Hello guys!
We are now initializing the project "thelearningtriangle".
The main idea of "The Learning Triangle" is a game which consists of two major parts: A world and some creatures. These creatures want to survive inside the world, but in this project there isn't a player controlling the creatures. An algorithm must deal with this exercise.
The world is created randomly with walls, eatable fruits, energy/deadly fields and poisones fruits. The following image shows a first design and isn't final.
 | |||
| first sketch of a random world |
Creatures are symbolized as triangles. Each of them can move inside the world and explore the field. They need energy to keep themselfes alive so they have to find and eat eatable fruits. Creatures can replicate themselfes with a fixed cost plus 50% of the remaining health. They also have a running-speed and field of view. The last few visited places can be saved in a triangle.
Triangles Goal:
Walking distance as far as possible.
We want to program this project in python and java with Google's API for maschine learning "Tensorflow".
You can find our project on this GitHub page.
Hey guys,
ReplyDeletei've been reading through your posts and found the topic very interesting. As far as I understood, you are trying to make the "monsters" or "traingles" learn the best behavior pattern to survive in the labyrinth? So basically this is supposed to go into the direction of AI, isnt it?
Best of luck with your project!:)
Hey Daria,
Deleteyou're right, our idea is to let the triangles survive "on their own" in our world. Only the AI is possible to control them and has to find out the best way to do it.
Thank you for your support
TheLearningTriangel-Team
Hey guys,
ReplyDeleteI agree with Daria, it's a really nice idea! However, I have some questions:
1. You say there is no player controlling the game, but who then plays? Is there a player who can create the AI algorithm?
2. Are there different triangles in one match or only one with its duplicates?
3. How does the triangles find there ways? Can they scan the complete map or only a few fields?
Greets Hendrik
Hey Hendrik,
DeleteFirst, there is no player. The only thing the user can do is watch how the triangles move, die or explore the map. Everything is controlled by the AI once its started.
Next, we want to create different type of triangles with different stats and abilities, but the exact types aren't clear yet. We will discuss this in an upcoming blog entry.
Last but not least, the triangles move on their own (controlled by the AI), so first it will be random which direction they choose. They have a field of view and can't see everything, which means that the more the AI moves the triangles around, the more of the world will be scanned. So over time, the AI will know where to go to find energy and the movement of the triangles will be more "organized".
Hope I answered all questions, if not, just ask :D
Greets TheLearningTriangle-Team