Loomio
Thu 27 Dec 2018 11:35AM

Mini-Grant for Development of a Transradial Prosthetic Arm

BP Bhargav Parthasarathy Public Seen by 120

> I had to cut out this introduction and the references section and post them separately because my thread was exceeding the Loomio thread character count, so I thought it made sense to post these two sections beforehand. I will post the rest of the thread right after this one; this introduction would have been the first thing, and references would have come last.

Hi everyone, my name is Bhargav Parthasarathy and I am an engineering undergraduate student. Over the past 18 months or so, I have spent much of my time researching the priorities and patterns involved in transradial prosthetic arms in comparison to more dexterous and biomimetic robotic arms. These interests have led me to my current work of building a competitive transradial prosthetic arm that improves upon current designs.

After detailing my project here and if given approval from the community, then submitting a proposal, ultimately I hope to gain as much funding as possible to support the full cost of the project. As a college student, I am unable to set aside enough money right now to fully fund this project due to tuition and housing, which is why I was interested in the micro-grant process.

Seeing as my goal here is to eventually submit a proposal anyway, I figured it would make more sense to organize my project details in the micro-grant proposal format which I am sure you all are more familiar with reviewing. I tried to make this as detailed as I possibly could and would appreciate any feedback. Thank you.

References:

Stanford study: https://web.stanford.edu/class/engr110/2011/LeBlanc-03a.pdf
Prosthetic arm cost breakdown: https://health.costhelper.com/prosthetic-arms.html

BP

Bhargav Parthasarathy Thu 27 Dec 2018 11:42AM

Description of Proposed Project:

The Problem:

Establishing the problem first, according to a Stanford study published by prosthetics specialist Maurice Leblanc the current amputee world population is now well over 3 million, about 80% of which are living in developing countries, and about 59% of which are below elbow (transradial) amputees. The latter two figures here are particularly important to note because with the far majority of amputees living in developing countries and majority being transradial amputations, this helps to explain why there is such a large focus (especially now) on designing low-cost prostheses that are in turn more accessible, with many if not most designs being transradial prosthetic arms. Looking at the average cost of a transradial prosthetic arm according to CostHelper Health's prosthetic arm cost breakdown, the shift toward low-cost transradial prostheses is further understood with the average cost of a lower arm prosthetic arm being $20,329 - a price most families simply cannot afford, excluding additional costs due to physical/occupational therapy.

Of course keeping the main goal of restoring as much motion and freedom to the amputee in mind, I have aimed to achieve this by specifically focusing design around balancing dexterity and production cost. After researching and understanding models like Open Bionics' Hero Arm, Exii's Hackberry arm, and Ottobock's Bebionic 3 among others, the effective balance of these two considerations is what I feel to be a gap in today's design of transradial prostheses.
If we look at the three aforementioned models more closely, this is helpful in describing the current state of prosthetic design and helps to justify the need for a more effective design.

1) With the Hackberry prosthetic arm being to the lower end of the cost spectrum, this open source design is impressive in keeping production costs to as low as $200, however, lacks in terms of dexterity. This is seen with the absence of wrist rotation, two-segmented fingers dependent on elastic force, relatively few degrees of freedom, off-hand dependent operations such as with thumb and preset control/selection, and an unreliable infrared distance sensor for control.
2) With the Open Bionics Hero Arm, this is an example of a transradial prosthetic arm in the middle of the cost spectrum at about $6,980. This design is certainly more expensive, but also more dexterous than the Hackberry Arm and even features an adaptive fit. However, it is still limited dexterity wise due to relatively few degrees of freedom with two-segmented fingers and a rotate-only thumb, reliance on the off-hand for preset selection and wrist rotation, and the fingers' reliance on elastic force like the Hackberry arm.
3) With the Bebionic 3, this is an example of a transradial prosthetic arm on the higher end of the price spectrum costing at least $11,000 and ultimately as much as $30,000 according to other users. Being much more expensive than the other two models, this model boasts better control algorithms and more advanced myoelectric hardware, but is still limited dexterity-wise due to relatively few degrees of freedom with two segmented and non-adaptive/rigid fingers, limited presets and non-instantaneous selection, and reliance on the off-hand for rotation of the upper thumb.

With these models in particular being similar to many others at similar price points such as with the similarity between the Bebionic 3 and Touch Sensitive i-Limb Quantum, they are good representations of the market and thus the gap in current prosthetic designs becomes more evident. This gap is summarized by the lack of dexterous ability in lower and even to a lesser extent higher cost transradial prosthetic arms due to:
- Few degrees of freedom
- Non-adaptive grips that are limited in terms of the range of objects that can be handled (ex. more delicate objects)
- Limited presets and non-instantaneous preset selection (manually scrolling through presets to find the desired mode)
- Unnecessary reliance on the off-hand (takes away from prosthetic hand's autonomous function)

As a result of these flaws, what we see then is the inability of current designs to, again, effectively balance production cost with dexterous ability.

