Cartpole-RL: Solving the Cartpole Problem with RL Algorithms

This repository contains implementations of various Reinforcement Learning (RL) algorithms to solve the CartPole-v1 environment from OpenAI Gym. The primary method used here is the Cross-Entropy Method (CEM).

---

Getting Started

Prerequisites

Ensure you have the following installed before running the scripts:

• Python 3.x
• OpenAI Gym
• PyTorch

You can install the requirements for the repo using the following command:

pip install -r requirements.txt

To run the training script,

python3 cartpole_crossentropy.py

To run the test script,

python3 cartpole_crossentropy_test.py

The logs during training appear like the following:

Running on mps
1: loss=0.662, reward_mean=27.5
2: loss=0.646, reward_mean=44.9
3: loss=0.626, reward_mean=61.5
4: loss=0.602, reward_mean=82.3
5: loss=0.597, reward_mean=57.9
6: loss=0.573, reward_mean=79.5
7: loss=0.573, reward_mean=67.5

The loss and reward graphs for crossentropy model are as following: