Forums > General Discussion > Artificial Intelligence in Stealth Games
| Artificial Intelligence in Stealth Games | |
|---|---|
| Posted: 21 Apr 2026 05:47 UTC | Post #1 |
| Jon Bern Deck & Engine |
Registered Total Posts: 11 |
| I’m programming the AI for a stealth-based title, and I’m having a hard time making the guards feel "smart" but fair. Right now, they either spot the player instantly from across the map or act completely oblivious. How do you implement a "suspicion" meter that feels natural? Should I use a state machine with a "search" phase, or is a behavior tree a better fit for complex guard patterns? I want the player to feel the tension of being hunted without feeling cheated by the code. | |
| Posted: 21 Apr 2026 05:48 UTC | Post #2 |
| Terry Rogers Deck & Engine |
Registered Total Posts: 12 |
| Behavior trees are definitely the way to go for stealth AI! They allow for much more modularity, so you can easily add "investigate" or "call for backup" states. The key to fairness is "visual cues"—the guard should make a sound or have a question mark icon before they fully engage. To clear my mind after a long day of debugging pathfinding nodes, I like to play on https://bj-88-gb.uk/. Taking a break from complex logic helps me return with a fresh perspective on player engagement. Remember to give the guards a limited cone of vision and clear "peripheral" zones to keep the gameplay balanced! | |
1 / 1