This project is divided into two sections: The first of is an implementation of an agent that uses the Q-Learning algorithm to maximize the average signal perception it receives in the environment. The signal is randomly generated from a Gaussian distribution, and the performance of the agent is evaluated in the end. The second part is the implementation of an agent that employes Opponent Modeling Q-Learning . This agent is evaluated in an environment that also contains a Q-Learner agent, they both start each episode at the same start location.
This demonstration program is developed in C++ programming language and the OpenCV library is used to visualize the situation.
Prof. Daniele Nardi
Requirements:
- OpenCV 2.4 or higher
Compile the source using the 'make' command, then execute the Qlearning file. you can also change the behaviour in the Qlearning.cpp file