The Evolution of Planetside 2's Level Design
By the time I had joined the team, the framework for Planetside 2's environments was forming but still very much in flux. Indar was on the 2nd of its 4 total revamps. The world team had chosen a different take on Facilities than Planetside 1's cloned underground labyrinths. Since PS2 was on such an abbreviated development cycle, we were unsure of whether or not we'd get the tech in time to not only creating large underground spaces, but also lighting them without melting your GPU.
Opting instead for a variety of freestanding buildings, partitions, platforms, and terrain props, the sequel's capturable levels were slated to be unique, handcrafted, and about half a dozen times more numerous. The modular method of building bases was not only quicker to crank out content, but also leaned heavily on our instancing system that reduced the cost of commonly used props. Now that we were creating outposts with a different building method and only about a year left, we needed to quickly learn what made a good Facility, and bang one of them out every 3 days. Lots of R&D, lots of OT.
This writeup goes over where we started and what we found worked across the 3 layers of Planetside's level design, giving a brief overview of what we tried to end up where the game is today.
The addition of the Lattice system focused players into larger groups and moved them along more predictable routes, creating fights.
The Lattice did not exist at the game's launch, requiring friendly territory to only touch the enemy's for it to be contestable. Implementation of the Lattice was talked about before we had even launched beta, but it remained a contentious feature from Planetside 1 we held off on. Combined with the number of capturable outposts, many facets of the other layers suffered from the lack of structure: smaller fights, difficult to predict attacks, and increased difficulty when making Facilities. Since way more bases were vulnerable at any given time from many more angles, bases were constructed radially instead of linearly along the Lattice, bloating the time required to create solid levels due to the increased difficulty.
Base construction is tied into its location on the lattice and how many connections it has. Linear bases work best with <4 lattice connections, have a defined back end, and are quicker to implement. Radial bases are more difficult to put together, but perform much better at Lattice junctions where they're expected to be attacked from all sides.
We opted to give Indar the Lattice treatment 4 months after launch in February 2013. Our choices when planning for the first Lattice included making connections based on what made sense visually, removing outposts that fit poorly with the lattice, limiting the amount of bases with 4+ lattice connections, and better balancing each third of the map. My role on this last Indar revamp was also to manually tie each hex to a Facility, as well as better align key elements of each base to fit onto the lattice. Adjusting about 80% of Indar's bases averaging 4 hours each and the data work for hexes, the revamp ended up taking approximately 300 hours over the course of five weeks.
Despite the first implementation resulting in an asymmetrical Lattice gently jammed onto an asymmetrical continent, the improvements to Indar were still significant. The unintuitive canyons region in southeast Indar, for instance, now progressed in a logical manner instead of allowing armies to skip entire lanes. It was overall a success, preferred by the majority of players who enjoyed the new structure the feature gave continents.
We continued retrofitting older continents with the lattice over the next year, updating Esamir mid-2013 and Amerish in early 2014. Hossin, our latest continent, was designed with the Lattice from its conception, making it the most balanced at the highest layer of the game's level design.
Battlegrounds weren't a very developed concept at launch and weren't picked apart until we looked at the space between Quartz Ridge and Indar Excavation site post-Lattice. We were seeing fights float between these two outposts for hours at a time without either of them making much progress. They both functioned properly, and fights would take place at the outposts and be fun, but capturing territory always seemed just out of reach. An empire would have their soldiers and vehicles advance to the other side, fight their way over, and upon respawn would find themselves back at their old base on defense all of a sudden. What was causing this fight to stagnate?
While the answer didn't lie totally with the World team's domain, the majority of it had to do with the negative space between the two Facilities, the implementation of the Lattice, and coincidence. With the Lattice being laid on top of Indar, the dynamics of some areas changed with the reduced connections between outposts. Quartz Ridge and Indar Excav are a good example of this, as the open area previously was more flexible in where you could attack from.
