Environment Configuration — Parallelism
Set the number of environments, choose their spacing, then choose whether they share scenery and position.
1. Set the number of environments
env.num_envs overrides env.scene.num_envs:
srb agent rand -e _manipulation env.scene.num_envs=16
srb agent zero -e _manipulation env.num_envs=128
Environments share one batched DirectEnv; cross-environment collisions are
filtered.
2. Set environment spacing
Visual sensor isolation has not been established. Spacing alone does not guarantee cameras cannot see another environment.
env.spacing, when set, supplies env.scene.env_spacing; otherwise the scene
default is used:
srb agent rand -e _manipulation env.num_envs=64 env.scene.env_spacing=20.0
srb agent rand -e _manipulation env.num_envs=32 env.spacing=10.0
If env.stack=true, spacing is forced to 0.0.
3. Choose environment stacking
env.stack controls whether environments share scenery and position:
| Value | Effect |
|---|---|
false | Independent scenes and positions; greater diversity, initialization cost, and memory use. |
true | Shared scenery and position; faster initialization and lower memory use, but less diversity. |
With particles and multiple environments, SRB forces env.stack=false and
logs a warning.
srb agent rand -e _manipulation env.num_envs=16 env.stack=true
Note: These registered task IDs default to
env.stack=true:excavation_demo,lunalab,velocity_tracking,locomotion_velocity_tracking,waypoint_navigation,locomotion_waypoint_navigation,orbital_waypoint_navigation,formation_following,formation_following_decentralized,formation_following_orbital,moonfall,traversal, andlanding. Stacking shares scenery and forces spacing to zero, so passenv.stack=falsefor visual observations or independent scenery, for example:srb agent train --algo dreamer --env waypoint_navigation \ env.num_envs=64 env.stack=false