are fully flexed, which will provide the force needed to grasp whatever object is in the palm. Due to the natural compliance of the springs, the fingers will be able to grasp objects of differing widths.
Final Design
For our final design, we elected to have the linear actuator drive all 5 fingers with 1 actuator link that breaks out into 5 joints. This design allows us to apply force at every finger. The entire palm assembly is designed around this idea of a rail system, and since our linear actuator only provides linear translation force input, we designed the top 4 fingers to take in linear translation input and output rotational motion. This is equivalent to an under-actuated slider-crank mechanism where the input is the slider and the output is the crank.
Figure 7: Slider block (piston-esque) attacked to a short linkage. The entire block needs only 1/2in of linear travel to fully actuate the finger assembly
The thumb actuation is where we ran into some challenges. Since the thumb does not face forward like the rest of the fingers, we needed a way to turn the linear translation in the Y direction into a force input that is greater than 10 degrees from the axis of translation. To do that, we came up with a double rack and pinion system where the actuator link will have a gear rack coming out of the opposite side relative to the other 4 fingers' actuation joints. This gear rack will translate with the actuator link and apply motion to the thumb gearbox that includes an input pinion gear, an intermediate direction reversal gear, and another gear rack that will drive the under-actuated slider-crank mechanism attached to the thumb.
Figure 8: Thumb with gearbox attached
After figuring out all of the moving parts, we designed an enclosure that acts as a mount and a rail system for all the finger mechanisms and the actuator link. This enclosure needs to also hold part of the linear actuator and the entirety of the microcontroller that we are using. It needs to also be in two pieces with an acrylic plate that lets the inner workings be visible.
Figure 9: Palm enclosure, Actuator Link, Acrylic plate, and Top frame
Lastly, we designed an enclosure for the rest of the electronics which included a mount for the linear actuator that also serves as a mount for the L298N H-bridge and a clamshell-style forearm that will hold the battery and act as a handle to hold onto the robotic hand. This shell-style forearm needs to be realistic in size but also have enough volume to hold all of the cables that will be tucked into it.
Figure 10: Linear actuator/H-bridge mount, battery holder, and Forearm
Finally, after putting everything together we arrive at our final Minimally Actuated Robotic Hand.
Figure 11: Final assembly of Minimally Actuated Robotic Hand
| BOM | Qt. | Actual Cost Paid (Adjusted by amount used) | Cost | Link |
|---|---|---|---|---|
| PLA Filament | ~ 1 lb | Free from TIW | $6.30 | |
| Seeeduino Xiao SAMD21 | 1 | $9.99 | $9.99 | Link |
| Compression Springs | 2 | $0.13 | $0.13 | Link |
| Linear Actuator | 1 | $19.99 | $19.99 | Link |
| 11.1V LiPo Battery | 1 | $13.05 | $13.05 | Link |
| M2 Screw Set | 1 | $7.99 | $7.99 | Link |
| Electrical Connectors | 1 | Free from parts box | $1.00 |
Table 2: Projected Bill of Materials for Final Project