Demonstrates operator approval gates in octos pipelines — the robot pauses before each navigation and waits for approval.

Your chemical plant has 3 inspection zones: Zone A (routine), Zone B (high-voltage equipment), and Zone C (clean room). The robot patrols all three zones autonomously. Last week, maintenance was working in Zone B with the power panels open. The robot drove in anyway — nobody told it to skip Zone B. The near-miss triggered a safety review: "How do we let the robot run autonomously while keeping humans in control of where it goes?"

Full autonomy is fast but blind to context. Full manual control is safe but slow. Most real operations need mixed autonomy: the robot handles the routine steps, but a human approves high-stakes decisions. This is not a rare pattern — it's how aviation (autopilot + pilot), surgery (da Vinci + surgeon), and nuclear power (automation + operator) all work.

Gate nodes give you this pattern in 3 lines of DOT:

gate_b [label="Approve Zone B?" type="gate"]

The pipeline pauses. The policy checks. If approved, the robot proceeds. If rejected, it skips that zone and continues. No code changes, no redeployment — just a policy update.

• Gate nodes in DOT pipelines (type="gate")
• Gate approval/rejection policies
• Skipping pipeline steps when gates are rejected
• Industrial pattern: human-in-the-loop safety

The pipeline has 3 gates (before stations A, B, and home). The GATE_POLICY controls which gates are approved:

• Default: "Station A,home" — approves A and home, rejects B (zone under maintenance)
• Robot navigates to A, skips B, returns home

cd 04-human-gate
dora up && dora start dataflow.yaml --attach

[gate 2/10] APPROVED: Approve navigation to Station A?
  Reason: policy: 'station a' matches gate
[step 3/10] navigate_to({'waypoint': 'A'})
  Checkpoint: at_A
[step 4/10] read_lidar({})
[gate 5/10] REJECTED: Approve navigation to Station B?
  Reason: policy: gate not in approved list [Station A,home]
[step 6/10] SKIPPED: navigate_to_B (gate rejected)
[step 7/10] read_lidar({})
[gate 8/10] APPROVED: Approve return to home?
  Reason: policy: 'home' matches gate
[step 9/10] navigate_to({'waypoint': 'home'})
  Checkpoint: at_home

Mission Summary:
  Approved gates: ['Approve navigation to Station A?', 'Approve return to home?']
  Rejected gates: ['Approve navigation to Station B?']
  Tools executed: [get_robot_state, navigate_to, read_lidar, read_lidar, navigate_to, get_robot_state]

Edit GATE_POLICY in dataflow.yaml:

Try 05-slam-nav-sim for visual MuJoCo simulation.