Parallel Reinforcement Learning

Parallel reinforcement learning (RL) aims to accelerate the training of RL agents by distributing computations across multiple processors or machines. Current research focuses on developing efficient frameworks that minimize communication overhead and maximize hardware utilization, employing techniques like asynchronous parallelization and specialized architectures such as the reactor model. This approach significantly reduces training time for complex tasks, such as robot control and multi-agent scenarios, enabling faster experimentation and the development of more sophisticated RL agents for robotics, gaming, and other applications. The resulting speed improvements are crucial for tackling the sample inefficiency inherent in many RL algorithms.

Papers

September 12, 2024

Learning Skateboarding for Humanoid Robots through Massively Parallel Reinforcement Learning
William Thibault, Vidyasagar Rajendran, William Melek, Katja Mombaur
Reinforcement Learning LeArning Abstract Humanoid Robot Human Like Locomotion Behavior Locomotion Policy Universal Robot 5 Parallel Reinforcement Learning

May 19, 2024

Going into Orbit: Massively Parallelizing Episodic Reinforcement Learning
Jan Oberst, Johann Bonneau
Deep Reinforcement Learning Reinforcement Learning Benchmark Orbital Motion Reinforcement Learning Library Parallel Reinforcement Learning

December 11, 2023

Spreeze: High-Throughput Parallel Reinforcement Learning Framework
Jing Hou, Guang Chen, Ruiqi Zhang, Zhijun Li, Shangding Gu, Changjun Jiang
Reinforcement Learning Reinforcement Learning Algorithm Parallel Reinforcement Learning

December 7, 2023

Efficient Parallel Reinforcement Learning Framework using the Reactor Model
Jacky Kwok, Marten Lohstroh, Edward A. Lee
Reinforcement Learning Multi Agent Reinforcement Learning Fast Simulation Nuclear Reactor Parallel Reinforcement Learning

November 16, 2023

JaxMARL: Multi-Agent RL Environments and Algorithms in JAX
Alexander Rutherford, Benjamin Ellis, Matteo Gallici, Jonathan Cook, Andrei Lupu, Gardar Ingvarsson, Timon Willi, Ravi Hammond, Akbir Khan, Christian Schroeder de Witt, Alexandra Souly, Saptarashmi Bandyopadhyay, Mikayel Samvelyan, Minqi Jiang, Robert Tjarko Lange, Shimon Whiteson, Bruno Lacerda, Nick Hawes, Tim Rocktaschel, Chris Lu, Jakob Nicolaus Foerster
Multi Agent Reinforcement Learning Multi Agent Challenge State of the Art JAX Parallel Reinforcement Learning

September 22, 2022

Parallel Reinforcement Learning Simulation for Visual Quadrotor Navigation
Jack Saunders, Sajad Saeedi, Wenbin Li
Reinforcement Learning Non Humanoid Robot Simulation Framework Real World Experiment Quadrotor Navigation Parallel Reinforcement Learning