Pre-Lattice, these bases could just attack everything that was next to them, and those bases could also fight whatever was next to them, so large forces rarely carved any sort of discernible path through the map. Enemy dug in too deep at a base? Push past and they'll give it up eventually. The Adjacency system led to more of a creeping, shapeless blob that grew and supported lots of smaller fights, and often one large fight that could be sieged indefinitely, a dynamic caused by difficulty in keeping player populations together.
Post-Lattice, bypassing either of these two bases meant a long detour, so they now had strategic value as well. So not only has the region suddenly become this high-value chokepoint, the half-kilometer of land between the bases coincidentally happened to be flat and barren, allowing whoever is standing on the high vantage points of either Facility to plink at the serpentine dodging infantry below. Good luck making it across!
So why did the fights stall? Originally, Facilities could expect attackers from 4-6 different directions and the pre-existing terrain had generally worked fine. If it didn't, going around a tough base was much easier. After the Lattice, blobs happened far less as a result of the new skeleton, and forces moved in more predictable ways while still being offered strategic choices at junctions. This led to some growing pains in cases like the segment between these two, as the empty space here was now all of a sudden a problem: the bone had no muscle.
Lack of cover on an early 2013 version of the battlefield (left) caused players to execute lots of failed attacks from bad locations. The current version (right) shows their tendency to gravitate around cover, as well as how additional sunderer locations mid-way allow for increased attack success.
Our biggest change to this Battleground were the coral smokers and large building intended to cover a mobile spawn point, regardless of which base you were attacking from. We also added some small rocks outside of both Facilities, which infantry then stuck alongside both it and the larger cover, changing their previous out-of-the-way behavior now that better options are available. Additional mobile spawn spots are available along the length of the Battleground as well, creating layers and redundancy when destroyed; the next spawn is more likely to be <200m away instead of >400m.
It was a great test ground to iterate on and see how the fight played out after the changes went live. We knew that the bases were fun, shared the same lattice, were on the edge of the map, and that it was a blank slate; our changes were totally additive. With the knowledge of how the fight functioned with nothing on it, we could directly observe changes in player behavior. Ultimately, less cover means longer Battleground fights and more killing fields, and adding more cover pushes the fight closer towards the Facilities.
As we found out how to build better Facilities, we learned that vehicles needed to be somewhat restricted inside them, so these open field fights provided much of the combined arms fights. The march across an open field with dozens of other soldiers and the subsequent clashes formed an iconic, immersive experience that was distinctly Planetside. We saw this idea cemented at industry events and conventions as well, where players would completely lose it upon seeing a full 100+ player fight in full swing.
The Facility layer is by far Planetside 2's most visible, concrete aspect of the game's level design, and the one that got fleshed out first. A facility consists of a capturable man-made settlement and the area immediately around it. PS2 was notable for having dozens upon dozens of these on a continent, each one handcrafted (save for the major facilities and construction sites) and themed to fit the game's sci-fi post-Earth world.
Making Facilities for Planetside 2 proved to be difficult early on, and we found ourselves revising our work often as we learned to craft better fights. As mentioned before, the game's unique features often meant unknown results when put into action, and the game's scale meant testing how something played usually required a live player population. 50-man internal playtests were not going to happen with the quantity and launch schedule we had, so the ability to mentally spin as many plates as possible when decision making, as well as being really good at guessing player behavior were extremely valuable skills when planning a base.
Our Facilities on northern Indar are pretty indicative of our simple, initial attempts at making levels: put a cap point in a building, craft a ring around it to block vehicles and prop it out. Theme!? This is where they put the crates! While these old style bases still function, their design is a little obsolete and bland, and we learned to build bases to take advantage of PS2's unique mechanics better over time. We also got much better at theming, where familiarity with the prop pool and practice gave way to interesting, world-appropriate structures, despite never having added any new environment art.
