--- key: lesson/npc-perception type: lesson id: npc-perception lang: en title: NPC perception and target memory summary: >- Separate world state from character knowledge: visibility, last observation, forgetting and readable reactions. updatedAt: '2026-10-20' status: active url: 'https://gamemechanics.pro/en/lessons/npc-perception/' jsonUrl: 'https://gamemechanics.pro/api/v1/knowledge/en/lesson/npc-perception.json' markdownUrl: 'https://gamemechanics.pro/content/knowledge/en/lesson/npc-perception.md' aliases: [] tags: - npc-ai - perception - state-machines sourceIds: - source-52855da41e5f6f4d sources: - id: source-52855da41e5f6f4d title: 'Martin Walsh: Modeling Perception and Awareness in Splinter Cell Blacklist' url: >- https://www.gameaipro.com/GameAIPro2/GameAIPro2_Chapter28_Modeling_Perception_and_Awareness_in_Tom_Clancy%27s_Splinter_Cell_Blacklist.pdf verifiedAt: null relations: - type: requires direction: outgoing label: Before you start key: lesson/state-machines title: States and transitions url: 'https://gamemechanics.pro/en/lessons/state-machines/' jsonUrl: 'https://gamemechanics.pro/api/v1/knowledge/en/lesson/state-machines.json' markdownUrl: 'https://gamemechanics.pro/content/knowledge/en/lesson/state-machines.md' - type: requires direction: incoming label: Continue learning key: lesson/npc-action-selection title: How an NPC selects an action url: 'https://gamemechanics.pro/en/lessons/npc-action-selection/' jsonUrl: >- https://gamemechanics.pro/api/v1/knowledge/en/lesson/npc-action-selection.json markdownUrl: >- https://gamemechanics.pro/content/knowledge/en/lesson/npc-action-selection.md - type: explains direction: outgoing label: Apply the principle key: mechanic/boss-fight-atlas--boss-cover-line-of-sight title: Cover and line of sight url: >- https://gamemechanics.pro/en/knowledge/mechanic/boss-fight-atlas--boss-cover-line-of-sight/ jsonUrl: >- https://gamemechanics.pro/api/v1/knowledge/en/mechanic/boss-fight-atlas--boss-cover-line-of-sight.json markdownUrl: >- https://gamemechanics.pro/content/knowledge/en/mechanic/boss-fight-atlas--boss-cover-line-of-sight.md - type: explains direction: outgoing label: Apply the principle key: mechanic/voice-audio--family-22 title: 'Stealth, lures and noise control' url: 'https://gamemechanics.pro/en/knowledge/mechanic/voice-audio--family-22/' jsonUrl: >- https://gamemechanics.pro/api/v1/knowledge/en/mechanic/voice-audio--family-22.json markdownUrl: >- https://gamemechanics.pro/content/knowledge/en/mechanic/voice-audio--family-22.md - type: practices direction: outgoing label: Put it into practice key: practice/maintenance-patrol title: Maintenance patrol url: 'https://gamemechanics.pro/en/knowledge/practice/maintenance-patrol/' jsonUrl: >- https://gamemechanics.pro/api/v1/knowledge/en/practice/maintenance-patrol.json markdownUrl: >- https://gamemechanics.pro/content/knowledge/en/practice/maintenance-patrol.md - type: related direction: outgoing label: Related materials key: lesson/state-machines title: States and transitions url: 'https://gamemechanics.pro/en/lessons/state-machines/' jsonUrl: 'https://gamemechanics.pro/api/v1/knowledge/en/lesson/state-machines.json' markdownUrl: 'https://gamemechanics.pro/content/knowledge/en/lesson/state-machines.md' - type: reviews direction: outgoing label: Review through a lens key: lens/schell-028 title: The State Machine url: 'https://gamemechanics.pro/en/knowledge/lens/schell-028/' jsonUrl: 'https://gamemechanics.pro/api/v1/knowledge/en/lens/schell-028.json' markdownUrl: 'https://gamemechanics.pro/content/knowledge/en/lens/schell-028.md' - type: reviews direction: outgoing label: Review through a lens key: lens/schell-063 title: Feedback url: 'https://gamemechanics.pro/en/knowledge/lens/schell-063/' jsonUrl: 'https://gamemechanics.pro/api/v1/knowledge/en/lens/schell-063.json' markdownUrl: 'https://gamemechanics.pro/content/knowledge/en/lens/schell-063.md' - type: contains direction: incoming label: Learning paths key: path/npc-behavior title: NPC behavior url: 'https://gamemechanics.pro/en/learn/npc-behavior/' jsonUrl: 'https://gamemechanics.pro/api/v1/knowledge/en/path/npc-behavior.json' markdownUrl: 'https://gamemechanics.pro/content/knowledge/en/path/npc-behavior.md' placements: [] supersededBy: [] attributes: category: engineering level: intermediate minutes: 6 endsAt: null provenance: >- Original gamemechanics.pro material. Check illustrative examples in your own project. license: CC-BY-4.0 codeLicense: MIT attribution: gamemechanics.pro contributors --- # NPC perception and target memory id: npc-perception type: lesson key: lesson/npc-perception