2016 Spring AI Project – LD33 & Neural Networks

As my senior year of high school began to slow down and the college hype began to increase, I looked for simple projects to work on while in the midst of the craziness. Thinking that it had been a while since I had made anything relating to artificial intelligence, I wanted to refresh my mind on Neural Networks and machine learning in general. Since I needed a game to attach an AI to, I naturally thought back to my Ludum Dare 33 game (linked here), as it was a rather simple game that wouldn’t be hard for a computer to master.

I used a Neural Network (read more about Neural Networks here) with 2 input nodes (one to tell when an arrow was close and another to tell when a house was close), 2 hidden layers of 5 nodes each, and 4 output nodes (controlled each of the 4 keys required to play the game: punch, block, jump, move right). The activation function was again a sigmoid (1 / (1 + e^-4.9x)). I used a genetic algorithm with a 10% mutation rate on each gene passed down in order to weed out bad Networks and incentivize growth of good Networks.

Screen Shot 2016-03-25 at 11.11.00 PM

Graphic displaying the structure of the Neural Network I used for this project. Inputs were either a -1 or a +1 depending on whether or not its specific object was close (-1 for false, +1 for true). A key was pressed if its respective output value was > 0.5 (the range for the sinusoid I used was (-1, 1)).

Despite my belief that my game was easy, initial runs of the simulation proved that although scoring points was easy, there were too many ways to earn points, which confused the networks. The Networks easily figured out that moving right was a good way to earn points as points were given based on how many civilians are squished, but it had a hard time figuring out that it could squish civilians longer if it blocked arrows. However, after many many iterations, Networks could generally figure out a better strategy for scoring (whether it be jumping on houses to destroy them or blocking arrows). Despite the Networks learning how to score points better, I have yet to see a perfect strategy formed (one that jumps on houses AND blocks arrows), which may be solved if I ran more iterations of the simulation.

ld33aistart.gif

Starting out you can see that it doesn’t do anything. Usually, if it does anything, it holds down random keys, causing its actions to be sporadic at best.

ezgif-2212419532.gif

After ~5 minutes of the simulation running, the AI finds out that walking right is a good strategy for getting points.

ld33aiBLOCK.gif

After a considerably long time, the AI has figured out blocking is a good combo with walking, as it can walk longer distances than if it wasn’t blocking. However, it hasn’t found out that destroying houses is a good source of points.

All in all I think this was a great refresher for AI. Since I’ve been mostly working on games these past few months, it was also a good break from the grind that asset creation is becoming. Also, learning that machines aren’t always as bright as I may think they are is a great lesson to take going forward in my AI career. If you want to see the code for this you can see it here: https://github.com/mccloskeybr/LD33

Advertisements

2016 Spring Game Project (In Progress) – Adventures of Mark

A lot has happened since my previous update nearly a month ago. The main area of my focus has been switching from a top down view to something more isometric (think games like Earthbound), and the perspective challenges that are associated with that. I’ve also added a lot of new objects, tiles, and items (including various shops). Making this game has really opened my eyes to just how many assets and things of the like are required for not just indie-games like this, but professionally done AAA titles as well, and how difficult and tedious asset creation can be at times.

One of my more recent undertakings has been overhauling the way I create levels in the game. Before, I used a system of images where each pixel was a tile in the level, which was tedious and inefficient for a variety of reasons. Memory wise, it required a relatively large amount of effort to load each level. Making the levels themselves was also difficult because I had to use a specific set of colors with precise R, G, and B values to get the desired outcome, and didn’t get to see what the final outcome would look like until I loaded it into the game. Now, I use a level editor that I made myself that really facilitates level making because you can actually see the tiles you’re placing instead of just solid colored pixels.

Screen Shot 2016-03-15 at 12.47.26 PM

Although it is a simple editor, it has everything I need to make levels quickly and efficiently. The lighter tiles are floor tiles while the darker tiles are walls. The blank spaces are place holders while I create more tiles to be input into the game.

Screen Shot 2016-03-15 at 12.48.14 PM

The map I created inside the level editor loaded into game.

It did take a little while to complete the map editor, and I still have to add objects, but the benefits definitely outweigh the costs. Now, the tile IDs are saved to a text file that can be easily read and made into a tile map that’s loaded pretty quickly.

Screen Shot 2016-03-15 at 5.06.39 PM

The text file created by the editor to be loaded into the game. This particular text file creates the level shown above.

When I first thought of making a level editor, I thought the idea was a good one, but backed away because I wasn’t familiar with a lot of the window components needed to make one (for instance JButtons and JToolBars). However, I’m glad I worked through and figured out how everything meshes together because making it was fun and interesting, and I can make levels more efficiently from now on.

If you want to see the game as it progresses in detail, you can visit the github repo here: https://github.com/VolitionDevelopment/The-Adventures-of-Mark

2016 Spring Game Project – Plane Challenge

As my senior year of high school began to slow down, I began to look at more relaxing and laid-back  projects. Games, to me, are fairly low stress as there isn’t that much problem solving (at least at the early stages), so I decided to make a couple of games during my final semester of High School.

On my spring break trip to Grand Cayman, I decided I wanted to see how much of a game I could make on the plane ride down. This resulted in a fairly simple game that was completed in roughly 2 hours (give or take 10 minutes). Granted, I did start laying the ground work while I was waiting to board, but most of the work was done on the plane itself.

The game itself ended up being about planes. As I didn’t have too much time to work on it, I decided to make a simple endless runner in which the player dodged obstacles while trying to get as far as they could get without dying. The player, in my game, controls a plane that can move left and right, dodging debris while also picking up health packs.

Screen Shot 2016-03-06 at 11.38.42 AM

As you can see, I really whipped out my AP Drawing art skills on this one. In reality, I wanted to spend more time coding than on asset creation (which definitely shows in the final product).

One of the challenges of making this game was creating an endlessly looping background. The way I ended up doing it was having one image that was the size of the window (500 x 500 pixels) that was stacked on top of itself 3 times. The resulting picture scrolls down and when it hits a specific point (close to the end), it jumps back up exactly 1 image height and continues to scroll down, resulting in the illusion of an endlessly scrolling image.

If I had a bit more time to work on this project, I’d first add a menu, then a scoring system that went up the longer the player was alive.

I’m now going to start uploading all of my projects to Github, so exploring the code is a little bit easier. If you want to see the code for this project, you can see it here: https://github.com/mccloskeybr/PlaneProject