A couple of years ago I designed my own modular dungeon tiles. I bought a bunch of gesso boards and sprayed black chalkboard paint onto them. When dried, I carefully drew black 1" lines across the boards to create a grid. The idea is that you can create a campaign by designing different rooms and connecting them together. You can draw on the tiles with white chalk markers, which can just be washed afterwards.
It looks a lot nicer than the dry-erase marker systems that are very popular. I wanted something that looked really nice with the cool miniatures!
I actually don't have many friends who are into D&D so unfortunately I never really got to play a real campaign with them. However, it was definitely great inspiration, and some of the things I read about and learned while working with these miniatures and tiles definitely helped me to form my own ideas about procedural games.
Minotaur's Maze doesn't use fully modeled rooms that snap together, which is a very common practice for procedural games, and for good reason: It lets artists really hand-craft each room and give them personality and visual quality. I decided that Minotaur's Maze was going to be a true-to-form roguelike with very granular level generation, which happens at the tile level, rather than with entire rooms at a time. Since it's just me, and I don't have dedicated environment artists to actually create all of those rooms, there's a certain economy in defining everything with algorithms and parameters.
About six months before I started working on Minotaur's Maze, I decided to try my hand at making a traditional ASCII roguelike. I wrote everything from scratch in C++, and even created my own ASCII table that was a combination of Japanese and western character glyphs from the 80's. I used to draw ANSI art when I was a kid so this was something I took very seriously. I also designed a custom color palette. This gave the game a unique look that felt plausibly retro but still felt different.
The game had a unique lighting engine as well, which included shadow casting, but also supported semi-opaque occluders in the world. In environments like forests or whatnot, the light would continue to travel giving the lighting a volumetric look. Neat!
I never finished the game, but I did have the engine in a functional state, with a world editor and a combat engine. This game did not have procedural environments. I got pretty far with it but abandoned the project once I started work on Minotaur's Maze. Some of the pathfinding and AI logic was ported over but that's it.