Proposed Solution

As stated above with my transradial prosthetic arm design centered around balancing production cost and dexterity, this model will be made to incorporate the following features:

1) 16 degrees of freedom total (3 from each finger and thumb, 1 from wrist)

With 16 DOF, from this the prosthetic hand is much more biomimetic in both appearance and function with normal 3-segmented fingers that interact well with objects by allowing the full surface area of all three phalanges to be used. This relatively high DOF count in general has greater potential for more complex interactions and functions - much like the human hand.

2) Myoelectric control (using EMG sensors for movement of fingers and hand)

This will be accomplished by using two Myoware EMG sensors to measure the voltage on the opposing sides of the forearm muscles to determine when the servos will open or close the fingers. Because there are two EMG sensors here mounted on opposing sides of the forearm, these sensors can be used to both control the opening/closing of the fingers and bidirectional rotation of the wrist. For example if a single contraction is used to contract the left/right side flexor muscles, this can translate into opening/closing the fingers. Conversely if the users contracts the left/right side flexor muscles twice in rapid succession, this pattern would be detected to translate into the clockwise/counterclockwise direction rotation of the wrist (using the wrist mounted servo). The one function that will require the use of the off-hand will be the rotation of the thumb's base (metacarpal joint). This is because fitting a servo to do this motion was very difficult and determined to not be worth the added weight and cost.

3) Adaptive grip (designed to maximize surface area while gripping objects using a passive mechanical interface)

With the goal here being to allow the fingers to handle as many objects as possible, this is enabled by using servo driven gear actuation as opposed to traditional tendon and rubber band driven actuation. Because most prosthetic hands depend on rubber bands or some sort of elastic force to restore the fingers to the resting position (ex. Hackberry arm and Hero Arm), this also limits the ability to for example handle more delicate or brittle objects that break easily under force. This is especially the case when that force is not distributed over as much surface area as possible. Back to the first point, because rubber bands are not programmable to change how much elastic force is exerted/needs to be overcome in order for the fingers to move, as soon as one finger phalanx encounters resistance rather than just stop and allow the next phalanx to continue wrapping around the object, that phalanx will continue to exert an increasing amount of force on the object because the rubber band is continuing to stretch. If you are handling an egg for example, this is difficult because there is a fair chance that the egg shell will break if accidentally picked up using the bottom finger phalanges, and thus, you are forced to hold the egg or other delicate objects using the prosthetic hand's finger tips. This is even worse when considering the rigid design of the expensive Bebionic 3 because of the reliance on rigid mechanical links that make it such that if the bottom phalanx of a finger is stopped when picking up a rock for example, then the top phalanges will stop as well. My design aims to solve these issues because by using passive gear driven actuation, a very, very small amount of force (force of friction between gears) is required to overcome the resistance that each phalanx offers, and thus, the fingers are able to completely wrap around the object at hand before the servos apply any real force on the object. Using the previous analogy, this design would allow the fingers to completely wrap around the egg while hardly applying any real force to the egg at all, giving a much better grip. Furthermore depending on the preset that is active, how much force is applied can be restricted for example in a "delicate mode" preset where the current sensors tell the servos to stop applying force when a programmed threshold is reached.

4) Force sensing (using current draw sensors to detect load on servos)

As briefly mentioned above, force sensing will be enabled by the use of 5 current draw sensors for the 5 servos because of how the amount of force exerted by each of the servos against resistance is proportional to the amount of current that is drawn by the servos. Accordingly, for each of the presets that are programmed for handling more delicate or stronger objects, different current draw threshold values can be used with these current sensors. This will allow the current sensors to stop the servos once the threshold values are met.

5) Adaptive fit (inspired by Hero Arm design brace wrapping design)

Similar to the way the Hero Arm works, this adaptive fit works by allowing two outer covers on the left and right side of the stump's brace to tighten and loosen around the brace using a cable connecting both, and is controlled by turning a knob on one of the covers. This is useful because as the amputee's stump and brace change size, the outer covers can simply move apart to accommodate and wrap around the brace, without having to be replaced as often. Even when a new set of outer covers is needed to accommodate a large enough change in brace size, the outer covers can simply be 3D printed and mounted again using 4 bolts because of the modular design of these components.

