The above video was recorded on 3/3/2025 to demonstrate the most current version of the gameplay at the time

Above is a flow chart I created upon deciding to take this project seriously describing the general gameplay loop, including the menu and the general steps the procedural generation system would take in order to create levels. I created this flow chart a couple of weeks after deciding to start this project fresh with a new project file. The reasoning for this is I was working in an old project file that I had created for a different game I had abandoned. Due to this, not only was the name of the project file no longer representative of the project, but it had a lot of bloat in the files and temporary, hacky code meant to just make things work. Since I was making a new project file from scratch, I decided it would be best to get the gameplay loop written down visually so it could guide me through the design process. This flow chart has since been made obsolete after deciding to redesign how objectives and consequently the floor generator as a whole would function, but this still serves to explain the general gameplay loop.

This video is a demonstration of the level generator in my old project file. The generator itself is extremely simple, almost completely randomly placing rooms on any open door it can find. It only knows about doors so you can see every so often a room will flash into existence before immediately disappearing, as it spawned colliding with another room. This version of the generator did have a couple improvements I made to it in order to make it slightly smarter than completely random. Firstly, I created a system for giving rooms weights. In the data table of rooms, you could assign a float value no smaller than 0.1. This would then be multiplied by 10 and truncated to be used to indicate how many instances of that room’s class reference should be in the array that the generator would randomly pull from to pick what room to spawn. The other improvement I created was tile sets. As you can see, some of the rooms in the generation are a purple color. These rooms are of a different tile set than the white rooms. The way this functioned is each tile set would have a data table indicating the rooms and weights for that set. These data tables would then be referred to in a larger tile set data table which also included a tag name, for example “Tileset1”. Finally, each door on a room would be tagged with 1 or more of those tags to indicate what types of tile sets would be allowed to spawn off of that door. While this did function decently, It ended up being wildly inconsistent and tended to create small patches of the purple tile set on the fringes of a level.

Since transitioning to the new project file, the floor generator has improved massively, with nearly the whole generator being rewritten a couple of weeks ago (2/19/2025). The generator went through a handful of “improvements” intended to help it create more interesting and consistent generation that ultimately ended up being scrapped and replaced with a much more elegant solution in this rewrite. The rewrite essentially involved doing away with room weights, tile sets, and sections in favor of a new socket system and hand-picking room count ranges for each floor. This came with several massive improvements to the generation. The most significant of these, in my opinion, is the removal of the old sections system. Essentially, this system would break the generation up into several distinct sections that could each have their parameters defined in the floor data. The problem with this is that the sections would always generate in a linear fashion, i.e. section 1 leads to section 2 leads to section 3 and so on. The new system defines “sockets” for each doorway. These sockets are essentially a matching game — rooms may only connect to each other if their touching doors share a socket. This new system provides several benefits over the old one. First and foremost, it allows me to create sections of somewhat different looking generation without necessitating that these sections flow linearly. This also means that in theory I could create micro-sections and really fine tune generation exactly how I wish. The other major improvement this allows for is having different door types in the same socket. Since the system distinguishes between different connection types, each of these sockets can have a predefined set of valid doors for the doorway. Since tile sets were removed, I had the opportunity to change how rooms are picked for a level as well. Instead of using predefined sets of rooms with weights, I transitioned to using a system of hand picked rooms. For each floor, I decide exactly which rooms I want to be on it as well as in what quantities. The system not only allows me to pick specific values (including 1, if I want to have a room that spawns once and exactly once on a floor), but also ranges of rooms. This new method not only provides that capability for unique rooms, which I wanted but wasn’t sure how I’d implement before, but also means that generation is more consistent and is far less likely to generate a disproportionate number of one room type or fail to generate another room type.

After doing this rewrite, I was far happier with the floor generation system. A few days afterwards, I had a conversation with my art guy Aidan (who you can find here) about the rewrite. We discussed what it would take to start getting some rooms modeled for the game so that the grey boxes can be replaced. This lead to us writing out a list of parameters to act as guidelines for how to design rooms in a consistent manner. These parameters are mostly size constraints that, other than the minimums, are allowed to be pushed for particularly special rooms in order to distinguish them from the normal ones. Once we finished writing out these parameters, we decided it would be best for me to design a brand new floor from scratch utilizing these new parameters to see how they feel and adjust them if necessary. Surprisingly, the only parameter I felt should be changed was our minimum room size, which we originally started at 500 x 500, but I upped it to 750 x 750. About 4 days after this meeting, I considered the prototype floor to be complete, including multiple door types, across 2 different sockets, a large focal point room, and a handful of new rooms. With these rooms made and being happy with them, I decided it was time to move on to working on some objectives.

As I started to consider objectives, there were two major improvements I wanted to make to the generator in order to facilitate more interesting objectives. The first of these is taking a hint from the improved room spawning system and implementing sockets for objectives. This allows me to segment objectives based on their size, theme, or otherwise. It wouldn’t make much sense for a broken pipe to spawn in a cubicle or a stack of untidy papers to be on the floor of a hallway. This socket system fixes this problem. The other problem I had was that I wanted to have objectives that are unique to certain rooms. The current system sort of could do this, by having a socket that only exists in that room and only has 1 objective, but the problem is the existing system wouldn’t guarantee that objective would spawn. The solution was just to pre-place these objectives in their respective rooms. As the system worked, this would work just fine, though the generator would over-generate objectives. To fix this, all I had to do was add another step to spawning a room that detected these pre-placed objectives and took note of their point values. With this improvement, I think I can finally actually call the prototype level complete. You can see all of these features in the video at the very top of the page.

In terms of next steps in order to get to that alpha version I want to reach this year, I’m mostly focused on getting a few floors, each with unique generation, several objective types, including at least one unique objective per floor, and as many enemy types I can think of and implement. Beyond that, there’s several other features that I’d like to implement into the game. The main and most important one of these is a large series of puzzles for players to solve in order to discover secrets and progress the game’s story, similar to Black Ops: Zombies’ easter egg hunts. These puzzles would span multiple floors and likely involve not only running through the game several times, but also solving meta puzzles potentially outside of the game. I’m also interested in adding a system of crafting that encourages more exploration of the levels by having blueprints that would be discovered in the level as well as the crafting stations themselves spawning randomly in the level. I’m still undecided whether or not I’ll add this system, but it seems like it could add a lot of interesting opportunity, especially when it comes to the puzzles.