[Submitted on 12 May 2026 (v1), last revised 16 Oct 2026 (this version, v2)]
Abstract:We present OrbiSim, a novel robotic simulation paradigm that redefines world models as a fully differentiable physics engine for embodied intelligence. Unlike prior world models that focus on unconstrained imagination in latent or visual domains, OrbiSim establishes a unified, physically-grounded pathway that bridges structured scene assets, neural dynamics, and downstream reinforcement learning. By enabling end-to-end differentiability throughout the entire simulation loop — spanning from explicit state transitions to visual observation generation — OrbiSim supports tasks traditionally intractable for classical simulators, such as differentiable contact modeling, gradient-based policy optimization under sparse rewards, and intuitive physical inference. Empirical results demonstrate that OrbiSim significantly outperforms state-of-the-art world models in both predictive fidelity and control performance. Furthermore, its consistent responsiveness to asset configurations and physical parameters suggests its potential as a differentiable tool for enhancing robot simulation and policy training.
Submission history
From: Jiajian Li [view email]
[v1] Tue, 12 May 2026 13:43:53 UTC (13,721 KB)
[v2] Fri, 16 Oct 2026 15:22:03 UTC (15,953 KB)
