Principle:Google deepmind Mujoco Sleep Wake Optimization
| Knowledge Sources | Domains | Last Updated |
|---|---|---|
| Google DeepMind MuJoCo | Performance Optimization | 2025-02-15 |
Overview
Description: Sleep/wake optimization strategy that disables computation for bodies that have come to rest.
Context: In scenes with many bodies where only a subset are actively moving, MuJoCo's sleep system can significantly reduce computational cost by skipping dynamics computations for inactive bodies.
Theoretical Basis
The sleep optimization monitors the kinetic energy and velocity of each body over a configurable time window. When a body's activity falls below a threshold for a sustained period, it is marked as sleeping. Sleeping bodies are excluded from the forward dynamics pipeline, constraint solving, and contact detection. When an external force or contact from an awake body disturbs a sleeping body, it is woken and re-enters the active computation. This optimization is particularly effective for stacking and settling scenarios.
Related Pages
Implementations
Workflows
- (none yet)