Principle:Google deepmind Mujoco Sleep System
| Knowledge Sources | Domains | Last Updated |
|---|---|---|
| Google DeepMind MuJoCo | Physics Simulation, Performance | 2025-02-15 |
Overview
Description: Body sleep/wake state management system that tracks activity levels and transitions bodies between active and dormant states.
Context: MuJoCo's sleep system manages per-body and per-island activation states, ensuring that groups of connected bodies (islands) sleep and wake together to maintain physical consistency.
Theoretical Basis
The sleep system operates at the constraint island level. An island is a connected component of bodies linked through active contacts or joints. If all bodies in an island have velocities below the sleep threshold for a duration exceeding the sleep time, the entire island is put to sleep. Waking occurs when any body in a sleeping island receives a non-negligible external force, control input, or contact from an awake body. Island-level sleep management prevents artifacts such as a sleeping body failing to respond to a contact from another body in the same island.
Related Pages
Implementations
Workflows
- (none yet)