This repository provides a PyTorch implementation of the Deep Deterministic Policy Gradient (DDPG) algorithm, a reinforcement learning method for continuous action spaces. It is suitable for researchers and practitioners looking to experiment with or apply DDPG in environments like Pendulum-v0 and MountainCarContinuous-v0.

How It Works

The implementation follows the DDPG algorithm, utilizing an actor-critic architecture. It employs a replay buffer for experience replay and a random process for exploration, with utility functions adapted from the keras-rl repository. This approach allows for stable learning in continuous control tasks by decoupling the gradient updates.

Quick Start & Requirements

• Primary install / run command: ./main.py
• Non-default prerequisites and dependencies: Python 3.4, PyTorch 0.1.9, OpenAI Gym.
• Links: DDPG Paper, PyTorch, OpenAI Gym, keras-rl

Highlighted Details

• Implements Deep Deterministic Policy Gradient (DDPG).
• Includes utility functions for replay buffer and random process from keras-rl.
• Demonstrates training on Pendulum-v0 and MountainCarContinuous-v0 environments.

Maintenance & Community

• Contributions are welcome.

Licensing & Compatibility

• License: Not specified in the README.
• Compatibility: PyTorch 0.1.9 is a very old version, potentially incompatible with modern PyTorch releases and Python versions.

Limitations & Caveats

The project relies on a significantly outdated version of PyTorch (0.1.9), which is likely incompatible with current Python versions and may lack features or performance improvements found in modern PyTorch. The README also does not specify a license, creating uncertainty for commercial use or integration into other projects.