Although I do like the general concept of AI in stealth games having a “vision” cone (and a aural one, in the case of Mark of the Ninja), I think there’s some tweaking to be done. It’s just way too incredibly silly to be able to stand just outside of the cone, in front of an enemy, and not be seen. But as soon as your avatar is touching the cone, you’re spotted.
The most recent stealth game I played, Assassin’s Creed Chronicles: China did one interesting thing to help. The cone will slowly fill up with red as you sit inside of it’s range, indicating impending detection. It’s still a little obtuse though, as it still counts against the player as being “detected” even if they exit the cone before it fills completely.
I think to make a truly realistic 2D stealth simulation, the amount of input the AI receives needs to be expanded. Their sight cone should have stages to it; the closest being “complete awareness,” a middle section being “aware of movement” and the furthest being “unaware.” Things like objects in the environment, shadows, color, and most importantly light, should play an important factor in dynamically adjusting their awareness cone. Peripheral vision also needs to be taken into account, both for the player’s benefit and the enemy’s.
And as for sound, I think it’s less important how loud something is, and more important how far away the sound is, as well as how different the sound is from other sounds that are already happening in the environment. The player could use the soundscape to their advantage to help them mask certain things, such as knocking an enemy out while a truck passes by, to avoid others nearby from hearing it.
May 14, 2015