Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Environment Configuration — Robot

Tip: List registered robots and categories with srb ls --live robot; list cached names with srb ls robot. Use supported _ categories in -e.

Robot Selection

srb agent rand -e _manipulation env.robot=ur10

srb agent rand -e _ground env.robot=unitree_go1

srb agent rand -e _aerial env.robot=crazyflie

Select a robot with env.robot=<name>. This replaces task values and resets the robot’s owned fields, end effector, and payload. Kinova on Franka gets its gripper; Husky gets its default payload.

  • Attach a task tool with env.robot=ur10+<tool>; selecting a robot drops task tools.
  • Set env.robot.end_effector=... or env.robot.payload=... in any CLI order. Set a single field with an override such as env.robot.end_effector.<field>=....
  • Set or clear asset fields explicitly when changing selectors. srb agent eval and --continue replay embodiment by reapplying .hydra/overrides.yaml, but values supplied in --cfg YAML can remain through a selector change.
  • Selecting a named top-level asset, such as env.sample=moon_rock, resets it too. Selecting env.robot.mobile_base=... or .manipulator=... preserves that component’s sub-assets.

End Effector for Manipulators

Select an end effector with base+end_effector:

srb agent rand -e _manipulation env.robot=ur10+shadow_hand

Payload for Mobile Robots

Select a payload with base+payload:

srb agent rand -e _ground env.robot=spot+cargo_bay

Payloads can be static rigid geometry or ActivePayload articulations. Active payloads mount as separate scene articulations and add actions after the base. For landscaping, the action order is two filtered RaphRover drive channels, then one continuous RaphShovel velocity channel.

srb agent rand -e terrain_landscaping env.robot=raph_rover+raph_shovel

Landscaping maps normalized [-1, 1] actions to robot/cmd_vel slice [0:2] (linear scale and saturation 0.4 m/s; angular scale and saturation 1.0 rad/s) and payload/joint_vel slice [2:3] (scale and saturation 0.04 m/s). These are simulator limits, not hardware-calibrated. See Terrain Landscaping Training.

Asset-instance CLI customizations. Each srb agent or srb real_agent invocation rebuilds task config through Hydra.

  • extract_defaults_from_class serializes assets as {"name": ...}, with action_mode, nested assets, and all serializable plain fields (numbers, strings, booleans, or lists, tuples, and dicts thereof), including class defaults. Serializable config fields pass through; nonserializable manager-term configs are skipped.
  • AssetResolver calls asset constructors with declared values for model_post_init and the spawner. It also sends overrides such as env.scenery.flat_area_size=3.0 to simulation.

Combine Mobile Manipulators

In mobile-manipulation environments, set env.robot.mobile_base and env.robot.manipulator:

srb agent rand -e _ground_manipulation env.robot.mobile_base=unitree_a1 env.robot.manipulator=unitree_z1

srb agent rand -e _ground_manipulation env.robot.mobile_base=anymal_d+cargo_bay env.robot.manipulator=ur10+shadow_hand

Compose a mobile manipulator in one env.robot token. Use a registered mobile robot, manipulator, and optional end effector. Tasks whose default robot is not a composite mobile manipulator accept only ground mobile bases here:

srb agent rand -e _ground_manipulation env.robot=unitree_a1+unitree_z1
srb agent rand -e _ground_manipulation env.robot=husky+ur10+robotiq_hand_e

Change Action Modes

Set action_mode with a Hydra override on the robot, mobile base, manipulator, or tool. Each robot, active payload, and active tool adds its action_mode to the action space. This works for registered tasks whose composed robot exposes the field:

srb agent rand -e _manipulation env.robot.action_mode=joint_pos_rel

srb agent rand -e _manipulation env.robot=ur10 env.robot.action_mode=joint_pos_rel