Skip to content

Safe Exploration for an RL agent in a dynamic wireless network environment using a constrained Markov Decision Process (CMDP).

Notifications You must be signed in to change notification settings

OluwaseyiWater/meta_learnedRL

Repository files navigation

meta_learnedRL

This is a collection of code for a meta learned reinforcement learning project. The project combines reinforcement learning with meta learning to learn how to learn reinforcement learning policies. The code is written in jax and jumanji.

The project is split into four main components:

  • mLN/environment.py: This contains the reinforcement learning environment. It is a cellular network environment in which the agent must learn to allocate resources to different users.

  • mLN/agent.py: This contains the reinforcement learning agent. It is a basic DQN agent.

  • mLN/meta_agent.py: This contains the meta learning agent. It is a meta learning agent that wraps the reinforcement learning agent.

  • mLN/train.py: This contains the training code. It trains the meta learning agent to learn how to learn reinforcement learning policies.

  • mLN/

    • environment.py
  • Installation

    • pip install --upgrade -r requirements.txt
  • Run the code

    • chmod +x scripts/*
    • ./scripts/run_maml.sh
    • ./scripts/run_recurrent_attn.sh
    • ./scripts/run_recurrent_ml.sh

About

Safe Exploration for an RL agent in a dynamic wireless network environment using a constrained Markov Decision Process (CMDP).

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •