Desk Robot ( StayFit Assistant )

Introduction

Spending extended periods sitting at a desk can be detrimental to health. As someone who values both health and innovation, I envisioned a way to remind myself to stay hydrated and take regular breaks to stand up and walk. Combining this with my interest in creating a tiny, cute desk robot, I found the perfect opportunity with the Grove Vision AI2 module. Its efficiency make it ideal for this project.

Project Concept

My goal is to build a health alert AI robot that monitors my sitting time and prompts me to hydrate and move at regular intervals. The robot will be equipped with the Grove Vision AI module for accurate posture detection, and it will have a charming design inspired by a concept generated by DALLE.

Key Objectives

• Health Monitoring: The robot will detect prolonged sitting and remind me to stand up, walk, and hydrate.
• Aesthetic Appeal: The robot will have an adorable and engaging design to make it a delightful addition to my desk.
• Efficient Functionality: Leveraging the capabilities of the Grove Vision AI module, the robot will provide timely and accurate alerts.

Refining the Approach

The complexity involved in designing this robot from scratch led me to consider alternative approaches. I decided to start with existing designs, test the algorithm, refine it, and then proceed to the final design.

Inspiration and Approach

While searching for existing designs, I came across a Star Wars-inspired droid that caught my attention. You can find it in this link.My plan is to print this part and redesign the enclosure to accommodate the Grove Vision AI2 module, camera, battery, and charging circuit.

You can find the parts to 3d Print from this Link.The modified parts I have added along with the Project attachment.

Deploying Pre-trained Model

I utilised Sensecraft AI, which offers a variety of public AI models that are ready for testing and deployment.

Lets use Face detection model and save the images to an SD card.

And in the output options, I selected "save image to SD card" to store the processed images directly on the device.

This is working really well without a single line of code.

Now, let's train our own model.

I started training the model to recognize when I am sitting in the chair and when I am not. I captured multiple images of myself sitting at the desk and away from the desk. After collecting a sufficient dataset, I published the model to Sensecraft AI.

Now, let's put all the electronics inside the enclosure and test the model to ensure it’s working properly. I have added a USB-C connector provision below the enclosure to connect the board with the computer.

Let's print the rest of the parts and assemble them according to the instructions from this link. I used insert nuts to connect the bolts.

Disadvantages

• Since the forward part contains a lot of weight, the bot tends to topple down quite often.
• Could have added some LEDs to notify me.

Next Steps

Now, for the next steps, I need to include Seeed_Arduino_SSCMA library and code using the Arduino IDE to calculate the time I spend sitting on the chair. It should alert me to take intervals and I need to train the model to check if I am drinking water too. Additionally, I need to add a buzzer and, if possible, an audio module, along with some servo mechanisms to make it move.

Let me know your thoughts in the comments!

I will update soon. Happy hacking!