Running in Circles

So last week’s post was a lot about the planning process and reconstruction coming out of the decision to restart the project from scratch. This week I’m happy to report that in terms of the basic functionality, the floor generator is just about finished and notably more functional than the original proof of concept version! Other news this week includes a complete reconsideration of how objectives will work in the game that renders most of the flow chart from last week obsolete… You’re gonna have to trust me on this one. Lets talk about it.

So before truly diving into making the floor generator smarter and more capable of generating interesting environments, I figured I’d do some competitive analysis on the game’s main inspiration, Lethal Company. I really like the generation in this game, so it’s only natural to try and see if I can figure out how it works.

I started out with decompiling the game using JetBrains dotPeak with the help of some of the modding resources I found online. This let me read through the raw code for how the generator works, but due to being decompiled, it’s pretty hard to just read it like a book and I found my eyes glazing over a little bit. Despite this, I did glean some information about some possible improvements I may make to my own generator down the line, but nothing I’ve implemented so far. After this, I figured I’d just dive into the game and explore the levels after they’re generated to see if I can work out any patterns. With the help of a mod for the game known as Imperium, I was able to get some really good looks at how the generation works. There’s a handful of useful tools this provided me such as god mode, pausing the timer, and night vision. That said, the most notable thing this mod provides is the map.

Imperium’s map

This map was invaluable for collecting data. Compared to the in game map the game itself provides at the terminal, this one allows me to pivot the camera, zoom in and out, toggle certain visibility channels, and just generally get a much better view of what’s going on. The first thing that became immediately obvious is that the game’s generator works entirely on a grid system, and I do mean entirely — any room that seems to break the grid is just a clever use of angles within a room — no room in the game as far as I can tell has a door that deviates from 90 degree angles. That said, there were a handful of rooms that it took a moment to figure out.

The confusing rooms, placed in a PureRef board for comparison.

At first glance, these rooms seem to break the rules of the generator. They either have doors that seem to be off angle or off grid. As I hope is obvious now from this comparison, these rooms accomplish this with clever room design. The left room has a cheeky angle and strange cutoff points for if a room doesn’t spawn off its branches. Despite this, all its connection points are completely on grid and angle. Similar deal with the room on the right — the room is comprised of a hallway and a room. The hallway then has 4 connection points — one on either end and 2 on the left side near the top. These revelations gave me a bit of confidence that maybe I could also just use a grid system, which will become important a little later, but before I committed to that, I wanted to do a similar observation of the game’s newest tile set, the mineshaft.

The cave section of the mineshaft level

As you can see, even the cave part of the mineshaft is very clearly on a grid. Rooms like the underwater tunnel or diagonal rooms are just that, larger rooms that help break up the grid. With that confirmed, it was decided — my generator would use a grid system. Now, to understand the reason for this, let’s review how exactly the generator works.

A rough image I drew when explaining this concept to a friend.

At its most basic form, the generator creates a floor like a tree, starting from one point and spreading out. This is described by the picture above. Green doors would generate open, blue doors would generate closed — the red door is the one we’d like to possibly be open, but currently generates closed. In the old project file, I did have a solution for this. Each door frame would have an overlap that sticks out half way. This would mean that if 2 door frames lined up, these overlap boxes would overlap each other, then they could detect this and open their frame. This sort of worked — it did open doors and create loops — but this meant that every door that lined up would open, which meant in order to avoid this, I’d need several different intersection pieces to make sure you can’t walk from anywhere to anywhere in basically a straight line. The problem is the generator itself doesn’t have references to the rooms it’s creating, so it can’t be the decider of which doors open and which stay closed. The solution I came up with for this in the new version of the generator is to use the game time since creation of each room to establish seniority and let the older room in any situation make the decision of whether or not to open the path.

A level generated with this new system for opening paths

An example of a long loop generated in one of these levels

I’m happy to report that this system basically worked flawlessly, opening up meandering and maze like routes throughout these levels comprised entirely out of 4 way intersection pieces. This system along with some more varied room types is bound to make levels that are easy to get lost in. Now, with that element out of the way, lets take a step back and talk objectives.

