Unity enemy AI mechanic / 13 sec
Unity Enemy AI: Threat Priority and Target Arbitration
See a live visible target correctly preempt older evidence in a Unity stealth AI system without deleting the guard memory of the previous clue.
What this clip demonstrates
- Evidence is tracked independently by source and stimulus type.
- Priority, confidence, and recency decide which threat deserves action.
- A new live target can interrupt a stale investigation without erasing it.
Implementation notes
The strongest awareness value is not always the best target. A saturated body-discovery track may keep the guard highly alarmed, but a player currently in view is the more urgent thing to act on.
Argus separates overall suspicion from threat selection. Priority rules choose the action target, while independent memory tracks preserve the earlier evidence. That makes interruption and resumption behavior deterministic and debuggable.
Build the underlying mechanic
Design a readable suspicion meter for Unity stealth AI using evidence tracks, tier thresholds, decay, memory, hysteresis, threat priority, and player-facing feedback.
Read: Stealth suspicion meter design