lang: en status: active url: https://gamemechanics.pro/en/lessons/npc-perception/ Separate world state from character knowledge: visibility, last observation, forgetting and readable reactions. ## Prerequisites Game code knows every object's coordinates. An NPC does not have to. Here, AI means a character behavior algorithm rather than a language model. ## The principle Separate observation from decision. Perception reports “target visible” or “noise heard,” memory retains the last known location, and behavior selects an action. Martin Walsh's Game AI Pro 2 chapter examines perception and awareness as game models where consistent rules and feedback matter. Start with [States and transitions](/en/lessons/state-machines/). Explicitly define which information each state can access. If an NPC uses hidden player coordinates after losing sight, that should be a deliberate game rule rather than accidental data leakage. ## Practical example ### An observation drone In this original teaching model, a drone detects the player within a 90° sector and 8 m range if a ray to a target sample point is unobstructed. Radius alone is insufficient: a wall blocks observation. A ray to one point is also a simplification and can behave oddly at cover edges. While the target is visible, the drone updates its last known position. After losing sight, it searches that position for two seconds, then resumes its route. It does not update memory through a wall. Illustrative pseudocode, not code for a particular engine: ```text visible = within_range(target) and within_cone(target) and clear_ray(target) if visible: memory.position = target.position memory.expires_at = now + 2.0 state = tracking else if now < memory.expires_at: state = searching search_position = memory.position else: state = patrol ``` Before the first observation, memory is assumed to be expired. This variant provides short memory without modeling accumulating suspicion. Add suspicion only if the gameplay rule needs it. ## Common mistakes - Range checks without obstruction checks let the NPC see through walls. - Search uses the hidden target's current location, making memory effectively omniscient. - States change without a cue, leaving the player unsure whether detection occurred. ## Verify it - Display the debug sector, ray, state and memory position. - Check a target in front, behind, behind a wall and at the range boundary. - Move the player after losing visibility and confirm the memory point stays fixed. - With discrete updates, check memory expiry on the first tick after its deadline. ## Prompt for an AI agent > Implement separate observation, memory and drone state. Prepare four test scenes and log transition reasons. State the ray and update-frequency simplifications. Visually identify searching without relying on color alone. Review through the [State Machine](/en/lenses/#schell-028) and [Feedback](/en/lenses/#schell-063) lenses. ## Connected materials - Before you start: [States and transitions](https://gamemechanics.pro/en/lessons/state-machines/) (key: lesson/state-machines) - Continue learning: [How an NPC selects an action](https://gamemechanics.pro/en/lessons/npc-action-selection/) (key: lesson/npc-action-selection) - Apply the principle: [Cover and line of sight](https://gamemechanics.pro/en/knowledge/mechanic/boss-fight-atlas--boss-cover-line-of-sight/) (key: mechanic/boss-fight-atlas--boss-cover-line-of-sight) - Apply the principle: [Stealth, lures and noise control](https://gamemechanics.pro/en/knowledge/mechanic/voice-audio--family-22/) (key: mechanic/voice-audio--family-22) - Put it into practice: [Maintenance patrol](https://gamemechanics.pro/en/knowledge/practice/maintenance-patrol/) (key: practice/maintenance-patrol) - Related materials: [States and transitions](https://gamemechanics.pro/en/lessons/state-machines/) (key: lesson/state-machines) - Review through a lens: [The State Machine](https://gamemechanics.pro/en/knowledge/lens/schell-028/) (key: lens/schell-028) - Review through a lens: [Feedback](https://gamemechanics.pro/en/knowledge/lens/schell-063/) (key: lens/schell-063) - Learning paths: [NPC behavior](https://gamemechanics.pro/en/learn/npc-behavior/) (key: path/npc-behavior) ## Sources and provenance Original gamemechanics.pro material. Check illustrative examples in your own project. - [Martin Walsh: Modeling Perception and Awareness in Splinter Cell Blacklist](https://www.gameaipro.com/GameAIPro2/GameAIPro2_Chapter28_Modeling_Perception_and_Awareness_in_Tom_Clancy%27s_Splinter_Cell_Blacklist.pdf) (id: source-52855da41e5f6f4d) CC-BY-4.0 · gamemechanics.pro contributors