Part of the reason I hopped into Lethal Company to do some competitive analysis is I was starting to realize that even the simplest objectives I was thinking of for Maintenance Crew ask a lot of the floor generator. For example, the simplest objective I thought of was a keycard retrieval task. Basically the players would have to find a keycard in the map and bring it back to the elevator. This already means that there needs to be a step where the generator spawns all objective-relevant items before moving on to other points of interest. This also creates a new failure point in the system if for some reason the generator were to fail to spawn all objective-relevant items. Additionally, we probably also want to make sure we generate this relatively close to the elevator so you don’t get screwed over having to get to the farthest possible room and back. Anything more complicated than this could need specific rooms generated in specific sections of the generation, and I was starting to think that it was all too complicated and possibly not fun for players.

While I was doing my analysis of the rooms in Lethal Company, I was thinking about how similar games handle objectives. Lethal Company has 1 generic objective: collect scrap. Content Warning, a similar game, though without procedural generation this time, also has 1 generic objective: record scary things. These games can get away with confusing and maze like maps because no matter where you go, you’re probably going to find something that will help progress your objective. With the objective system I was thinking of, either the game would need some kind of system that shows you where your objectives are, a system for finding those objectives via easier to find elements in the map, similar to GTFO, or a system that generates multiple of the objective to increase the chances of players stumbling into them. The first option invalidates most of the generated map by just guiding players right to where they need to go, the second feels complicated out of the scope of this game, and the third feels lazy and like a band-aid solution. Due to this, I started thinking about how I can simplify my objectives into a generic one that happens every floor. The first step of this was coming up with a short sentence like the ones for Lethal Company and Content Warning to describe this generic task: perform maintenance. The game, after all, is about meeting a minimum quota of maintenance in an old derelict facility — the players should go into each floor with a quota to fulfil.

Instead of having a whole system of objectives that put a big strain on the generator, I decided it would be a good idea to simply make use of the planned point of interest system instead. With this, I can design a bunch of smaller tasks for players to take care of that contribute to their quota for the floor. This new distinction also means that I can rearrange my data tables so that Floors references Tile Sets which further references Objectives so that I can have better control of what objectives spawn in what tile sets. This also has the fun side effect of letting players push the limits a little bit on how long they can stay on a floor. Since every floor is always gonna spawn way more objectives than are needed for their quota and because I intend to reward money for objectives completed, there’s an incentive in staying longer than strictly necessary, which wasn’t there before.

With all of that said, the one things this loses out on is specific control over the objectives spawning in a particular floor, which did make me somewhat sad. This did lead me to coming up with a new idea during a conversation with one of my friends about the topic. In essence, the idea is to have secret objectives that can be assigned to specific floors. This idea is still fairly early on in its conceptualization and is not currently on my todo list for getting a playable prototype ready by the end of the year, so take all of this with a grain of salt. The basic idea is on a particular level, the generator will generate a special room, objective, item, or otherwise. This thing is specific to that floor and is tied into a larger floor-spanning objective. In this case, lets say floor 2 generates a special keycard somewhere. The players could take this keycard back to their elevator and take it with them. Then, on say floor 5, maybe a special room generates in the floor that requires that keycard to open. As you can see, this has some fun gameplay implications that are not a massive strain on the floor generator and can allow me to have some fun with storytelling and puzzles. In theory, I could have an entire section of the floor locked behind this hypothetical room without needing to change anything about the floor generator as it exists right now — it could simply be a new tile set that’s only used there.

With this past week of work behind me, I think I’ve finally got a clear vision of what this game is going to look like, at least in its first iteration. Up until now I haven’t felt completely confident that I’d be able to have a prototype by the end of the year, but now I think it’s a real possibility! I’m not sure if I’ll be able to post a public build of the game when it’s ready (currently all the online functionality relies on Steam) but it’s definitely something I want to look into when I get closer to that point! With most of the work on the level generator itself out of the way, expect to see more in the way of actual gameplay elements coming soon.