Give the player (and AI) options
The core experience of a good FPS such as F.E.A.R. is the dynamic conflict between the player and seemingly-intelligent, active enemies. This means that both parties need meaningful combat verbs to exploit-- expressive movement, a wide variety of attack types-- as well as spaces which encourage and highlight the use of these verbs.
The worst place to roll out these combat mechanics is in an empty hallway-- no cover, no lateral movement potential, no interesting geometry for the AIs to interact with, no strategy, no surprise. Conversely, the best space is arena-like and varied, with an emphasis on flanking opportunities. The closer any given encounter space drifts towards the hallway model, the less interesting the gameplay there is going to be.
The primary elements of a good FPS encounter space are these:
1. Varied, clustered cover. Players and AI both need useful and varied cover for any kind of tactics to arise. Half-height and full-height cover each serve a purpose, as the verticality/laterality of each is significant (full-height cover is useful against elevated enemies, while half-height cover is invalidated; full-height cover forces actors to alter their lateral path while half-height can be vaulted, etc.)
Clustering of cover is important, as cover which is too evenly distributed becomes undifferentiated and leads to a flat experience. Cover should exist as discrete islands with meaningful no-man's land between each. This gives the player meaningful moment-to-moment choices to make ("should I risk exposing myself to enemy fire in favor of running for a better vantage point?") and causes AIs to be out in the open and moving laterally to the player's view on a frequent basis as they seek new cover, allowing for a shooting gallery experience of trying to take the enemy down before he reaches safety. The idea is to create meaningful points of emphasisinstead of an undifferentiated field of scattered, equally-useful cover nodes.
The most useful cover should be placed in the arena's mid-orbital, the dense ring between the outer edge and the central point of the encounter space. This encourages the player to move into the thick of the action instead of hanging on the periphery, and leaves the central dead zone as a no-man's land that remains risky to advance through, encouraging circular navigation.
Changes in elevation are also recommended, as high ground from which to fire down on enemies can be just as useful as a solid piece of cover to hide behind. Mid-field rises also provide the opportunity to observe the space mid-fight, allowing the player to reassess the situation and adjust his tactics accordingly.
2. Circular navigability. This goes back to the "as little like a hallway as possible" point. A good encounter space gives the actors options, and encourages variability each time an encounter plays out there. This requires not just a wide hallway with islands of cover distributed throughout it, but an open arena that is circularly navigable-- one with pathways around the edges which allow defended flanking movement. This encourages the player to advance and be mobile, and allows the AI to surprise the player by swooping in on their starting position from the side. A wide hallway with cover in it still boils down to advancing battle lines, while defended flanking corridors on the peripheral encourage the actors to circle around one another, take risks ("should I risk flanking into the thick of the enemy force to gain a better close-range firing position?") and generally be active instead of sticking to a single safe point and taking potshots. Circular arenas should give the player a multiplicity of options while keeping him wary of possible enemy flanking maneuvers, dynamics which are conversely defused by the binary flow of a linear hallway no matter how wide or cover-strewn.
3. Observability. As the player approaches an encounter space, he should be able to observe its major features and devise an initial plan of attack. This means that the entry point should feature a vantage point, often elevated, that illustrates the layout of navigable space, cover points, and interactive objects (explosives, water hazards.) All relevant features of the space should be visible and readable, and any element of the space that is obscured should be intentionally so (for instance, the terminal point of a flanking corridor might be obscured to increase the player's feeling of risk in attempting a flanking maneuver by reducing his knowledge of what lies at the other end.) The player, having observed the space, may hereby think beyond arm's reach once he's in the thick of a fight by relating his current position to the overview he saw before the encounter began. Should the player die during the fight, this initial vantage point on respawn provides a reminder of the space's layout, to aid his survivability for the next go round.
Assessing a space for these high-level principles should lay a strong groundwork which can be further refined-- by line of sight tuning, strategic item placement, lighting readability-- to form the basis for an excellent encounter.
The second aspect of setting up the encounter is blocking out the placement and initial behaviors of the enemy AI that the player will be facing. This determines how the player enters the fight, and ultimately how he walks away from it. In an FPS that features expressive combat mechanics and active enemies, the best place for the player to begin the fight is right in the middle of the action; how does one encourage him to dive in, instead of plinking at his foes from the sidelines?
One way is to give the player the first move-- let him get the drop on his enemies. This ties into the observability factor, while also encouraging the player to set up the fight to his advantage and close the distance before fighting starts.
In this scenario, the player approaches the encounter space and observes his opponents standing or patrolling around in the center or at the far end, unaware of his presence. These enemies should be spread out enough that a single grenade blast won't take them all out, and having backup waiting in the wings is important. The player may observe the enemies' movements undisturbed as long as he doesn't attack or advance too close. This presents the player with options-- does he hang back on the outer ring of cover and line up a headshot on one of the enemies? Does he plant some proximity mines around the flanking corridors then toss a grenade at the group to make them scatter? Does he close the distance and open up with automatic fire just as they notice his presence? The player is allowed to choose his tactics and consider his approach. This is invaluable from a player experiential standpoint.
The opposite experience is often encountered in F.E.A.R. 2: as the player steps through a doorway into the fight arena, enemies are already aware of his presence and spraying the entry point with suppressive fire. What options does the player have now? The only valid ones are to retreat and use the edge of the entry door as cover, or to dash blindly forward into a hail of bullets, which is most often suicide. An unaware enemy is key-- it allows the player to strike the match setting off the encounter, instead of being purely reactive to his opposition's opening moves. It allows the player to take up an optimal position for beginning the fight, which a good level designer makes sure is significantly deeper into the arena than the entry door. It allows the respawning player to intentionally alter his tactics upon retry, instead of being forced to deal with the exact same setup each time.
An unaware enemy is subject to extermination by an opening headshot or grenade, but this is a small price to pay-- backup can be spawned at the far end of the arena to replace any intial fodder as a second wave of enemies advances into the encounter. The gains in player control over the fight's initial moments are worth it.
An alternate approach is the ambush-- the player observes a quiet arena, and advances into the middle, only for the enemy to pop out of hiding and attack (rappel down through skylights, jump down off of balconies, swarm in through multiple entry doors, burst through a wall, etc.) This is a fair approach in the back half of the campaign, as the player should be experienced fighting his enemy and could use some variety to encounter setups. However, the ambushing enemies should nonetheless have terrible reflexes-- enemies that pop out guns blazing will merely frustrate the player. Rappelling/door-bashing/balcony-diving/wall-busting ambushers should take a while to ready their weapons and draw a bead on the player, allowing him to make it to cover and get the first shot off. The idea is for the player to retain some initial advantage while still being thrust suddenly into the middle of an encounter.
My experience with F.E.A.R. 2 is that it unfortunately often misses the principles that made the encounters in the original game so engaging-- frequent are restrictive, linear encounter spaces without flanking corridors, precognitive enemies that begin firing on the player before he gets a chance to enter the space, and unobservable spaces without clear flow or points of emphasis. This not only makes the player's role in combat more frustrating, but makes the enemies appear less intelligent-- with fewer navigational options, they tend to remain stationary more and surprise the player less. Smart AI is only half the equation-- smart arena design is required to convincingly demonstrate your enemies' innate abilities. Hopefully the points above will help guide your encounter design towards showcasing your game's AI in the most flattering possible light, making the enemies look-- and the player feel-- as smart as possible.
