Building an Assistive Robotic Arm

We are building an assistive robotic arm for my daughter who recently became paralyzed in her left arm.

Similar projects worth following
My daughter, Lorelei, was diagnosed with Acute Flaccid Myelitis a rare polio-like syndrome and lost the use of her arm. Traditional treatments would have given her about a 5% chance at recovery, and state of the art treatments were closed to us or were too costly. Without any experience, we decided to tackle this challenge by setting up an open project and reaching out to experts for help.

[Update] Lorelei is now able to move her arm again! It has regained about 50% of its strength, however, her shoulder remains paralyzed.

Follow our work here:


Arduino code for muscle​ sensor monitoring.

ino - 10.49 kB - 09/08/2016 at 15:29


  • How the connected world empowered my daughter

    Bodo Hoenen11/22/2019 at 14:56 0 comments

    Lorelei and Bodo Hoenen present their robotic exoskeleton at HLTH2019

  • Job Opportunity - Help us develop open-source code for project

    Bodo Hoenen01/29/2018 at 15:41 0 comments

    We are looking for some help to develop open source code for this project 

    We aim to develop an open-source platform for muscular activity signal detection —recorded non-invasively on the skins surface— and using that for exoskeleton control.  

    The muscle signals will be recorded using the Myo Armband and processing/exoskeleton control will take place on a Raspberry Pi. The goal is to detect 2 to 5 patterns from myo signals using machine learning / pattern recognition methods.

    Many open-source libraries are available and can be used in the project subject to minor modifications. Guidance and support will be provided.

    Here is a Wiki that we have for this project

    We already have a working model for this, but it is using proprietary software, We now need to develop this software as open source so that we can freely share it. 

    Required skills

    • Python programming
    • Basic Bluetooth communication principles
    • Basic signal processing (i.e. filtering, sliding window processing)
    • Basic machine learning / pattern recognition (e.g. classification using the sklearn library)
    • Raspberry Pi (& Linux operating system)
    • Version Control (e.g. Git/Github)

    Breakdown of work

  • The end of our story and the beginning of many more

    Bodo Hoenen09/26/2017 at 17:43 0 comments

  • Build your own robotic assistive arm

    Bodo Hoenen02/18/2017 at 19:01 0 comments

    Do you want to build a robotic assistive arm, we have the instructions here:

  • We were on TWIT

    Bodo Hoenen02/18/2017 at 19:00 0 comments

  • Our story so far

    Bodo Hoenen11/23/2016 at 01:33 0 comments

    Over the course of the last few months, a group of innovators from around the world have joined my daughter and I to build an open source robotic assistive arm. This is our story so far, come join us and make it your story too! :

    Assistive Robotic Arm

  • We are getting there!

    Bodo Hoenen11/22/2016 at 15:36 0 comments

    We need help, if anyone knows how to code machine learning to recognize patterns in myoelectric signals, you could change the world with this!

    Over the course of the last few months, a group of innovators from around the world have joined my daughter and I to build an open source robotic assistive arm. This is our story so far, come join us and make it your story too! :

  • Moulding 3D print

    Bodo Hoenen11/05/2016 at 20:42 0 comments

    We printed out our first attempt at the arm brace. We decided to use PLA plastic as it would allow us to print out the brace flat and them mold around Lorelei's arm after we heat it. To make that easier and to make sure I don't burn lorelei, we made a cast of her upper and fore arm, and molded the brace around those instead. It worked really well!

  • Testing out designs

    Bodo Hoenen11/03/2016 at 11:08 0 comments

    We have been testing out the arm brace design a little by printing it out on paper. And Now that we have a design that should work well we sent it off to the 3D printer, and will hopefully get it by the weekend.

  • Using Machine learning to boost MyoElectric signal recognition

    Bodo Hoenen10/21/2016 at 02:04 0 comments

    Super Excited today. Lorelei and I figured out how to reliably get a signal from her arm to control the robotic prosthetic! (Disclaimer the video and lighting is poor but I was just excited to quickly show you all).

