Isaac Sim

Quick Reference – Scripts

1. Manufacturing Scene (manufacturing_scene.py)

Creates a full factory environment: ground plane, conveyor belt, shelving racks, Franka Panda robot arm, boxes on the conveyor, work table, industrial lighting.

# Basic scene creation (200 sim steps)
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh manufacturing_scene.py --steps 200

# Save as USD + capture image
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh manufacturing_scene.py \
  --save-usd /output/factory.usd \
  --capture /output/factory_scene.png \
  --steps 200

2. Pick and Place (pick_and_place.py)

Commands the Franka Panda to pick a box from the conveyor and place it at a target location. Uses Isaac Sim's built-in PickPlaceController.

# Default pick-and-place
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh pick_and_place.py

# Pick box index 2, place at custom position
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh pick_and_place.py \
  --pick-index 2 \
  --place-pos "0.5,-0.8,0.5" \
  --steps 2000 \
  --capture /output/pick_place_result.png

Arguments:

• --pick-index N – Which conveyor box to pick (0-3)
• --place-pos "x,y,z" – Target placement position
• --steps N – Max simulation steps (default: 2000)
• --capture PATH – Save final state image

3. Robot Control (robot_control.py)

General-purpose Franka Panda control with four actions:

# Return to home position
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh robot_control.py --action home

# Move to specific joint positions (7 DOF)
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh robot_control.py \
  --action move_joint \
  --joints "0.0,0.5,0.0,-1.0,0.0,1.5,0.8"

# Move end effector to Cartesian target (uses RMPFlow)
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh robot_control.py \
  --action move_to \
  --target "0.4,0.0,0.4"

# Gripper control
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh robot_control.py --action gripper --state open
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh robot_control.py --action gripper --state close

Output: Prints a STATE_JSON:{...} line with joint positions, velocities, and gripper state. Use --output /output/state.json to save to file.

4. Spawn Objects (spawn_objects.py)

Spawn primitive objects (box, cylinder, sphere) with physics.

# Single box at a position
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh spawn_objects.py \
  --type box --position "0.3,0.0,1.0" --scale "0.1,0.1,0.1"

# 10 random spheres in an area
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh spawn_objects.py \
  --type sphere --count 10 --area "-1,1,-1,1" --height 1.5

# Static colored cylinder
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh spawn_objects.py \
  --type cylinder --position "0.5,0.5,0.5" --color "1.0,0.0,0.0" --static

Arguments:

• --type box|cylinder|sphere
• --position "x,y,z" – Exact position (for single objects)
• --scale "x,y,z" – Object dimensions (default: 0.05 each)
• --color "r,g,b" – RGB color (0-1 range)
• --count N – Number of objects
• --area "xmin,xmax,ymin,ymax" – Random spawn bounds
• --height H – Spawn height for random objects
• --mass M – Object mass in kg (default: 0.2)
• --static – No physics (visual only)
• --seed N – Reproducible random placement

5. Capture Viewport (capture_viewport.py)

Render a camera image from the simulation. Can load the manufacturing scene or a custom USD file.

# Capture manufacturing scene
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh capture_viewport.py \
  --output /output/scene.png \
  --scene manufacturing

# Custom camera angle
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh capture_viewport.py \
  --output /output/closeup.png \
  --scene manufacturing \
  --eye "1.0,-1.0,1.0" \
  --target "0.0,0.0,0.5" \
  --resolution "2560,1440"

# Empty scene
/home/ubuntu/.openclaw/workspace/isaac-sim-demo/scripts/run_sim.sh capture_viewport.py \
  --output /output/empty.png --scene empty

Arguments:

• --output PATH – Image output path (required; use /output/ inside container)
• --eye "x,y,z" – Camera position
• --target "x,y,z" – Camera look-at point
• --resolution "W,H" – Image resolution (default: 1920×1080)
• --focal-length F – Lens focal length in mm (default: 24)
• --scene empty|manufacturing|custom
• --usd-path PATH – Path to custom USD file (with --scene custom)
• --warm-up-steps N – Render warm-up frames (default: 30)

Typical Workflows

Create and Capture a Manufacturing Scene

# 1. Create the scene and capture an image
run_sim.sh manufacturing_scene.py --steps 200 --capture /output/factory.png

# 2. The image is at isaac-sim-demo/output/factory.png
# Return it to the user with the read tool

Robot Pick and Place Demo

# 1. Run the pick-and-place
run_sim.sh pick_and_place.py --pick-index 0 --capture /output/result.png

# 2. Show the result image

Interactive Robot Control

# 1. Home the robot
run_sim.sh robot_control.py --action home

# 2. Move to a position
run_sim.sh robot_control.py --action move_to --target "0.4,0.0,0.4"

# 3. Close gripper
run_sim.sh robot_control.py --action gripper --state close

Output Files

All outputs (images, USD files, JSON state) go to: /home/ubuntu/.openclaw/workspace/isaac-sim-demo/output/

This directory is mounted as /output/ inside the container. To return images to the user, use the read tool on the output file path.

Intent Mapping

Notes

• All simulation is headless (no display required).
• First run of each script is slower due to shader compilation and asset loading. Subsequent runs use cached data in ~/docker/isaac-sim/cache/.
• The run_sim.sh wrapper auto-creates and removes containers. Set KEEP_CONTAINER=1 to preserve them for debugging.
• Isaac Sim uses Nucleus for asset paths. Cloud-based assets require internet access on first load.
• Each script is self-contained: it creates its own scene, runs, and exits. For persistent scenes, save as USD with --save-usd and reload with capture_viewport.py --usd-path.