NB Roomba




So the project is to build a robot which can carry a camera and walk around. Roomba is cheap and easy to develop on it.

This is a Nvidia Jetson tx1 board. It has a GPU along with a powerful ARM chip on it. The greatest stuff here is this board only consumes 10W power.

I already opened up a Roomba, and try to mount this board on it. If you don't know how to open it, you can read this inversely.

Step 1: Open It Up and Drill Holes

After removing the top&bottom cover, wheels, battery, bumper, motherboard. Here is what your Roomba should look like.

I drill four holes for the tx1 board, but if you got other boards, make sure you measure the right length before drilling anything.

Step 2: Put Motherboard Back and Connect Pins

Step 3: Put Front Pad and Screws Back

Step 4: Connect Bumper Wire, Put Bumper Back

Step 5: Put Top Cover and Screws Back

Step 6: Put Wheels, Battery, and Back Cover Back to Its Place

Step 7: Mark Useless Part and Cut It Away

I found out that the Tx1 board is actually larger than the circle shape of the Roomba after a while. So I decided to cut the dust bin open and live more space to the board.

The red line on the bin indicate the cutting line.

Step 8: Prepare Power/Reset Button

As the power/reset button on the board is hidden inside the Roomba, I soldered wires to buttons and later mount it on the dustbin.

I should choose two different color to distinguish Power and Reset... 0_o

Step 9: Put All Component Together

Step 10: Add a Bottom Cover to Keep the Board Away From Dust

Step 11: Now Write a Simple Face Detection Program on That. It Will Find and Follow the Largest Face LOL



    • Weaving Challenge

      Weaving Challenge
    • Epilog X Contest

      Epilog X Contest
    • Organization Contest

      Organization Contest

    2 Discussions


    Question 9 months ago on Introduction

    Great Build. Can you provide the software for face detection and follow program please?

    1 answer

    Answer 9 months ago

    Hi, if you familiar with OpenCV, it won't be difficult to imply its face detection algorithm. If I remember correctly, it's called haar cascade. The follow program is simple. You just first detect the face coordinate, if it's left on the picture just program the robot turn to same direction.