Compare
Simulators
Compare robotics and physical AI simulators: license, physics engine, GPU acceleration, ROS support.
To understand what these simulators are for, our world models and sim2real pillar explains the data factory they feed.
Geographic footprint
| Countries | entries |
|---|---|
| United States | 5 |
| Switzerland | 2 |
| United Kingdom | 1 |
Hover a country for details, click to filter the table.
Click any row to reveal details, sources and verification date.
| Name | Organization | License | Open source | Physics engine | GPU accel. | ROS |
|---|---|---|---|---|---|---|
| NVIDIA Isaac Sim + Isaac Lab | NVIDIA | Apache-2.0 (composants Omniverse Kit propriétaires) | yes | PhysX 5 (+ moteur Newton intégré dans Isaac Lab) | yes | yes |
|
Isaac Sim 5.0 and Isaac Lab 2.2 reached general availability at SIGGRAPH 2025, with the Isaac Sim application open-sourced on GitHub under Apache-2.0 (the underlying Omniverse Kit remains closed source). It is the de facto industry platform for humanoid and manipulation policy training, now compatible with the Newton physics engine.
Uses : RL training, synthetic data, sim2real, digital twin
Shifts
Sources: NVIDIA Technical Blog (2025-08-11)GitHub (isaac-sim) (2026-07-09)NVIDIA Developer (2026-07-09)NVIDIA Technical Blog (2025-06-16) Last verified 2026-10-07 |
||||||
| MuJoCo | Google DeepMind | Apache-2.0 | yes | MuJoCo (moteur natif, contacts articulés) | yes | n/a |
|
Open-sourced by DeepMind in 2022, MuJoCo remains the most cited physics engine in robot learning research; the 3.x series added MJX for accelerator hardware, and MuJoCo Warp (GPU-optimized, up to ~70x faster on humanoids) also powers a solver inside NVIDIA's Newton engine.
Uses : RL training, sim2real, model-based control, benchmarking
Shifts
Sources: GitHub (google-deepmind) (2026-07-09)GitHub (google-deepmind) (2026-07-09)GitHub (google-deepmind) (2026-07-09)Google DeepMind (official blog) (2021-10-18)Google DeepMind (official blog) (2022-05-23) Last verified 2026-10-07 |
||||||
| Genesis | Genesis AI (issu d'un projet académique multi-universités) | Apache-2.0 | yes | Moteur multi-physique unifié maison (rigide, MPM, SPH, FEM, PBD) | yes | n/a |
|
Launched in December 2024 as an academic collaboration, Genesis became one of the fastest-adopted robotics simulators ever and its development is now backed by the startup Genesis AI (repackaged as genesis-world). It loads MJCF, URDF and mesh formats and targets the full spectrum from rigid robots to soft-body and fluid simulation.
Uses : RL training, synthetic data, soft robotics, generative simulation
Shifts
Sources: GitHub (Genesis-Embodied-AI) (2026-07-09)GitHub (Genesis-Embodied-AI) (2026-07-09)GitHub (Genesis-Embodied-AI) (2026-07-09)Genesis-Embodied-AI (depot official) (2024-12-18) Last verified 2026-10-07 |
||||||
| Gazebo (nouvelle génération, ex-Ignition) | Open Source Robotics Foundation (Open Robotics) | Apache-2.0 | yes | Pluggable : DART (défaut), Bullet, TPE | no | yes |
|
Gazebo Jetty, the 10th major release, shipped on September 30, 2025 as an LTS supported until September 2030. Physics runs on CPU (GPU is used for rendering and sensors), which keeps it more suited to systems integration and testing than to massively parallel RL training.
Uses : CI testing, ROS development, digital twin, multi-robot simulation
Shifts
Sources: Open Robotics Discourse (2025-09-30)GitHub (gazebosim) (2026-07-09)gazebosim.org (2026-07-09)Open Robotics Discourse (equipe Gazebo) (2024-12-02) Last verified 2026-10-08 |
||||||
| PyBullet / Bullet Physics | Communauté Bullet Physics (Erwin Coumans) | zlib | yes | Bullet | no | n/a |
|
Development has slowed to maintenance mode (latest release 3.2.7, January 2025) and most new RL research has migrated to GPU-native simulators, but PyBullet remains widely used in education and as a lightweight CPU baseline.
Uses : RL training, prototyping, education, benchmarking
Sources: PyPI (2025-01-30)GitHub (bulletphysics) (2026-07-09)pybullet.org (2026-07-09) Last verified 2026-10-08 |
||||||
| Webots | Cyberbotics | Apache-2.0 | yes | ODE (fork Open Dynamics Engine) | no | yes |
|
Born at EPFL in 1996 and open-sourced in December 2018, Webots is the leading European (EMEA) robot simulator; the latest stable release R2025a (February 2025) added new robot models and improved ROS 2 support, with nightly builds still active in 2026.
Uses : education, prototyping, ROS development, competitions
Shifts
Sources: GitHub (cyberbotics) (2025-02-04)GitHub (cyberbotics) (2026-07-09)Cyberbotics (2026-07-09)Cyberbotics (depot official, release taguee) (2018-12-18) Last verified 2026-10-08 |
||||||
| CoppeliaSim (ex-V-REP) | Coppelia Robotics | Propriétaire (Edu gratuite pour l'enseignement, Pro commerciale) | no | Multiples au choix : Bullet, ODE, Newton Dynamics, Vortex, MuJoCo | no | yes |
|
Successor to V-REP, developed in Switzerland; version 4.10 shipped in May 2025. The Edu edition is free but restricted to schools and universities, while commercial use requires a CoppeliaSim Pro license.
Uses : prototyping, education, factory simulation, algorithm validation
Sources: Coppelia Robotics (2026-07-09)Coppelia Robotics (2025-05-14)Coppelia Robotics (2026-07-09) Last verified 2026-10-08 |
||||||
| Newton | NVIDIA + Google DeepMind + Disney Research (gouvernance Linux Foundation) | Apache-2.0 | yes | Moteur propre construit sur NVIDIA Warp (solveurs incluant MuJoCo Warp et VBD déformables) | yes | n/a |
|
Announced at GTC in March 2025 and released in open beta at CoRL 2025 under Linux Foundation governance, Newton is the major new entrant of 2025-2026: it plugs into Isaac Lab and MuJoCo Playground, and is already used by Skild AI and Samsung for fine manipulation and deformable-object training.
Uses : RL training, sim2real, differentiable physics, deformable simulation
Sources: NVIDIA Technical Blog (2025-03-18)NVIDIA Newsroom (2025-09-29)GitHub (newton-physics) (2026-07-09) Last verified 2026-10-04 |
||||||
8 / 8 entries
Christian Verbrugge · D·Fairy
Bringing a physical AI solution into European industry?
15 years at KUKA on automotive OEM programs (€100M), now focused on edge AI and autonomous agents. I support robotics and embodied AI companies on their EMEA market entry: OEM and Tier 1 access, co-funded pilots, AI Act readiness.
Book an intro call