- Jan 13, 2009
- 2,333
- 18
- 81
First of all, sorry for the incoming wall of text. I'll try to pepper it with gifs to make it more readable and to point the unlucky ones that have yet to play this gem.
Playing Shadow Of Mordor and Alien: Isolation recently got me thinking about this.
I was optimistic about the future of gaming when I played the awesome FEAR game but it turned out to be an exception rather than becoming the norm.
I wish I could say it became the standard but games that have come out since don't have nearly as sophisticated AI as FEAR did.
Typically, stepping up the difficulty in games doesn't produce smarter AI, it's either more of equally stupid bots at once to make objectives
harder to achieve or it's just a boost in health points for the bots to give them slightly better odds vs. the player.
More often than not, this results in frustration rather than a challenge or forces you to heavily exploit weaknesses you find in AI behavior.
The outcome is far from desirable, dying because you didn't get to click the button 5 times in a second or still breezing through a supposedly harder game isn't the right solution.
FEAR wasn't the first game to have enemies with a decent FOV, without severe glaucoma or late stage cataracts or to have NPC's attempt to flank you but it nailed that part and did a lot of other things right too.
IMO, FEAR was one of the the first titles that gave depth to AI behavior. If you run into a squad of 6 and eliminate 5 of them quickly or brutally, the last guy may give up and run.
When they engage you, they'll use the environment against you. They crouch and go prone to shoot at you, knock objects over to create cover.
If there are 2 or more enemies present, they may lay suppression fire on you to make it easier for the other to flank you. If you slaughter one of them, the next guy might refuse the order to step into his place.
The spotting mechanics were realistic, enemy reacted to you using the flashlight properly unlike many games with NPC's oblivious to your actions.
Even HL2, which had better AI than the average game of it's time looked flawed in comparison, an example: http://youtu.be/e0WqAmuSXEQ?t=36s
That was one part of the equation, the other was getting feedback since the player was listening to enemy radio chatter. The dialogues helped the immersion and some of the exchanges were genuinely funny.
The lead designer said they had to cut 40% due to limitations. Here is an example just to size up the scope, NPC's had a lot to say, took an arrow to the knee wouldn't have stood out: http://www.youtube.com/watch?v=qX73XG9e5BE
The in game physics helped elevate the game to it's status. 9 out of 10 games feature indestructible/immovable books and magazines, you are lucky if you can get a bullet hole decal on them.
FEAR let you shred paper to pieces if you wanted to. Back to the AI. Monolith used a different system by the name of STRIPS, developed by MIT.
Being just a end user and a fan of a game nearly forgotten, I can't figure out why this hasn't been utilized more.
Here is an image from a document that goes deeper into detail, also linked below:
Link to the PDF detailing the "three states and a plan" approach:
http://alumni.media.mit.edu/~jorkin/gdc2006_orkin_jeff_fear.pdf
Playing Shadow Of Mordor and Alien: Isolation recently got me thinking about this.
I was optimistic about the future of gaming when I played the awesome FEAR game but it turned out to be an exception rather than becoming the norm.
I wish I could say it became the standard but games that have come out since don't have nearly as sophisticated AI as FEAR did.
Typically, stepping up the difficulty in games doesn't produce smarter AI, it's either more of equally stupid bots at once to make objectives
harder to achieve or it's just a boost in health points for the bots to give them slightly better odds vs. the player.
More often than not, this results in frustration rather than a challenge or forces you to heavily exploit weaknesses you find in AI behavior.
The outcome is far from desirable, dying because you didn't get to click the button 5 times in a second or still breezing through a supposedly harder game isn't the right solution.
FEAR wasn't the first game to have enemies with a decent FOV, without severe glaucoma or late stage cataracts or to have NPC's attempt to flank you but it nailed that part and did a lot of other things right too.
IMO, FEAR was one of the the first titles that gave depth to AI behavior. If you run into a squad of 6 and eliminate 5 of them quickly or brutally, the last guy may give up and run.
When they engage you, they'll use the environment against you. They crouch and go prone to shoot at you, knock objects over to create cover.
If there are 2 or more enemies present, they may lay suppression fire on you to make it easier for the other to flank you. If you slaughter one of them, the next guy might refuse the order to step into his place.
The spotting mechanics were realistic, enemy reacted to you using the flashlight properly unlike many games with NPC's oblivious to your actions.
Even HL2, which had better AI than the average game of it's time looked flawed in comparison, an example: http://youtu.be/e0WqAmuSXEQ?t=36s
That was one part of the equation, the other was getting feedback since the player was listening to enemy radio chatter. The dialogues helped the immersion and some of the exchanges were genuinely funny.
The lead designer said they had to cut 40% due to limitations. Here is an example just to size up the scope, NPC's had a lot to say, took an arrow to the knee wouldn't have stood out: http://www.youtube.com/watch?v=qX73XG9e5BE
The in game physics helped elevate the game to it's status. 9 out of 10 games feature indestructible/immovable books and magazines, you are lucky if you can get a bullet hole decal on them.
FEAR let you shred paper to pieces if you wanted to. Back to the AI. Monolith used a different system by the name of STRIPS, developed by MIT.
Being just a end user and a fan of a game nearly forgotten, I can't figure out why this hasn't been utilized more.
Here is an image from a document that goes deeper into detail, also linked below:
Link to the PDF detailing the "three states and a plan" approach:
http://alumni.media.mit.edu/~jorkin/gdc2006_orkin_jeff_fear.pdf
Facebook
Twitter