Principle:Google deepmind Mujoco Mass Matrix
| Knowledge Sources | Domains | Last Updated |
|---|---|---|
| Google DeepMind MuJoCo | Physics Simulation, Linear Algebra | 2025-02-15 |
Overview
Description: Computation and factorization of the joint-space mass (inertia) matrix for multibody systems.
Context: The mass matrix M(q) is central to MuJoCo's dynamics computations. It appears in the equations of motion M*a = f and is used in both forward dynamics (computing accelerations from forces) and inverse dynamics (computing forces from accelerations).
Theoretical Basis
The mass matrix is a symmetric positive-definite matrix whose (i,j) entry represents the inertial coupling between generalized coordinates i and j. MuJoCo computes it using the composite rigid body algorithm and stores it in a sparse format that exploits the banded structure arising from the kinematic tree topology. For forward dynamics, the mass matrix is factorized using sparse Cholesky decomposition (L*D*L^T), enabling efficient solution of the linear system. The factorization is also used for computing constraint-space quantities and analytical derivatives.
Related Pages
Implementations
Workflows
- (none yet)