A Review of Nine Physics Engines for Reinforcement Learning Research

TL;DR

This paper reviews nine physics engines used in reinforcement learning research, comparing their features, usability, and performance to guide researchers in selecting appropriate simulation tools.

Contribution

It provides a comprehensive comparison of nine physics engines for RL, highlighting MuJoCo as the top choice and discussing challenges in engine selection and usability.

Findings

MuJoCo is the most recommended engine for performance and flexibility.

Unity is easy to use but less scalable and less accurate.

There is a need for improved usability and transparency in simulation engines.

Abstract

We present a review of popular simulation engines and frameworks used in reinforcement learning (RL) research, aiming to guide researchers in selecting tools for creating simulated physical environments for RL and training setups. It evaluates nine frameworks (Brax, Chrono, Gazebo, MuJoCo, ODE, PhysX, PyBullet, Webots, and Unity) based on their popularity, feature range, quality, usability, and RL capabilities. We highlight the challenges in selecting and utilizing physics engines for RL research, including the need for detailed comparisons and an understanding of each framework's capabilities. Key findings indicate MuJoCo as the leading framework due to its performance and flexibility, despite usability challenges. Unity is noted for its ease of use but lacks scalability and simulation fidelity. The study calls for further development to improve simulation engines' usability and performance and stresses the importance of transparency and reproducibility in RL research. This review contributes to the RL community by offering insights into the selection process for simulation engines, facilitating informed decision-making.