arif2705 / mazerunner

Final Project for Artificial Intelligence 2016-1 class at UNICAMP.

Maze Runner

Final Project for Artificial Intelligence 2016-1 class at UNICAMP.

Local navigation of the bibed robot NAO on a V-REP simulation using reinforcement learning (Q-Learning). The desired behavior is for the robot to use its sensors to avoid walls and attempt to get as close as possible to a tag placed at the kitchen.

Installing

This project has some dependencies that are not managed by pip. They must, therefore, be installed manually:

• V-REP simulator
• Choregraphe Suite
• Aldebaran NAO Python SDK

Finally, install MazeRunner:

python setup.py install --user

Executing Navigation with Q-Learning

1. First, start Choregraph:

   /opt/Aldebaran\ Robotics/Choregraphe\ Suite\ 2.1/bin/naoqi-bin -p 5000

2. Open the V-REP simulator:

   /path/to/vrep/vrep.sh

   Then open one of the scenes in mazerunner/scenes folder and run it.

3. You have two options now:

4. 1. Train a navigator by executing python mazerunner/examples/train_navigator.py. This script will move the agent and train it to achieve the goal set (kitchen, by default).
5. 2. Navigate using a previously-learned behavior by executing python mazerunner/examples/navigate_with_trained_agent.py