This week I'm wrapping up, getting packed & ready to leave Providence, RI. I had an invaluable learning experience here at Brown University. I am very grateful for the DREU giving me this opportunity to intern here with the Brown Robotics Lab team. It was very inspiring to hear the logic and thought process of PhD's and graduate students as they helped me with my difficulties and project's shortcomings. My mentor Dr. Jenkins is an awesome individual that I will definitely be keeping in contact with and probably be asking for a letter of recommendation in the near future. I really wanna thank CRA & DREU for having such a wonderful program for undergraduates. I couldn't have spent my summer better. I got to see a new state & be educated at the same time.

In these two weeks I successfully got the irobot to recognize the tag & move forward toward it and stop & to also turn and face it if it is not directly in its view, basically getting the irbot to follow the tag. I also got the irobot to tackle the tag when close enough. This was done using the nolan.py program in the Brown ROS package and making some slight changes to it.

Problems experienced with accomplishing this were:

1. The webcam on the mini notebook that I used had a bad resolution correction system, so I had to hard code a desired resolution that would work better with the ar_recog.

export GSCAM_CONFIG="v4l2src device=/dev/video0 ! video/x-raw-rgb,width=320,height=240 ! ffmpegcolorspace ! identity name=ros ! fakesink"

2. The main goal initially was to have the tag attached to the football and have the irobot use the tag to seek out & tackle the/kick the ball but I found difficulty in devising a means of doing this without the tag toppling over and the webcam losing sight of the tag. So instead simply attached the tag to a box to display the irobot working with the tag as though the box was a ball. the box substitute being a little better in this case seeing it would never topple over.

Below are videos I recorded showing the irobot performing a find & tackle and video showing that the irobot is programmed to follow the tag if it were to be moving.

This week was spent trying to figure out different ways to have the iRobot successfully seek out and find a ball & play soccer with it. Referring to the post docs & graduate students for advice, it seems that this is easiest done by setting up a camera system on the robot that will then in turn located tags using the ROS ar_recog package, to locate an image/tags over the ball. these tags are fond via corner-finding. The ar_recog package looks for an object with 4 corners much like a square with significant contrast between the two colors balck & white. Once this is found, the ar_recog package outlines it in green reporting metric values of distance & position, which can be then used to add the iRobot in using position tracking to locate the iRobot & evidently move towards it.

Below are examples of AR tags:



Now using this method of AR Tag Recognition, I can now write a program that will have the iRobot locate and follow the tag which will be mounted on the ball & hopefully have the iRobot play soccer with the ball.


 Today I also got the AR Tag Recognition up & running. Below you can see it finding the AR Tag held in front of it & outlining it in  a green box.

This week I spent time continuing the odometer calibration of the iRobot Create and also getting myself familiarized with writing programs using the ROS system to control the robot's motor skills. I had to calculate the Average Drift per Theta for the robot's forward motion & also the rotaional motion.

Below are pictures of me doing the measurements of the Average Drift per Theta:





This calculations were based on me having the iRobot run experimental forward motions 10+ times for several different distances.

For the rotational measurements a constant theta turn of 360 degrees was used & this was tested 10+ times as well.

This week's assignment involved me having to use the ROS software to design different exercises for the Create iRobot carry out to test it's odometer. For example, attempting to answer questions such as: "Does the iRobot really go as far as its position tracker sensor says it does? and if not, How far does it really go and at what angle if any?". With this information we can design a fail safe calibrated to compensate for this odometer error to make for more accurate & precise movement of the iRobot in the x & y plane as it navigates to different coordinates.

This week I successfully got the iRobot carry out, what is called an Enclosure Escape maneuver where I program the iRobot to be able to find its way out of a maze of enclosure using the bumper sensors located on the front of the iRobot. When the iRobot encounters a wall or obstacle it intelligently moves away and tries another direction until finally escaping its enclosure.