Instruction Graphs

This is a 15-400 research project based off the work on instruction graphs described in https://www.cs.cmu.edu/~mmv/papers/14aamas-MericliEtAl.pdf.

Dependencies

• python2
• ply (installable via pip install ply)
• ros-indigo-turtlebot-simulator (try sudo apt-get install ros-indigo-turtlebot-simulator on Ubuntu)

Running the interpreter

• git clone https://github.com/anbenson/instructiongraphs
• cd instructiongraphs/IGinterpreter
• roslaunch turtlebot_stage turtlebot_in_stage.launch &
• ./main.py getcoffee.ig

Notes about Implementation

• The interpreter by default requires a running instance of the turtlebot simulator, which is very slow and will cause your computer to become slow to respond. You have been warned. Alternatively, you could edit the code pretty easily in dynamics.py to disable calls to the simulator and merely print things instead.
• The move command takes two arguments. The first refers to distance, and the second to angular rotation.
• In order to detect obstacles, the interpreter relies on user input. Enter 'y', 'yes', or an empty string to respond with 'yes', else 'no'.

Demo

A short video demonstrating the 'getcoffee.ig' program in action is included in the file ig_turtlebot_demo.mp4. You might wonder why I filmed it with a phone camera. This was mostly due to the fact that the turtlebot simulator bogged down my VM so badly that I was worried what a screencast program might do to it.

Report

The final report for 15-400 is in the report subdirectory. The paper itself is paper.tex, but the report also contains several appendices which are the other tex files in the directory.