6) Vibration feedback (proportional to force applied by servos)

Here, vibration feedback will be implemented as a way of allowing the amputee to "feel" how much force the servos are applying. This will work by using PWM on the micro-controller (an Arduino Nano) digital pins to alter the speed of the vibration motor depending on the current draw readings from the current sensors. Also to conserve battery, these vibration motors will only be active when there is a finger-controlling EMG signal active.

7) Voice control (using voice recognition module for much more instantaneous preset selection)

Because of the problem of having to manually scroll though different presets evident with the Bebionic 3 for example, this is somewhat tedious for amputees, and as a result acts as a limiting factor to the amount of presets that can be created. Rather, the approach with this design relies on voice recognition methods where upon saying a command phrase, the micro-controller will quickly activate the corresponding preset. This allows for many more presets to be programmed for different tasks or situations, because so long as the amputee can remember the preset commands, any one preset can be quickly selected by saying the correct phrase. A voice control module will be used for this.

8 ) Completely 3D printed frame (PLA)

Self-explanatory but just to elaborate anyway, the frame of this transradial prosthetic arm will be completely 3D printed using PLA due to its cheap cost and ease to print with. Of course with most 3D printed designs that are made out of relatively cheap thermoplastics like PLA or ABS, this means that should a part break or need to be replaced, this can happen at a very low cost and even be done by the amputee.

9) Some extra details

In an effort to have individual finger control using 5 servos, I also tried to keep a balance by providing as much torque to the fingers as possible while keeping the size of the servos down so that the arm is not too large. From this I have selected and designed around fairly powerful servos that at the 6V being provided will provide about 20 kg-cm of torque per each of the four fingers, and at least 4 kg-cm of torque for the thumb. This should make the hand also relatively strong compared to current models which often rely on 1) one servo to selectively control the 4 fingers and thumb, or 2) much more expensive linear actuators to accomplish the same function. As stated before, the prosthetic arm is going to be controlled using an Arduino Nano and will be powered by a 7.4V 4000 mah rechargeable battery pack. To power the servos and motor, this voltage will be reduced to 6V using a step-down voltage regulator. For force sensing, the current draw sensors will use the analog-in pins on the Arduino Nano. A protoboard on the side of the servo housing will be used for the circuits involving the servos, voice module, vibration motor, etc. Concerning the manufacturing of the gears, I am currently working on increasing print resolution so that the gears (which are fairly small) can be accurately printed using PLA, but for now this has been done by laser-cutting acrylic. For the connection between the forearm mounted servos and the finger mounted gears, this will be done using flexible driveshafts. Also by having the bulk of the weight from circuit boards and especially servos located near the stump in the forearm region, this reduces the amount of torque being exerted on the stump which is important in keeping the arm as comfortable and lightweight for the user as possible.

Expected results/impact:

After finishing building, testing, and refining this transradial prosthetic arm, I hope to see it become applied to an actual amputee. This would allow testing I imagine over short and then gradually extended periods of time to observe how the arm holds up in daily conditions.

Relative to the other transradial prosthetic arms currently available to consumers, I think that this arm will prove to be a competitive option that provides much more value given its relatively large amount of functionality as a prosthetic arm on the lower end of the cost spectrum. I also think that by moving forward to somehow make this design open source, this would increase its impact because of the accessibility and rapid community-contributed design improvements that come with ideas on open-source platforms.

Estimate of work effort involved:

What is already done

So far, I have the design fully modeled in CAD software which will be necessary for 3D printing components and for determining how to assemble the arm. As far as 3D printing/laser-cutting is concerned, all gears have been laser-cut and about half of the 3D printable parts have been printed, with some having to be redone due to quality issues.

