About Me

I am CS PhD student in robotics & machine learning at the Robotics & Embedded Systems Lab (RESL) , advised by  Gaurav Sukhatme . ...

Sep 1, 2014

ET Robocon

During my initial years at MathWorks, I was given the opportunity to work on the official MathWorks entry to ET Robocon Tokyo! I've put videos of the work we did for Robocon 2013 and 2014.

ET Robo 2013

Balance a Lego MINDSTORMS NXT robot, while following a line and doing the following:
1) Go up a ramp
2) Jump off onto a platform. Balance on moving platform
3) Jump off the platform and go down a ramp

Here's a video of the complete trial run. My work is during 0:57 - 1:16.

ET Robo 2014

Detect and track objects in real-time. I had a camera providing a birds-eye view of the arena. I had to track two Zumobots and two spheros. Here is a video of the trial run (notice the blue sphero on the left. It is not participating in this one).

This is the view from the camera at the start of the run (with SIFT feature points marked for the zumobots)

Notice that I have added pictures on the zumobots, which help in detecting and identifying one zumobot from another.

Using these feature points, I create a bounding box for the two zumobots, and send the position of the zumobots to the zumobot controller. Here is a sample view with the bounding boxes and a sphero.

Finally, detecting and tracking the spheros was easier due to their colour. I used HSV colour space and could easily detect the spheros in the arena. Here is the output after using hue and saturation to filter out the orange sphero. Using this binary, I did blob detection to get the centroid of the sphero.

Overall, both fun projects to work on. Gained valuable experience computer vision and low-level controls for robotics!