Third Hand Upgrade
Author: Rodolfo Antonio Salido Benítez
Completion Date: Sep 16, 2018
Software: Autodesk Fusion 360, PhotoScanPro, MeshLab, Simplify 3D
Place of Creation:Biomedical Research Facility II. (UCSD)
Techniques: Photogrammetry, geometric modeling, organic modeling, computer aided design & manufacturing, rapid prototyping, and additive manufacturing.
Materials: Polylactic Acid (PLA), Printed Circuit Boards (PCB's), Arduino micro-controller, servo motors, proximity sensor, cable tensioners, aluminum extrusion, and lithium battery.
The Third Hand Upgrade is a biomimetic prosthetic augmentation attachment that interfaces with my body to serve as an addition limb. The project consists of an open source 3D printed prosthetic hand, developed by exiii under the name HACKberry, mounted on an aluminum pylon dressed with biomimetic fairings that interfaces with my right forearm through a mount with micro adjustable tensioners and an adjustable elbow joint. Thumb opposition, pinch-release, and power grasp movements of the prosthetic hand are mediated through three servo motors controlled by an Arduino micro that translates muscle movements sensed by an infrared proximity sensor into open hand and close hand commands. All plastic components were printed entirely in PLA using a personally fabricated RepRap 3D printer.
The HACKberry open source design uses two motors to contract and extend fingers, an independent motor for the index finger and a second motor for the coupled movement of middle, ring, and pinky fingers. A third motor is used to enable thumb opposition, which can be toggled by pushing a button on the back of the hand. A second button locks the movement of the coupled fingers in order to enable exclusive control of the index finger for pinch movements. The back of the hand also has an on/off switch and a third to activate a sensor calibration procedure.
This project was conceptually influenced by the popularization of cooperative design surrounding prosthetics, exposure to artistic ideas of technological body augmentation, and access to computer aided design and manufacturing technologies.
Globally distributed maker communities like e-NABLE and mission arm Japan have been building platforms for interconnectivity between volunteers and individuals in need of mobility assistive devices since 2014 and 2015 respectively. These communities have developed numerous engineering solutions to upper limb mobility needs through a philosophy of cooperation as opposed to competition. Their open source designs, often licensed under Creative Commons, are hosted in online 3D model repositories where anyone with access to additive manufacturing technologies can replicate them and where individuals with engineering expertise can contribute to further develop them. Platforms like these encourage individuals to work for non monetary rewards, an idea that resonates in me and that has pushed me to be an active participant in design and manufacturing methodology development for prosthetic devices.
Since my first exposure to these cooperative projects, I've researched and implemented additive manufacturing techniques coupled with computer aided design to develop a custom interface between an external mechanical device and my body. I focused my attention in developing and properly documenting accessible 3D scanning techniques in order to inform the maker communities of cheap methods for high fidelity socket design. I had the opportunity to dedicate an academic quarter, from January to March 2016, to a self funded undergraduate research project surrounding clinically established and state of the art prosthetic socket design. During this time I conducted primary literature review on socket design and manufacturing evaluation methodologies and applied the later to validate the utility of a newly introduced commercial 3D scanning product, Scanify by Fuel3D, for prosthetic socket design. Unfortunately, the product proved to be unfit for the desired applications but the core technology behind the device, photogrammetric 3D scanning, was further explored and implemented in the final product of this project.
"A prosthesis not as a sign of lack, but rather a symptom of excess."
The project was envisioned with a body augmentation goal as opposed to a rehabilitative prosthetics project as a tribute to the influential work of Australian performance artist Stelarc. Stelarc has been an active voice in the discourse regarding the limitations of the biological human body and its potential for technological evolution since 1980 and this project is an allusion to his Third Hand performances between 1980 and 1998. These
performances, and most of Stelarc's body of work, expose the capabilities of our symbiotic relationships with computerized systems, the possibility of mechanically redesigned bodies, and the directed evolution of the human body mediated by technology instead of random mutation.
In contrast with Stelarc's Third Hand, which was built with assistance of electric industrial company Imasen in Nagoya Japan, this project was partially designed and fully manufactured by the end user, myself, through materials that are easy to source and technologies that could enable wide distribution. Thus, it is not an exaggeration to state that technologically directed evolution is at our doorstep. The body augmentations designed by scientists and artists can be practically reproduced, adopted, and enhanced in cooperation across the globe.
Briefly, a HACKberry hand was 3D printed and constructed following the project's documentation then attached to a biomimetic artificial forearm that interfaces with a biological limb through a personalized adjustable mount. The biomimetic artificial forearm was designed to mirror the biological limb it's attached to and the adjustable mount was designed around a 3D scan of the same limb.
Photogrammetric 3D scan
A 3D mesh of my arm was generated through a dense point cloud reconstruction from 29 individual photos taken approximately ~10º apart from each other by circumnavigating my arm. In order to decrease surface uniformity proper of human skin, a temporary pattern of X's, spaced one inch apart from each other, was transferred to my arm using tattoo stenciling ink (the pattern improves the performance of the dense point cloud reconstruction algorithm by providing anchor points that are easily recognizable between photos while the spacing between X's allows for accurate scaling of the model). The 29 photos were imported to PhotoScanPro and the Camera positions in 3D space were inferred by aligning the photos using the Align Photos workflow of the software with the following parameters: Accuracy (Highest), Generic preselection (checked), Key point limit (40,000), Tie point limit (0). After aligning photos, a dense point cloud was generated using the Build Dense Cloud workflow with the following parameters: Quality (Ultra high), Depth filtering (Aggressive), Calculate point colors (checked). Finally, the 3D mesh was constructed using the Build Mesh workflow with the following parameters: Surface type (Arbitrary (3D)), Source data (Dense Cloud), Face count (High (444,181)).