What needs to happen

  1. My next step will be to draw out the circuit diagrams involving all of the electronic components which will be very useful in speeding up assembly time. It is also of course good practice in general to have this kind of documentation for future steps.
  2. As mentioned above, the next step for this project will be to finish 3D printing the rest of the parts. This is step 2 and not 1 because students are currently on break and I am waiting for 3D printing access when classes begin again.
  3. The next step will be to order parts which hopefully will occur if I am able to secure funding. For the parts that need to be ordered, I have the full parts list ready calculating the total production cost of the arm to be $425.02 (including 3D printing material cost approximation because not everyone has access to free filament). As far as how much in parts needs to be actually ordered, this number is $398.68 because it subtracts what parts I already have (one myoware muscle sensor) and also subtracts the estimated cost of 3D printed materials (because 3D printing materials are free for students). On the cost spreadsheet linked further down, the total amount of funding requested is $448.68 (added $50 to $398.68) as parts will very likely have to be replaced, and wiring supplies will have to be purchased (very difficult to quantify ahead of time). When ordering parts excluding what I already have, I will be ordering everything on the list except for the battery pack and flexible driveshafts which will be ordered after testing with code to confirm what the capacity of the battery pack should be, and modeling using tubing to measure how long each length of flexible driveshaft should be, respectively.
  4. After ordering parts and everything arrives, the next step will be assembly. Even though the CAD model shows everything to fit together correctly, there is still a chance that during assembly I might find that some parts need to be adjusted or re-printed. During this step, all such modifications will be made.
  5. After this is finished, I will need to write the program to control the arm using the Arduino Nano and subsequently test and refine this code.
  6. Next, I will use this code combined with extended testing to confirm what battery pack capacity is needed to power the arm, and after that is selected, the battery will be ordered. Subsequently, after measuring tubing to get an idea of what shaft lengths are needed, the flexible driveshafts will be ordered.
  7. Finally, the last step (after receiving and integrating the battery and driveshafts) will be to waterproof the arm's electronics to make it more durable in daily conditions. Any cosmetic modifications and finishing touches will now be added.
  8. With the prototype arm now complete, the next step will be to get in touch with a matching amputee (hopefully with e-nable's assistance) and conduct tests to receive feedback for improvements.
  9. After this cycle of testing and feedback reaches a point of satisfaction on the amputee's end, the design should then be ready for release through an open source platform.

Estimated timeline for completion:

The numbers corresponding with each time period below refer to the numbered list above and their respective tasks. Also, because this entire schedule is dependent on if/when funding is approved, for the sake of the timeline I am just going to assume that funding is approved on January 22nd if I submit this proposal ASAP with the community's approval. This is also when I get back to college:

  • January 1st - January 22nd: (1)
  • January 22nd - end of February: (2) (3) (4)
  • beginning of March - end of March: (4)
  • beginning of April - end of April: (5)
  • beginning of May - end of May: (6)
  • beginning of June - end of June: (7)

For steps (8) and (9), I really cannot give an accurate estimate for the timeline because of how many factors this relies on pertaining to the process of finding a matching amputee, the amputee's condition/availability, etc. which all have a very direct impact on how long testing and refining will take. Also as a note, the reason why assembly is predicted to take about two months is because of some of the parts suppliers I am looking to order from being located overseas in countries like China. From past orders, these deliveries have taken anywhere from 2 weeks to as much as a month. Lastly, not to whine too much but this upcoming semester's classes are expected to be much more difficult than the first, so should I be awarded funding for this project, I hope the people overseeing my progress understand if I fall behind my predicted schedule because of school work.

Names of individuals responsible for deliverables:

Bhargav Parthasarathy (me)

Amount of funding being requested:

Please refer to the following spreadsheet for all details regarding costs and funding:
https://docs.google.com/spreadsheets/d/1ue6cWVnkKPzd1UW-p4PodXCmFX0dN42_Daxiv5_ecCE/edit?usp=sharing
Please be sure to click on the highlighted cells to read the notes I have included for important explanations.

A brief overview of my background with e-NABLE:

I have been a member of my university e-NABLE chapter since the start of the 2018 fall semester, so my involvement began fairly recently.

BP

Bhargav Parthasarathy Thu 27 Dec 2018 11:44AM

Also here are some pictures of my design in CAD if anyone is interested.

YM

Yoav Medan Fri 28 Dec 2018 1:10PM

Thank you for the proposal. Based on our experience, you may want to look carefully into the following:
1. Total weight as well as cneter of gravity. Heavy hand with an unnatural location of CG seem to be unacceptable to users
2. Energy source and capacity. Besides contributing to weight, it may discharge too fast.
3. Phantom pain - Many amputtees suffer from phantom pain when using EMG driven prosthetics. We are therefore working on an EEG-driven control. Currently we are using an ankle worn motion controller
4. It is possible to reduce the #DOFs to 1 per finger and still get a biomemtic design. Happy to discuss off-line.
5. The Thumb is the most important and perhaps most complex finger. From the illustration it seems to have minimal articulation.

Happier and Healthier New Year wishes to the e-Nable community
Yoav

BP

Bhargav Parthasarathy Fri 28 Dec 2018 10:13PM

Mr. Medan thank you for your comments, and I will try to respond to your points.

