Mobirise Website Builder

Brain Controlled Robotic Prosthetics

September 2023 - Ongoing

GiyHub Repository

I am currently engaged in a year-long project that revolves around the development of mind-controlled robotic prosthetics, with a primary emphasis on harnessing EEG data to facilitate motor control functions as an alternative to myoelectric systems. Throughout the course of this project, one significant challenge that has presented itself is ensuring the safety and reliability of the mind-controlled prosthetic system. The aim is to reliably identify and interpret neural signals by leveraging advanced machine learning algorithms, which will then be used to control cost-effective prosthetic devices. This project is deepening my understanding of the field of robotics and healthcare, making advancements that enhance the quality of life for individuals with physical challenges.

$ 1.82B

Market Size

for Prosthetics and Orthotics in the USA in 2022

5.5%

CAGR

for the Prosthetics market

$ 2.93B

Projected Market Size

for Prosthetics and Orthotics in the USA by 2032

New Technology

Current prosthetics are either rigid or controlled through myoelectric systems

Currently robotics prosthetics use data acquired from nerves and muscle tissue to control motors

The proposed solution is to develop mind-controlled prosthetics that utilize brain data acquired through EEG scanners for precise motor control

Target Group

Amputees that are currently ineligible for robotic prosthetics for one or more of the following reasons:

  • Cost: High costs associated with the development, production, and maintenance of robotic prosthetics can be a significant barrier. Insurance coverage may vary, and not everyone can afford the expenses involved.
  • Limited Availability: Advanced robotic prosthetics may not be readily available everywhere, and accessibility can be a challenge. Limited distribution channels and specialized fitting requirements may make it difficult for individuals in certain regions to obtain such prosthetics.
  • Medical Eligibility: Not everyone is medically eligible for a robotic prosthetic. Certain health conditions or the absence of specific anatomical structures may restrict the suitability of these prosthetics for certain individuals.
  • Complex Fitting Process: Robotic prosthetics often require a precise and personalized fitting process. Individuals with unique anatomical considerations or those in remote areas without access to specialized fitting services may face challenges in obtaining a properly fitted robotic prosthetic.
  • Training and Adaptation: Learning to use a robotic prosthetic effectively requires time and effort. Some individuals may find it challenging to adapt to the technology, particularly if they have cognitive or physical limitations.
  • Incompatibility with Daily Activities: Certain occupations or activities may pose challenges for individuals using robotic prosthetics. Some prosthetics may not be well-suited for specific work environments or recreational activities, limiting their practicality for certain individuals.
Mobirise Website Builder
Mobirise Website Builder

Figure 1: Open-source 3D Printed Robotic Arm

Mechanical System Design

The first version of the prosthetic arm was built using 3D Printed parts modelled after the InMoov open-source project. A wire system was used to allow for the fingers to open and close without adding an excessive amount of motors. Each finger has two wires in it: one on the front and one on the back. When one of the two wires is pulled, it will shorten the distance between certain points of the fingers, allowing for grasping. This way, the entire hand can be controlled using only 6 Servos (1 for each finger and 1 for the wrist). 

Mobirise Website Builder

Figure 3: OpenBCI Headset


Electroencephalogram (EEG) data was obtained using the 8-channel OpenBCI Cyton board connected to dry electrode leads. A 3D printed headset was used ensure correct electrode placement, and electrodes were placed at the following positions:

FP1, FP2, C3, C4, P7, P8, O1, O2

The Cyton board communicates in real time with the computer using a wireless dongle, maximising ease of use and minimising bulkiness.


Software Design

In order to read and interpret the neural signals, ML code was written and adapted from existing repositories. A RNN model was used to send read the 8 brainwaves, and associate them with one of three values: Open, Closed or None. The None value was added to reduced noise, essentially making it so that it is not always interpreting a thought, but only doing so when it recognises that thought. The computer then sends the corresponding command to the Servo motors through a serial connection with an Arduino.

Demo Video

Mobirise Website Builder

Current State of the Project

This project is currently ongoing and a work in progress. New updates are expected throughout the course of 2024.

Hardware

Robotic hand fully built and working

Software

ML code able to recognise brain signals and identify corresponding action (better than random)
Python code able to control fingers + wrist independently or concurrently

Integration

Working, although subpar accuracy in identifying signals does not allow for reliable operation of the product

Carlo Colizzi 2024 - Olin College of Engineering

‌

Drag & Drop Website Builder