This project is part of an IT project unit of the ANDROIDE master's program. Its purpose is to train us in carrying out a team project from start to finish, from analyzing the subject to software development, as well as experimenting with the work done. Our project topic consists in implementing and evaluating a reinforcement learning algorithm, TD-MPC, within a specific library, BBRL.
TD-MPC is a combination of Temporal Difference (TD) learning and Model Predictive Control (MPC). This approach utilizes Temporal Difference predictions to enhance planning and action execution in Model Predictive Control, enabling better anticipation and adaptation to environmental changes for decision optimization. BBRL, which stands for BlackBoard Reinforcement Learning, is a simple and flexible library for reinforcement learning, derived from SaLinA.
BBRL: Inspired by SaLinA, source code: https://github.com/osigaud/bbrl Gymnasium: OpenAI library, source code: https://github.com/Farama-Foundation/Gymnasium
Our project is inspired by the research conducted by Hansen et al. See the source code below.: https://github.com/nicklashansen/tdmpc
The project is available on GitHub.
Before continuing, you need to install Python3.
Then, follow these steps in a terminal to install the project:
git clone https://github.com/ElDjeee/Codage_TD-MPC.git
cd Codage_TD-MPC
python3 -m pip install -r requirements.txtOnce this is done, you can run the project with this command:
python3 src/main.py