Regarding total weight/center of gravity, I realized upon reading this that I forget to include that in the thread, but I just ran the mass calculations in CAD and found the total mass to be 1174.55 g. As a reference point after doing some quick searching, I found that the average weight of the forearm and arm according to https://apps.dtic.mil/dtic/tr/fulltext/u2/710622.pdf was about 800 g and 450 g respectively. With the combined weight of the residual arm, however, the total weight of the amputee's arm and forearm should be closer to that approximation of 1250 g. According to this link http://bme240.eng.uci.edu/students/10s/slam5/considerations.html this prosthesis is a little bit heavier than supposedly the optimal weight of 1010 g. I also attached a picture of the center of gravity below, placing the point in the lower forearm region closer to the brace than the hand.

For the battery I will be selecting, you are certainly correct in that weight/discharge rate could be an issue, however, I have yet to actually confirm what capacity the battery should be and will determine it after assembly and code is written so that I can test for extended periods to find the current draw over time. So far, the highest capacity of 3.7V battery I could find was 4000mah which in order to produce 7.4V would require two to be connected in series, keeping the capacities from adding up.

For phantom pain, thank you for bringing this to my attention. Currently this design is intended to be EMG driven but for my purposes and if I even get to an amputee testing phase, should I have complaints regarding EMG resulting phantom pain, changes will of course be made. Because the EMG sensor setup I have right now does not distinguish beyond solely the presence of a signal in order to drive functions, the challenges that I see with converting to an EEG sensor would be not be too difficult beyond what wiring needs to be changed.

Regarding shortening the DOF to 1 per finger, I do not think this would be too desirable or allow for a very adaptive grip that could wrap around objects because only one joint would be available for extension/flexion. With an actual hand having 4 DOF per finger at 3 for each joint and 1 for adduction/abduction, the current design accomplishes this except for the 1 for add/abduction. The added cost as a result of 3 DOF per finger here is also not too high as all parts are 3D printed with the exception of 3 bolts and 2 standoffs.

For the thumb's range of movement, I think the picture may illustrate this poorly but the range of motion here is standard with most other prosthetic hands. This is given by 1 DOF for the swiveling movement in which that triangular thumb base you see can rotate 45 and 90 degrees, and 2 more DOF for the extension/flexion of each of the 2 phalanges.

CL

Chang Liu Fri 28 Dec 2018 2:54PM

Hello! Great job on this detailed proposal! I can't find the vote button, not sure if I'm being dense or if voting is open?
I have a few queries on your design: What is the weight of this prosthesis? Perhaps you can make the weight even closer to the socket, but also debulk the socket area as that might be a hindrance and non-cosmetic. A dichotomy, I know.
Is your forearm area designed to be attached to a very well-fitted and suspended socket in all ranges of motion? If not, your electrodes will lose contact and become frustrating to operate. I see from the design pic that this design won't fit someone with a long residual arm due to the electronics being housed in the distal forearm, but if you're okay with that, that's fine and a good place to start.
I have undergone an Ottobock training for the Bebionic hand, and it's actually very good for adaptive grip even with the smaller degree of freedom. It doesn't have the highest grip strength but the adaptive grip is sufficient to make it very functional. And it is already very breakable and high-maintainance, with the amount of parts it has, so perhaps you'd consider making your finger parts less complex for a similar functional outcome? The most impt parts for grasp patterns are mainly the index finger and thumb.
From what I see thus far, your prosthesis is great in that it seems like a much lower cost alternative to a high-end myoelectric hand, which is really great. A few of the advantages of the Bebionic hand actually comes from its thumb movement: by moving the thumb to and from opposition, it allows different grasp patterns to be generated using the same muscle contractions. (primary and secondary grasp patterns for each thumb position, and co-contraction and 4-channel controls). It's worth looking into it, if it can simplify your hardware and make it lighter! Or it might make it too complex, which is more often not the answer. The accuracy of the signals being picked up will need fine-tuning too, I'm not sure if the noise of the voice control will affect that. I'm not the most familiar with this, but I know that inaccuracy of each muscle contraction translated into the correct movement can be frustrating.
I think this is a great proposal, but it'll take some work to make it realistically functional for a user, especially to iron out some common clinical problems faced by users before you face the frustrations of these during user trial, so your timeline might be a bit tight, if you wish to amend that. I'd be happy to help identify some of them in greater detail if you'd like as you go along, as I'm sure the prosthetists here would, they have much more experience than me in this. :) Good luck!

BP

Bhargav Parthasarathy Fri 28 Dec 2018 11:31PM

Mr. Liu thank you for your comments, I will try to respond to your points. Also, you are correct in that there is no vote button; I wanted to post my draft proposal as a thread before actually submitting a proposal for voting so I could get some input on what should be improved first.