After exporting model, the complexity of the mesh was decimated using meshlab in order to achieve a smoother surface finish while still preserving the overall topology of the scan.
Organic Sculpting (CAD)
The 3D scan was imported into Autodesk Fusion 360 to be used as a mold for organic design of both of cosmetic fairings and the forearm mount. Under the sculpting environment of Fusion 360, T-Spline planes were created for each component to be sculpted, one for the mount and one for each fairing. Then, the Pull feature under the Modify submenu of the sculpting environment was used to make the vertices of the planes coincide with those of the mesh imported from the 3D scan. From there, faces and vertices of the T-Spline planes were modified until a satisfactory cosmetic product with high geometric fidelity to the biological 3D scanned limb was achieved. Finally, an adjustable elbow was sculpted using a Quadball T-Spline shape. The following videos can serve as a guide to the aforementioned processes: Modeling for the human body in fusion 360 and Mesh modeling.
Geometric Modeling (CAD)
The modeling environment was used to design mechanical components needed to attach the HACKberry hand and the cosmetic fairings to the aluminum pylon, a functional adjustable elbow joint based on a simple spring loaded locking mechanism, and mechanical mounts for the micro adjustable tensioner system sourced from Boa (BOA S2-S). The slotted grills of the fairings were designed to allow for tool access to the aluminum pylon, reduced component weight, and cosmetics.
Additive Manufacturing (CAM)
All 3D printed components were manufactured on a RepRap Kossel 3D printer using Polylactic Acid (PLA). All models were sliced using Simplify 3D at 200 micron layer resolution, printed with supports and a raft. HACKberry hand components were downloaded from this repository.
The final product is a prosthetic augmentation attachment with a micro adjustable limb interface, an adjustable elbow joint, adjustable wrist rotation, ulnar and radial deviations, an opposable thumb, and lockable non-index fingers. The device was intended to be controlled by the same arm it was mounted on but weight bearing of the device resulted in sensor noise that made it impossible to accurately control it, thus, the sensor was mounted on the opposite arm. Further development of the body augmentation device will explore the possibility of myolectric control through pectoral muscle contractions.
A button in the back of the hand activates a calibration procedure that lasts 8 seconds. During this time, the user contracts and relaxes the forearm muscles next to the sensor and the minimum and maximum readings of the sensor are stored as min and max variables. The range between min and max values is re-assigned to a range of 0-100 where values below 15 result in finger extension, values close to 25 result in fingers holding their position, and values higher than 40 result in finger contraction with variable speed.
The sensor works by measuring the changes in distance to the skin its mounted next to. When muscles contract, cushions surrounding the sensor compress and distance between the sensor and skin decreases. This sensing system has proven to be an effective linear actuator for prosthetic device control by Masahiro Yoshikawa et al. in their paper Trans-radial prosthesis with three opposed fingers
Rotational adjustments can be made by pressing spring loaded buttons around the device. Thumb opposition is activated by pressing a button on the back face of the hand which toggles the thumb motor between two positions, opposed to index finger or in the plane of the palm. Another button on the back of the hand locks non-index fingers in their current position. This can be used to isolate index finger movements for pinching gestures. These features allow the hand to be used to grasp irregular objects. The following video demonstrates all of the functions described and tests the efficiency of the hand at grasping a variety o different objects.
The Third Hand Upgrade has a variety of limitations. The full weight of the device mounted on my forearm caused an uncomfortable amount of shear stress on my skin. Consequently, a textile elbow compression sleeve was used as interface material between skin and plastic. This greatly reduced discomfort, but it was deemed necessary to modify mount design to extend to bicep for weight bearing.
The forces needed to support the weight of the device made it impossible to control it when placing the sensor on the same arm as the mount. Thus, it was necessary to use the opposite forearm to control the device, effectively limiting the function of two biological limbs in order to control the functions of a third artificial limb, a counterproductive solution.
The constant use of the analog motors of the prosthetic hand produces enough heat to trip the resettable fuses installed as a preventative measure to protect the motors. This means that after around fifteen minutes of constant use, the motors controlling the non-index fingers will stop behaving appropriately. This problem goes away easily after letting the motors cool down, however, this is a big limitation for individuals who intend to practically use the HACKberry hand as a prosthetic device.
I would like to explore the use of myoelectric sensors to control Third Hand Upgrade. This envisioned myoelectric controlled device would use pectoral muscle contractions for open and close hand commands. I would also like to encode thumb opposition and non index finger lock commands as muscle contractions to allow for more fluid control of the artificial limb. The mount would be redesigned to extend toward the bicep in order to distribute the weight of the device and prevent excessive torsion forces from acting on the forearm.
Ultimately, I would like to fit the HACKberry prosthetic hand to an individual with upper limb mobility needs. This project was essentially a pilot study to determine if I could effectively use photogrammetry and computer aided design to manufacture a prosthetic socket. I believe I acquired the necessary skills to achieve this goal and I will start looking for a candidate.
I want to express deep gratitude to the exiii team for placing their trust in the open source community to push their project forward, especially to Hiroshi Yamaura for his constant engagement in the HACKberry forum. I would also like to acknowledge Greg Humphrey for putting up with countless hours of 3D printing noise, and Satima Anankitpaiboon for her help with hardware sourcing and her technical support.