Autonomy level is a first-class entity attribute describing how much decision authority an autonomous system holds.
| Level | Name | Human role |
|---|---|---|
| 0 | manual |
Full human control |
| 1 | assisted |
Driver/operator assistance |
| 2 | partial_automation |
Human monitors automation |
| 3 | conditional_autonomy |
System drives within ODD; human fallback |
| 4 | high_autonomy |
System drives within ODD; human not required |
| 5 | full_autonomy |
No human intervention expected |
Autonomy level affects governance validation for:
| Domain | Influence |
|---|---|
| Decision authority | Max level capped by deployment profile |
| Recovery | Higher autonomy requires automated recovery policies |
| Human approval | Levels 0–3 require responsible person |
| Safety validation | Level 3+ requires trusted entity posture |
| Trust | Minimum trust tier escalates with level |
| Mission planning | Approval workflows scale with level |
| Readiness | High autonomy missions require readiness gates |
[governance]
autonomy_level = "conditional_autonomy"
Per-robot override via metadata:
# spanda.devices.toml metadata or device tree
[governance]
autonomy_level = "partial_automation"
CLI and API accept flexible aliases: level_3, l3, 3, conditional, conditional_autonomy.
Each deployment profile sets max_autonomy_level. Governance validation warns when entity autonomy
exceeds the profile maximum.