Regarding weight, as I am replying to all comments here I just detailed some information about weight I previously excluded in a reply to Mr. Medan which should give you some information. For debulking the socket area, actually your comment got me thinking about the right side of the socket (from the perspective of picture New_53), because that whole bulky portion is just the ratcheting mechanism to tighten/loosen the cables that connect between the outer braces, but this could just as easily be excluded and let there just be one of those spring locks for clothing instead like this image https://images-na.ssl-images-amazon.com/images/I/71QY1THUpEL._SX522.jpg which I see working just as well. For the left side of the socket, however, those gray and green portions you see are the battery and voltage regulator. Right now with everything in the forearm being tightly packed, it is difficult to relocate this elsewhere. Also being one of the heavier components, I thought it would make more sense to mount this as close to the socket as possible as this also helps to shift the center of gravity closer to the socket.

Regarding the attaching of the forearm to the socket, I don't think I discussed this enough previously but this is how the components are supposed to connect. The forearm is directly connected to the outer covers which hold the socket using a series of bolts to hold these components together, and another set of bolts to allow the outer covers to swivel open so that the socket can be removed easily, which in turn requires the cables holding the outer covers together around the socket to be loosened. I attached a picture below to help illustrate this.

Regarding the grip technique, I agree with you that the bebionic hand's finger construction is suitable for most scenarios after having seen videos of amputees using these arms to interact with different objects. That said, although probably not a very big issue, the bebionic arm's finger construction is relatively rigid because of the solid link between the top of the bottom phalanx and upper phalanx which forces the upper phalanx to rotate with the bottom phalanx in either direction. Because of this in some instances where given let's say an object's irregular geometry like with a rock for example, if you tried to wrap the bebionic hand's fingers around the rock chances are that both phalanges do not contact the rock at the same time unless you tried to orient/pickup the rock that way. This would mean that you would really only be picking up the rock with each finger's distal or proximal phalanx but probably not really both which means you also aren't using the full potential of each phalanx's surface area to enhance grip. The bebionic finger construction has been well done, but with my goal of improving wherever I can this is one area I thought could at least be improved slightly. Also regarding complexity and what you mentioned about the bebionic fingers being breakable and high-maintenance, in this case I do not think the complexity here is directly related to the breakability/maintenance of this arm. This is because although seeming complex, each finger is really a train of gears/standoffs secured by bolts on either side that will eventually be secured using loctite to ensure they don't slip out. With this, I cannot see really anything in this subsystem having to be maintained frequently, and because of the 3D printed construction can easily be replaced should something break. The ways I imagine this subsystem breaking is if enough torque is exerted to snap the gear teeth off, or if the finger is bent backwards suddenly which would force one or more of the three phalanges to break. To help reduce the chance of this however, I do plan on having a limiting current draw value to stop the servos from breaking the gear teeth if the motors can even supply that amount of torque, and printing those beige lower finger mounts out of flexible filament to cushion the blow if the fingers are bent backwards.

Regarding your comment about how the index finger and thumb are the main contributors in most grip patterns, I was thinking about this and did realize that there was potential to save some space/weight by replacing the high torque servos powering the middle, ring, and pinky fingers with lower torque servos. This would save 125.4 g (not too much money as both are similarly priced), and allow more room for housing electronics, but would significantly decrease grip strength. I think I will make this change anyways though; thanks for the suggestion.

Regarding the thumb, actually this thumb I think can do what you are talking about, as it is able to be moved to and from opposition which as you said allows for different grip patterns. This is done by using the off hand to rotate the thumb to either 45 or 90 degrees. For the fine tuning of the EMG signals, this also should not be a problem as I am not relying on the EMG sensor identifying different types of muscle contractions to translate into different finger movements. Beyond just an EMG sensor on either side of the forearm to detect when one or the other muscle group is active, no further distinction is made. As you stated with the difficulty experienced by EMG sensors in making this distinction in the forearm, I tried to avoid this by using presets instead which the user would select via voice control for whatever general type of task they were doing. For example if handling a computer mouse, this would work as the amputee would use the preset phrase for a computer mouse which would assign a different current threshold value to the current draw sensors, and work such that a muscle contraction detected by one of the EMG sensors would be a down-click and the other sensor's signal would be the release, or something similar to this.

Also, after posting my thread and thinking about this timeline a bit, I think for when I actually submit my proposal I am going to re-state my end goal to probably just be getting this to a developed proof-of-concept because of how difficult it is going to be for me to actually conduct long-run amputee testing and then actually getting this approved for clinical application. Also with many members having such a strong understanding of socket design and fitting and this being a field I am not that knowledgeable in, I think that releasing these files on a website and allowing someone to kind of take over that part of design might be better, especially if that someone already has experience with for example fitting amputees using methods like 3D scanning. Because the socket/outer cover portion of this design can easily be removed anyways by 4 bolts, I don't think this would be too difficult to change.