    Background: Up until now we have been struggling to get an accurate signal from her arm to control the robotics. We are looking for the signal that is being sent from Loreleis brain to her arm, but because of the motor neuron damage, the signal is really weak. The sensors we used until now essentially filtered and normalized the raw signal so that it can isolate spikes within the signal. When a spike reaches a given amplitude (threshold) we would trigger the robotic actuator to move the prosthetic arm, thereby helping lorelei move it. However, given the weakness of my daughters signal in her muscles, we had to set the threshold really low which in turn caused the robotic actuator to be triggered by signals from her heart or other muscles in her arm or body. Using this traditional approach resulted in a significant roadblock to our project.

    I was thinking that there must be a better way... Thinking of approaches that are used within things like image and voice recognition, we can train algorithms to recognize objects in pictures (this is how Google can tell you what is in your photos) or how we train algorithms to recognize speech (Like Siri and Google Now). In both these examples they train the algorithms with vast amounts of data, they do not filter the data or normalize it. What if we did the same with Lorelei's signals, could we not use a similar machine learning technique to look at the unfiltered, full signal that Lorelei's arm produces and have an algorithm learn to recognize the hidden signals within that noise.

    It turns out there is a company CoApt who are doing something really similar. After a few emails and a call, they kindly sent us an evaluation unit. After placing 17 sensors on her arm and a bit of calibration we got it to work. We were able to control a virtual arm you can see in the video. This is so amazing! Really amazing stuff CoApt, thank you!

View all 17 project logs

  • 1
    Step 1

    Do you want to build a robotic assistive arm like this? You can find all the instructions and step by step guids here:

View all instructions

Enjoy this project?



Jonathan Kelly wrote 09/08/2016 at 22:59 point

Really cool stuff.  

Just some thoughts (feel free to ignore, particularly as I am not an expert), if you have problems with the myoelectric sensor, one alternative might be to explore acoustic myography.  It may not work but something to think about if you hit a brick wall with the detection of muscle activation.  I have played a little bit with it to see if I could detect activation and it seems to have possibilities for stuff I am doing and there are a number of people who have projects out there in the public domain that are more advanced and refined that you could look at if interested.

Also wondering, if the signal from the bicep/tricep is difficult for her to generate reliably, have you thought about getting the signal from other muscles?  It would involve her learning to activate those muscles to move the arm but may be easier to get detection of muscle activity.

Again - cool stuff and feel free to ignore my ideas if they are not relevant (and again I am not an expert in this stuff just a tinkerer) :)

  Are you sure? yes | no

Bodo Hoenen wrote 09/09/2016 at 01:29 point

Hi Jonathan, thank you for taking the time to comment and share your insights! I'm looking at Myography now with interest. 

You are right we are struggling to get signals from her muscles because they are so weak, up until now we have been trying the standard approach (look at the amplification of the signal on a bell curve - this may not be the right way to describe it) This is not the raw signal but rather a filtered signal, which works really well for me, but does not work for my daughter. I come from a computer science, technical background and thought about using machine learning to analyze the full raw signal and over time have the system learn the unique patterns that my daughter can produce. then program a sensor to look out for those patterns and trigger the actuator accordingly. It turns out that there is a company doing just that CoApt, I just spoke to them (very cool people) and they are very kindly going to help me look at what options we may have with my daughter and the signals we can get from her arm.

Like you say, we could instead use signals from another part of her body, like her wrist, as those signals are stronger, however a large part of this project is about stimulating rehabilitation. The prognosis is that there is a very low chance that she will have a recovery (Based on other cases) and we dont want to just take that lying down. We have read research of how exoskeleton devices like we are building, that simulate, and use existing muscles are able to rehabilitate the growth of the motor neurons, where other approaches like traditional PT and OT are not able to.  Our goal is full recovery. Either through rehabilitation or assistive tech. And if it takes another 5 to 10 years, so be it. It beats accepting the prognosis and giving up :) 

Thank you so much for taking the time to share your experience and support what we are doing!

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates