That Skipable Preamble.
It’s only been about 3 months since I first opened Unity3D. The interface alone was daunting. Up till that point I was using Buildbox, a Stripped to the raw essentials game making tool. Unity3D is a beast, a titan in comparison. You have a high level view and can zoom all the way in to the raw code that makes everything work. With Buildbox, you only have the high level view. What lurks beneath is comply hidden from view. This is a blessing and a curse.
BuildBox has no AI. It’s all path driven behaviors. There’s nothing cool or fancy about it. The games you can make are super limited. Thus the market is flooded with Flappy Bird and Bullet Hell clones. I was spoiled with the luxury of not thinking as I assembled graphic art and put together basic games. But I wanted more, something real. I hate mediocrity. As much as I love “Rocket Shark” and “Drill Deeper” they were mediocre. Simplistic, fun games to pass the time while waiting for your coffee. There’s absolutely nothing wrong with that. But I want to give more, make bigger, better games. Of course time and money at this moment are at a premium, I need to start very small and work my ass off to get where I want. Yes, my ass is literally worked off. It’s numb as fuck sitting 16 hours in this worn out Ikea chair.
So I fronted some cash, and bought a prefab Platformer engine. Time again. I needed a shortcut. Something up to the task of getting a decent, quality game published. Boy that was a mistake. $500 in graphical assets and 100 redo’s later I’m finally on the right path. Yeah, I’ve been running myself thin, this is my break. A collection of fragmented thoughts.
Simple AI is NOT Simple
So that prefab engine I got, was a holy cow mess. The code was a nightmare. I have a programming background since College days, so I know shit code. Aside from that, I’ve spent the past year sitting in front of Udemy programming classes 8+ hours a day. Classes and book work does not make a good programmer though.. Getting out there and actually doing it does. So I opened the monster code and got terrified. It was like staring into the abyss and seeing it look back at you. The terror it sends down the spine shaking your very soul.
I spent the last 4 days working on the AI for one character, the typical “goomba”. Despite being a 2D engine the guy used a mashup of 2d and 3d Unity calls. It generally “worked” but weird things were happening. He used a maxX and minX variable with a Vector3 clamp to keep the character where it belongs. Not a big deal right? Wrong! When dropping a prefab into the scene, the monster “vanishes”. As I later learned, clamp is like a cage. It will forcibly put anything back within it’s grasp. So I unshackled my Goobma.
Without Gravity, without the chains to bind him, my goomba would just walk away. Through walls, through everything to escape the scene bounds. The guy (previous “developer”) cheated in an extraordinary way. That goobma didn’t give one f*ck. He was outa here.
So I set up a Test Lab
I needed a scene where my goomba could play happily without fear of being head smashed. I created a simple scene from some block prefabs and put them all together.

I hate that scene. But it yielded fantastic results. At first I thought I succeeded, but forgot to turn off clamp. Fuck “Clamp”. Then he still passed through walls, even with RigidBody and Boxcollider2D. Then I discovered the beauty of Raycast. Designed for bullets and projectiles, useful for so much more. But the damn goomba still walked through walls.
It’s all About the Angle
Have I told you I started Unity3d 3 months ago? Yeah, so I was just sending rays all over the place. I figured out how to use Vector2 directions to get down to a single Raycast in the direction my goomba was going. Bingo, he dodged walls. Next up, not falling off the platform.
This was simple right? Cast a ray at a 45 degree angle from the direction the goomba is walking? Well in theory yes. The original developer was tracking direction as a plus or minus 1. This needed to be converted to a Vector2. Finally the X coordinate was fine, I just needed to add in a negative Y and got my 45 degree angle.

Great. It worked. Set up a bool to tell if I want him to edge detect or not and I have a working goobma. But…
He ended up freaking out at the crates and “drilling” his way through the floor. After a few hours of “WTF Man”, I discovered it was a very simple fix. The BoxCollider2D was off by a few pixels. So the straight down RayCast was registering that one pixel gap as both an edge and not ground sending my goomba on a frenzied search for sanity.
I fixed all my prefabs and now he works.
There’s a Lesson in all this
It’s not like I lost 4 days development trying to rework shitty kludged AI. I learned a lot on how to work around Unity3D’s flaws. I had to simulate my own falling gravity arc for the goomba. I’ll have to do it again for other characters. But I now have a pretty decent platformer AI scaffold. From here I can build the other characters. The flying bats and skulls. The angry cats and pissy squirrels. The foundation of my goomba will become the basic AI of my more advance characters.
Oh shit philosophy!
Yeah, so I see the countless hours of “failure” as learning. I’ve adjusted one system only to break 10 more. Just today after completing the Goomba, I realized I had broken the Player’s Cling ability. That only took a few hours to fix because of everything I learned from the whacked out Goomba. Because of my struggles, the future implementations will become that much easier.
Anyway, holy cow this was a long “blog”. If you read this far, you’re awesome and thank you so very much.