E

ebubar Fri 28 Dec 2018 3:43PM

Great effort into this proposal! Kudos for the detail that you've provided. I think its wise to let other models out there inform your designs. I'd encourage looking into the Po Paraguay MyPo arm as well (https://www.thingiverse.com/thing:2409406).
I've also got some tips that may be helpful. I've had some experience with the myoware muscle sensors and would suggest that you'll want the sensor cable shield kit and cable (https://www.sparkfun.com/products/14109, https://www.sparkfun.com/products/12970) to increase comfort and robustness of the device. Without the cable shield, you are very limited in placement of the electrodes which is a problem. We also found them susceptible to breaking and giving unreliable signals without the cable shield. I may go further and encourage that you try to get this system working with a conductive fabric compression sleeve like this (http://www.advancertechnologies.com/2013/03/diy-conductive-fabric-electrodes.html) to further improve comfort. I also think the weight and power load that you'll have in this design are going to be the biggest considerations for getting the system working well, barring any considerations with fitting to a user. Any clever workarounds to lower the number of servos would be wise as this decreases both those issues. I would definitely suggest borrowing heavily from the work of groups like open bionics and hackberry, as they've had large teams working on these issues and (especially in the case of open bionics) managed to bring their products to markets. There are good reasons that they've made the choices that they have made. I think the timeline is a bit tight if you haven't already worked out the electronics for the myoware muscle sensing. They're not easy to get reliable readings from. I've had a student working on it for about 8 months (albeit sporadically) and they're telling me they're at about 80% reliability for recognizing a muscle signal to turn into a pose. I'm also curious if your design considerations and solutions are meeting the requirements of the intended users. Its really easy to get lost in thinking about what we might think is most useful and needed (i.e. getting a hand that is as human-like as possible) while neglecting what an end-user actually wants or needs. Are all those poses really necessary, given the added complexity in motors? I'm a fan of balancing function with simplicity as there is definitely a line to balance along. It might be worthwhile to seek out a potential user. Along those lines, I think you can conservatively add at least 1-2 years to your timeline once you start working with a person. At a University you'll need to leap over some liability hurdles (hopefully with assistance from a faculty supervisor). This is non-trivial to get accomplished. Best of luck with the proposal.

BP

Bhargav Parthasarathy Sat 29 Dec 2018 3:45AM

Mr. Bubar, thank you for your comments, I will try to respond to your points.

Regarding the myoware muscle sensor shield, right now I have two emg sensors designed to mount on either side of the stump for each of the two major muscle groups which after a quick search I believe are called the flexor policis longus and flexor digitorum profundus. I was just replying to Mr. Liu about this saying: "I am not relying on the EMG sensor identifying different types of muscle contractions to translate into different finger movements. Beyond just an EMG sensor on either side of the forearm to detect when one or the other muscle group is active, no further distinction is made. As you stated with the difficulty experienced by EMG sensors in making this distinction in the forearm, I tried to avoid this by using presets instead which the user would select via voice control for whatever general type of task they were doing. For example if handling a computer mouse, this would work as the amputee would use the preset phrase for a computer mouse which would assign a different current threshold value to the current draw sensors, and work such that a muscle contraction detected by one of the EMG sensors would be a down-click and the other sensor's signal would be the release, or something similar to this." From this, I have tried testing the Myoware EMG sensor on my own forearm before using simple wrist flexion and extension in either direction and was able to receive a simple readings that were high upon flexing the muscles, and low during relaxation without very much error. Because really these are the only signals I am looking to detect without trying to translate specific muscle movements into certain motions, I feel that the EMG sensor alone should be enough for this, but also because with everything so tightly packed together, it is difficult even after reducing servo size to house the Myoware sensors elsewhere besides directly contacting the stump, especially with the added size of the shields. About these sensors breaking easily, however, I was not aware of this and will take this into consideration by trying to add a protective case on the other side of its mounting place.

For the possibility of using a conductive fabric compression sleeve, I think this could work with my current design if the electrodes woven into the fabric could snap connect to the Myoware EMG sensor electrode points in the socket. However as I said with Mr. Liu: "after posting my thread and thinking about this timeline a bit, I think for when I actually submit my proposal I am going to restate my end goal to probably just be getting this to a developed proof-of-concept because of how difficult it is going to be for me to actually conduct long-run amputee testing and then actually getting this approved for clinical application. Also with many members having such a strong understanding of socket design and fitting and this being a field I am not that knowledgeable in, I think that releasing these files on a website and allowing someone to kind of take over that part of design might be better, especially if that someone already has experience with for example fitting amputees using methods like 3D scanning. Because the socket/outer cover portion of this design can easily be removed anyways by 4 bolts, I don't think this would be too difficult to change." Should the type of sensor also be exchanged for some other kind, assuming that a similar operating voltage is maintained and a similar wiring setup is maintained, interchanging this should also not be too difficult.

