Final_Presentation_RoboticsClub
Download as PPTX, PDF
1 like119 views
The document summarizes a prosthetic hand project that aims to develop a new control input for prosthetics using EEG sensors to detect brain waves. A physical prototype hand was created for testing, which uncovered areas for improvement. An EEG sensor system was also developed to measure brain activity and eye blinks for initial prototype testing. Future work includes refining the hand prototype, finalizing the EEG sensor design and electrode placements, improving the hand control code, and potentially presenting the design to end users.
Final_Presentation_RoboticsClub
- 2. Background Information/Goals Prosthetics are becoming increasingly common due to an aging population and many war veterans Currently hands and arms are controlled either by using a sling with various linkage that creates the hand movements based upon how the sling is moved at the shoulder or by finding individual nerve endings that connect to the phantom appendage Finding and rerouting nerve endings to different muscles is typical as then EMG (muscle activity) sensors can be used as control inputs Muscle activity amplifies the nerve signal, making it easier to measure against noise This method though requires extensive surgery for rerouting of nerves to different muscles The purpose of this project is to research and implement a new control input for prosthetics that utilizes EEG sensors (brain waves) A physical prototype prosthetic hand model was created for testing purposes
- 3. Physical Hand Prototype Hand created based off of open source design (Tact Open Hand) While constructing, it was found that multiple modifications were needed for the hand to suit our needs Upon construction and initial testing, it was found numerous improvements should be implemented for a more realistic prototype Passive method of reopening hand currently rubber bands are used to bring each finger open again, creating the problem of constantly applying an opening force It was found that DC motors do not like to hold their position against a force, thus this improvement is necessary for longevity of the design More secure method for attaching fingers to hand case Grippy coating on finger tips for more secure object pick ups Additional sensors Limit switches for each finger DC motors encoders work okay to provide position information, but can get lost quite easily Force feedback force sensors on finger tips to enable the control system to know how hard the hand is gripping an object
- 5. Hand Control Controlled with chipkit MX4 microcontroller (or can use most microcontrollers) Programmed in C using MPLAB X Takes feedback from each of the DC motors quadrature encoders to determine how much the motor has turned, and thus how much the finger is flexed Allows for each finger to be individually set to a specific flex Designed to take analog input and relate voltage to a finger flex for that specific finger Controls motors using H-bridges (HB5)
- 7. EEG Sensor Uses an OpenBCI 32 bit EEG amplifier board Supports 8 different electrode placements and two reference channels Allows for very low noise signal to be sent to computer (very good CMRR in instrumentation amplifier!) Electrode placement and configuration is still in progress Testing for functionality of the sensors has been done by measuring eye blinks Right: Electrode placement for initial tests Left: Blink recording and GUI interface Top: EEG Amplifier, electrodes, USB Dongle
- 8. EEG System Output In this trial, eye blinks were being measured. As can be seen three were made, represented by spikes in the signal of each electrode
- 9. What Was Learned Troubleshooting!! The hand prototype had multiple design issues that needed to be resolved 3D printing process Soldering Individuals new to soldering were taught the basics, as soldering was needed for motor connections Brain physiology Decisions needed regarding where to place the electrodes for best performance Controlling DC motors with encoders C coding for microcontroller in MPLAB X In general, this year was a huge learning experience for everyone involved
- 10. Future Work Determine all needed improvements to existing prototype hand Design and construct a refined prototype Finish calibration of EEG sensor and finalized headset design/electrode placements Improve hand control code for more reliable operations Possibly present finished design to an end user!
- 11. Mars Rover What we accomplished this year: Wrote a proposal to the Robo-ops competition Got a working drive system Created and implemented the electrical system to power the rover Created the drive control program Got the Rover Driving
- 12. Mars Rover What's in store for next year: Write another competition proposal Fine tune mechanical drive components Create and implement secondary systems such as vision and arm Add to the main program to include vision, control for secondary systems Tune driving and turning algorithm
- 13. Chess playing robotic arm
- 14. Team members: Marcus Blaisdell CS Vitaly Kubay ME William Conner Cole EE Kily Nhan Photo taken with Kinect camera we modified to use with regular USB
- 15. Objectives: Build a fully autonomous chess playing robot Increase members knowledge in robotics Vision processing Multi-joint robotic arm movement
- 16. Intention Use a pre-made kit arm as the base Last year, the design of the arm was constantly being changed This made it difficult to progress on any specific attribute as the requirements were different from week to week Using a pre-built arm allowed us to begin working on the coding part sooner We wanted to use the C++ version of OpenCV this year instead of the Python version we were using last year. The reason we wanted to work in C++ and not Python as we believed it would provide us with greater ability to achieve our goal
- 17. Progress: The first semester, Marcus was unable to spend much time working on this due to the demands of another project Vitaly took the lead on the programming portion and wrote the majority of the code for the object detection, movement detection, and shape detection Conner, Kily, and Marcus modified a Kinect camera to use regular USB with the intention of using this as the arms primary camera In the second semester, Marcus began working on the movement algorithm attempting to implement the Denavit-Hartenberg method As Marcus has not yet taken Linear Algebra, the transformation matrixes proved to be too difficult and he resorted to using basic trigonometry
- 18. Current state: There is a movement algorithm that moves one joint at a time The movement is smooth and effective The vision system exists in multiple components in various states of functionality We can detect the pieces We cannot yet determine their spatial position
Editor's Notes
- #7: Videos/ pictures here.