SIMULATION OF PREDATOR PREY DYNAMICS USING DEEP REINFORCEMENT LEARNING

CS 275: Artificial Life for Computer Graphics and Vision

Group Members:

1. Akshay Sharma
2. Anoosha Sagar
3. Maithili Bhide
4. Rahul Dhavalikar

Contents of this Repository:

1. index.html: web page of the project for quick understanding and navigation of the project
2. report.pdf: report of the project
3. videos: this folder contains all the videos of our simulation
4. img: this folder contains all the images used in the webpage and report
5. code: code of the project
6. code/ddqn.py: file containing the DDQN class
7. code/ddqn_run.py: main driver program of the project
8. code/multiagent/scenarios: this folder contains all the different scenarios used in our project
9. code/run: this folder contains all the shell scripts required for training and testing our project

How to Train a Scenario

To train a scenario, run any of the training shell scripts in code/run. For example, to train a simple 1v1 scenario, run the following script

./simple_1v1_train.sh

How to Test Out a Scenario

To test a scenario after training, run the corresponding testing shell script in code/run. For example, to test the simple 1v1 scenario after successfully training it, run the following script

./simple_1v1_test.sh

By default, the testing shell scripts contain the path to our pretrained models (e.g., ./save/simple_1v1_final). While testing your own models, this path should be changed to point to your model.