Kamino: GPU-based Massively Parallel Simulation of Multi-Body Systems with Challenging Topologies

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 practitioners; approximating the system 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 coupling. This capability facilitates high-throughput parallelized simulations that capture the true nature of mechanical systems that exploit closed kinematic chains for mechanical advantage. 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 that 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 can resolve contact dynamics without resorting to approximate models that simplify and/or convexify the problem. We demonstrate RL policy training on DR Legs, a biped with six nested kinematic loops, generating a feasible walking policy while simulating 4096 parallel environments on a single GPU.

Figures (9)

• Figure 1: DR Legs (left) with kinematic graph overlaid for half of the left leg (middle). As visible in the full kinematic graph (right), each leg contains a kinematic loop, with an addition loop created between the halves of each leg. Kamino models each rigid body with an independent pose and enforces kinematic relationships, including loop closures, as explicit algebraic constraints.
• Figure 2: Kamino is a solver within Newton, which defines a common interface and core building blocks for multiple solver back-ends, all built atop of NVIDIA Warp warp2022. For reinforcement learning workflows, Kamino will make use of the integration with NVIDIA's framework for robot learning, Isaac Lab, in the near future. Isaac Lab will provide environment wrappers, task definitions, and training pipelines on top of Newton's physics back-ends, enabling end-to-end RL training with Kamino.
• Figure 3: Kamino supports heterogeneous worlds. Each world may contain a different robot or set of robots with a variable amount of bodies and joints.
• Figure 4: Robotic systems simulated with Kamino. DR Legs and Olaf contain closed kinematic chains; BDX and Iron Man are tree-structured.
• Figure 5: Memory usage for sparse vs dense solvers. Per-world memory usage in Mb for each problem and solver, against the problem size.