Over time, we pieced together much of the mystery surrounding getting a solid outpost that worked well with the other layers, wether it was as broad scope as integrating battlements and ramparts into bases, or some of the following smaller details that played a huge part when put together.
Scale: With the Lattice being put into place, many pre-existing Facilities became obsolete design-wise, as they were too small to contain the now-commonplace larger fights. As a result, it was important to build bases to scale to larger populations while still allowing small skirmishes. We accomplished this with multiple layers that create additional lanes when the population gets high, as well as having facilities have a larger minimum size.
Quartz Ridge at launch (left) compared to the latest iteration (right). SInce this Facility became an important chokepoint with the introduction of the Lattice, increasing its footprint and playable space allows it to accomodate the large fights it attracts.
Combined Arms: Early on we wanted to extend the combined arms aspect to most of the Facility as well. It didn't mesh well in many cases, usually resulting in infantry catching a tank shell in the face when leaving the spawn room. Towers are an exception to this rule, but most outposts are now restricted to infantry only as a result of this. Vehicles found their roles in siege combat, taking place right outside of a Facility, and in the Battleground, where they are their most powerful.
Hard Spawns: Spawn rooms are how the owners of a Facility come back to life if they're killed nearby; they're shielded and provide a safe place to take a second, get your equipment and push out. Many of our early bases suffered from spawn camping problems where the attackers would surround this saferoom (from a distance) and wait for players to exit. Since it was predictable where they were going to come out of, fights got one-sided once they reached a certain point of the capture process, or if the defenders were outnumbered. We mitigated this by using larger spawn rooms with better vantage points, DoT areas around spawn rooms for camping attackers, shielded teleporter rooms to warp defenders across the base, and multiple cap points to spread out attackers.
Playstyles: Planetside 2 gave players a lot of options for abilities and weapons, and this variety of choices bled into the level design as well. Hackable equipment terminals allow infiltrators to switch gear, adapting to counter nearby threats. Placing a fence or wall to block an alley acknowledges that you want most players to have to go around, but don't mind Light Assault flying over it as a shortcut. Placing any building down (or suspended flat surface for that matter) means you're aware that Light Assault will probably fly around and murder people standing on top of it. Siege gameplay can be created by adding cover/ramparts on walls overlooking roads, which will often establish a symbiotic Medic/Engineer/MAX vehicle-repelling ecosystem.
Performance: Though high frame rates were always a game-wide goal, guidelines to ensure solid performance affected the way bases were built. Our hand-built outposts opted to use a collection of buildings, props and construction materials (platforms and blocks) with the idea of taking advantage of hardware instancing, favoring a few common props over of a ton of unique ones. This ultimately lead to a lot of creative reuse of props to keep base designs fresh. The Auger, for instance, is a laser drill built out of smokestacks and oil pipes.
Terrain: The open world nature of the game lead to some unexpected problems regarding vehicles, terrain, and how they interacted with a Facility. Since player behavior is pretty unrestricted, a base that is perfectly fine in its own vacuum may have completely unexpected dynamics because of map features 150m+ away. A distant hill may be easy to access and become a popular spot to shell infantry in a 'secure' courtyard, or a tall mountain or canyon can make aircraft nearly impossible to knock out of the sky with ground-based AA. Here, we placed Facility-owned turrets for active defense, and used LOS blocking props like trees to passively bring vehicles in closer.
NPCs: Developer-placed NPCs were primarily related to base functionality: spawn rooms, capture points, etc. For performance reasons we never pushed this envelope too far; adding an extra NPC or two per outpost meant the server performance suffered. Vehicle and Equipment terminals, however, provided areas where they were present as high value for both defenders and attackers. Defenders can use equipment terminals it to change roles or resupply ammo, and attackers can hack it to gain some extra flexibility and a foothold when pushing into a base. For vehicle terminals, these can also be hacked by attackers to buy a mobile spawn point, deploy, and essentially create a siege out of thin air.