Write the room. Then work it.

Real Python, in your browser, against the same simulator the studio runs. One cell builds a room out of a sentence, the next reads what is in it, the next sets the robot to work — and the room stays as the last cell left it, robot and all.

Unitree H1

What you can write

room.build(sentence) · room.add(kind, x, y) · room.empty(w, d, h) · room.objects(kind) · room.nearest(kind) · room.corners() · room.width · room.depth · room.entrance

robot.move_to(thing) · robot.pick(thing) · robot.place(x, y) · robot.bring(thing, place) · robot.turn(deg) · robot.say(text) · robot.takeoff(h) · robot.gesture(name) · use_robot(id) · robots() · kinds()

The first run fetches Python itself — about 6 MB, once, then it is cached. Everything after that is instant, and nothing you write leaves the browser.

A kinematic preview: routes, reach and carrying are rule-based, and nothing falls or collides with force.

Cell 1

# A room is one sentence. The layout engine reads it, the same one the chat uses.
room.build("Warehouse 16x10m with 4 storage racks, 6 crates and 2 barrels")
room

# A room is one sentence. The layout engine reads it, the same one the chat uses.
room.build("Warehouse 16x10m with 4 storage racks, 6 crates and 2 barrels")
room

Cell 2

# What is in it, and what the robot makes of it.
crates = room.objects("crate")
print(len(crates), "crates;", robot.name, "can carry", robot.can_grip)
nearest = room.nearest("crate")
print("nearest is", nearest.name, "at", nearest.position, "-", robot.distance_to(nearest), "m away")

# What is in it, and what the robot makes of it.
crates = room.objects("crate")
print(len(crates), "crates;", robot.name, "can carry", robot.can_grip)
nearest = room.nearest("crate")
print("nearest is", nearest.name, "at", nearest.position, "-", robot.distance_to(nearest), "m away")

Cell 3

# Now make it act. Every route is planned around the furniture; anything impossible is refused.
rack = room.nearest("Storage rack")
for crate in room.objects("crate")[:3]:
    robot.bring(crate, rack)
robot.move_to(room.entrance)
robot.say("done")

# Now make it act. Every route is planned around the furniture; anything impossible is refused.
rack = room.nearest("Storage rack")
for crate in room.objects("crate")[:3]:
    robot.bring(crate, rack)
robot.move_to(room.entrance)
robot.say("done")

Cell 4

# Change the cast, keep the room. Try "go2", "spot", "x2", "panda", "shadow-hand".
use_robot("go2")
for corner in room.corners():
    robot.move_to(corner)

# Change the cast, keep the room. Try "go2", "spot", "x2", "panda", "shadow-hand".
use_robot("go2")
for corner in room.corners():
    robot.move_to(corner)

Cell 5

# What it can see from there. Being in the room is not the same as being in sight:
# a crate behind a 2.4 m rack is hidden, and room.visible() leaves it out.
print(robot.name, "looks from", robot.eye_height, "m")
print(len(room.objects("crate")), "crates in the room,", len(room.visible("crate")), "in sight")
for hit in robot.scan(rays=8):
    met = hit.object.name if hit.object else ("wall" if hit.wall else "nothing")
    print(f"{hit.angle:>6.1f} deg  {hit.range:>5.2f} m  {met}")

# What it can see from there. Being in the room is not the same as being in sight:
# a crate behind a 2.4 m rack is hidden, and room.visible() leaves it out.
print(robot.name, "looks from", robot.eye_height, "m")
print(len(room.objects("crate")), "crates in the room,", len(room.visible("crate")), "in sight")

for hit in robot.scan(rays=8):
    met = hit.object.name if hit.object else ("wall" if hit.wall else "nothing")
    print(f"{hit.angle:>6.1f} deg  {hit.range:>5.2f} m  {met}")

How the loop works Open the Decision Lab