Regarding the factors of weight and power load, this was also an important point Mr. Liu brought up to which I responded: "Regarding your comment about how the index finger and thumb are the main contributors in most grip patterns, I was thinking about this and did realize that there was potential to save some space/weight by replacing the high torque servos powering the middle, ring, and pinky fingers with lower torque servos. This would save 125.4 g (not too much money as both are similarly priced), and allow more room for housing electronics, but would significantly decrease grip strength. I think I will make this change anyways though;"

Regarding drawing inspiration from models like the Hackberry Arm or those from Open Bionics, contrary to what my original critique might seem to convey I do actually admire these arms and have drawn design inspiration from these models in developing my own. Specifically, this was the case with the idea of a tightening/loosening set of outer braces featured on the Hero Arm, as well as with the idea of using gears mounted in one phalanx to articulate the next phalanx upon encountering resistance in the Hackberry Arm. I do believe, however, that there is definitely room for improvement in some areas which were those features I chose to discuss in the draft proposal.

About the requirements for the intended users, you are definitely correct about there being a line to balance function with simplicity. Actually with this model being my fourth prototype, I also realized this as I originally came from a design that looking at it now was absolutely crazy with each phalanx being able to be independently articulated from a total of 16 DC motors. Here, however, I have tried to make this a lot better and along the lines of what a lot of the field is doing with independent finger control but not really much else in terms of motor complexity. With the number of poses I mentioned in my original draft proposal, I think I meant to say that although this design has servos dedicated to independent finger control like other models, adding voice control to more quickly select presets also happens to mean that there is more potential for more poses, but not necessarily so many poses that it becomes unnecessary. In this way, I meant to say that more potential for poses is more a result of easier/quicker preset selection rather than having those 5 servos.

But as I was saying before with the amount of difficulty I expect to experience with getting to a point of working with someone on this, I am not ready to give up on the idea of trying to bring this to test with an amputee. But for the purposes of submitting a proposal, I will probably state my end goal as proving this design as much as possible to be a developed proof-of-concept just because I think there is some value in this design, and because at least that much I know can be an attainable goal.

BP

Bhargav Parthasarathy Fri 28 Dec 2018 11:32PM

Hi Patrick, you can find the link to the cost breakdown under the "Amount of funding being requested" section of my draft proposal.

BP

Bhargav Parthasarathy Wed 2 Jan 2019 10:56PM

Update:

Thank you all for your comments as they have helped me to review my design and make changes accordingly.

I completed a basic wiring diagram between all of the electronic components I will be using to make sure that I do not have too few pins anywhere or missing components which right now looks alright. As described in the reply to Mr. Liu, I made the change to debulk the area enclosing the socket by removing the ratchet-tightening mechanism to tighten the cables pulling the two outer braces together, because the same can be accomplished just using a spring lock like those seen on backpacks. Regarding the servos I am using, just to reiterate I am using two 13.8g servos for the wrist's rotation and thumb which have less torque than the four 55.6g servos for the four fingers. In the reply to Mr. Liu, I said that I might change from the the higher torque 55.6g servos for the middle, ring, and pinky fingers to the 13.8g servos, however, I am not so sure that I can do this anymore because of the 13.8g servos are not continuous rotation types. It is possible to tamper with them to make them continuously rotate, however this involves removing potentiometer functionality which will in turn take away speed control, which is needed. Even looking at other continuous rotation servos, this is much more limited in terms of the balance between torque and size as all servos I have seen that are on the smaller side have significantly less torque (2-4 kg/cm) for the same or a higher price. Because of this, I will likely keep the servo setup I have as I am unable to find better continuous rotation servos that offer a little bit less torque in exchange for smaller size, less weight, and a similar price. Lastly as stated before, I do plan to update the end goal of this project: "...with the amount of difficulty I expect to experience with getting to a point of working with someone on this, I am not ready to give up on the idea of trying to bring this to test with an amputee. But for the purposes of submitting a proposal, I will probably state my end goal as proving this design as much as possible to be a developed proof-of-concept just because I think there is some value in this design that others could improve on, and because at least that much I know can be an attainable goal."

At this point if no one else has any additional input, I will be getting ready to submit an actual proposal with the information and changes I have detailed.

Load More