A TinyML creative project
Low fidelity prototype

Prototype 1 + Prototype 2 + Prototype 3 = Low fidelity concept

My research work is finally paying off!

Last time, I was having a lot of difficulties with this project. The project seemed out of my league and that had an impact on my motivation and drive. Monday was the game-changer.

I still had not figured out how to import my own models onto the BLE sense microcontroller. Luckily, edge impulse was there to save the day. Finally, I could import my own models, where I had been stuck on Friday - Sunday. I changed my code to work with IC2 communication.

Conclusion: Monday was a very productive day!

And it made my motivation thrive!

Learning/Building blocks are done

I felt I had learned enough to finally make a low fidelity prototype of a small driving car that can respond to "yes". When this word is said: The servo's will start spinning and the motor will start to move forward.

This is done by these prototypes I have made in week 1:

  • Testing modded servo's spinning.
    • Value: 180 --> full speed
    • Value: 90 --> halt
    • value: 0 --> reverse
  • Serial communication between 2 Arduino's
    • Arduino BLE Sense (Brain) --> Machine learning models
    • Arduino Uno (Body) --> Component control
  • Machine learning prototype running on the BLE Sense
    • Tensorflow lite (importing existing models)
    • Edge impulse (importing my own self-trained models) <-- I chose this approach to get the best results
  • Battery circuit prototype
    • Making my own battery holder
    • Voltage theory (parallel circuits, etc)
    • Powering the Arduino boards and components with external power supplies was a difficult challenge that's why I made a prototype first to not fry any expensive components.

Other prototypes I had made that is not included YET in my low fidelity prototype

  • Testing my speaker module with the game of thrones theme
  • Ir sensors: Led activates when the Ir sensor detects input.
  • 2 dc motors linked to a Motor driver.

I put all of this together and this is the result:

Videolink (unlisted): https://www.youtube.com/watch?v=YX2flVePeAI&feature=youtu.be&fbclid=IwAR1VdQCLOxod4o31FavhFdkHZ814G1vuQ4pjYdyJ3VM9VJQDcP9ENqJ1V_U

Driving car that responds to the word "yes"

Agile working method

I have planed an agile working method that will focus on TinyML.

  • First week: Prototypes/experimenting/pathfinding/what works and what does not work?
  • The second week: applying the knowledge of my first week into a low fidelity prototype (a driving robot that can respond to a different kind of voice commands.
  • The third week: Making a polished first version of my robot (minding new opportunity areas), tech demo that showcases the machine learning of new sensor data. eg: accelerometer, camera, etc.
  • 4th week: polishing --> breathing room

My personal opinion

Monday and Tuesday have been a great motivation boost for me. This agile working method is keeping my spirits up. Sunday, I had nothing to showcase and 1 day later I have finally made something that I am proud of in this stage of development.

And I can't wait to make a V1 version of my product and showcase it in different kind of parkours/demos!