Eufloria > Eufloria Classic Mods

New Files Update

<< < (2/3) > >>

Aino:
Wish I was that creative, starting my own company with an idea that would be successful :)

I'm currently studying to become an engineer: Automation. I have a little over a year left of the study and then I'll officially become an engineer, but I don't think my future truly lies with engineering work. I'm hope I'll end up with a software developer job after my bachelor, which I guess is slowly becoming my dream job! :)
Besides my studies, I'm also working with C# as a hobby. Making my own little game engine just for fun is quite challenging, and working on it always reminds me I still have quite a lot to learn.

Already started with some basic setup code for the AI, just need to figure out it's behaviour and write the code. I'm also curious if Eufloria allows for multi threading, would be awesome to put that pesky and heavy neighbour checking work on another thread, alleviating the stress on the main game loop. I guess I'll have to look into it, and thread safety while I'm at it. I'll take on this project with a little more planning than previous projects though, so it might take some time.
I will most definitively not lose this one though, learning from your mistakes has that curious effect. :D

Edit: Worked on the AI some more today, looking promising. I'm weirded out by the fact that I actually don't know why the AI does what it does... It does what I expect, but currently I haven't added any sort of prioritization of planting trees or assembling to attack, but it does so anyways. I did check if the default AI is to blame, and it isn't. This is good I think, because that means I somehow made a pretty good foundation for the AI. It does have SOME A LOT of irrational behaviour that I hope to iron out soon; maybe the release will be sooner than expected? :D

Edit 2: Working on the AI some more I have finally made it behave as I expect it should. This is awesome, now to add some behaviours that is missing (defending and attacking), and see if it can beat MAI :)

Edit 3: I figure I can add a to-do list in this post too, so everyone curious can keep track of the progress:

--- Code: ---- Basic behaviour, like moving seedlings and exploring.  (done)
- Plant trees (done)
- Use one slot on extremely good asteroids for defense trees (not done, don't know if this is a great idea or not; losing out on good seedlings)
- Characterized by being the best asteroid in the empire.
- Defend properly ( 80% done :) )
- Attack properly
- Flowers:
- Plant flowers on the best asteroids in the empire
- Plant flowers on defense trees first
- Plant flowers on dyson trees second
- Mines
- Deploy mines on friendly threatened asteroids
- If no threatened asteroids, use for offensive purposes
--- End code ---

I think that covers the behaviours the AI should exhibit, I might have missed a few smaller points... :)

Edit 4: For the first time since starting this, the new AI beat my older one... It's rather curious, and it's because of an exploit in the older AI. I consider it a win nonetheless, considering I'm not done implementing any sort of offensive behaviour as of yet (the AI does attack, but it doesn't have any code directly telling it to be on the offensive). It will be interesting to see what happens once I add behaviour that makes it take more offensive actions :)

Oh, and it's really fun to see just how small adjustments in the mathematical formulas change the behaviour. It's not radical changes, but one adjustment in one formula made the AI go from loser to winner against MAI :D
- Note: this is a double-edged sword, one change can cripple the AI as much as it can boost it, but developing it from the ground up has it's benefits of making me see where some pitfalls are though :)

Lost Seedling:
Aino- glad to hear you’re almost done with your schooling. Maybe you’ll eventually find work developing artificial intelligence for self-driving cars or flying taxis- a possible future boom-industry. AI certainly seems like a sector of technology still with potential for enormous growth and opportunity. How would you like to work for a company like Boston Dynamics?
https://youtu.be/5iV_hB08Uns

Great stuff on the new Eufloria AI! I’ve got 9256 hours logged now in Eufloria, most of which is observing the various AI’s in action. Obviously, it is a source of endless fascination for me and I’ve learned a great deal about the strengths and weaknesses in Infected, Merchant, and Mutant and can usually predict which will win just by looking at the initial map configuration.
When I was working on Mutant, I sometimes found it could soundly beat Infected, but not Merchant. Or, with different optimization it could often defeat Merchant, but not Infected. They seemed to have nearly opposite strategies. Changing the many map options also had dramatic effects on the results. Because I tried to minimize Mutant’s drain on system resources, it means the AI is very slow to make decisions and take actions. This causes it to do poorly on maps with large numbers of asteroids or long send-distances.

Only by trying out many hundreds of different maps and configurations was I able to achieve a satisfactory win/lose ratio and at that point I stopped working on Mutant, despite some fatal flaws in its algorithms. I certainly would do things differently if I was starting again knowing what I know now!

Eclipse-B was an attempt to easily generate many different map scenarios to test the fitness of any AI combatants. Unfortunately the native Eufloria AI was not suitable to feature both because of its influence on other AI behaviors and the fact it was designed to intimidate and confuse human opponents and not the cold calculation of another AI. