Kamino: GPU-based Simulation of Rigid Multi-Body Physics with Kinematic Loops
Authors: Vassilios Tsounis, Guirec Maloisel, Christian Schumacher, Ruben Grandia, Agon Serifi, David Müller, Chris Amevor, Tobias Widmer, Moritz Bächer · CoRL 2026
Licences
- Open Access
- Paper: arXiv.org perpetual, non-exclusive license 1.0
Under arXiv's licence, everything below is written from the abstract only.
Why it matters
Robots with strongly coupled constraints such as kinematic loops are usually approximated as kinematic trees with extra loop-closure constraints or mimic joints. Kamino supports these couplings natively and simulates structurally different robots in one batch; the authors train a walking policy for a biped with six nested kinematic loops using 4,096 parallel environments on a single GPU.
Proposed solution
Kamino computes constraint forces with an ADMM-based constrained optimization that solves rigid multi-body forward dynamics as a nonlinear complementarity problem, resolving contact without simplified or convexified models, and runs batched heterogeneous worlds on the GPU.
Abstract
We present Kamino, a GPU-based physics solver for massively parallel simulations of heterogeneous highly-coupled mechanical systems. Implemented in Python using NVIDIA Warp and integrated into the Newton framework, it enables the application of data-driven methods, such as large-scale reinforcement learning, to complex robotic systems that exhibit strongly coupled kinematic and dynamic constraints such as kinematic loops. The latter are often circumvented by approximating the system's topology as a kinematic tree, and incorporating explicit loop-closure constraints or so-called mimic joints. Kamino aims at alleviating this burden by natively supporting these types of couplings. Moreover, Kamino supports heterogeneous worlds, allowing for batched simulation of structurally diverse robots on a single GPU. At its core lies a state-of-the-art constrained optimization algorithm based on ADMM, which computes constraint forces by solving the constrained rigid multi-body forward dynamics transcribed as a nonlinear complementarity problem. This leads to high-fidelity simulations that resolve contact dynamics without resorting to approximate models that simplify or convexify the problem. We demonstrate RL policy training on a biped with six nested kinematic loops, generating a feasible walking policy while simulating 4,096 parallel environments on a single GPU.
Links
- arXiv 2603.16536: https://arxiv.org/abs/2603.16536
- PDF: https://arxiv.org/pdf/2603.16536
- Code: https://github.com/newton-physics/newton/tree/main/newton/_src/solvers/kamino
- Project page: https://kamino.disneyresearch.com/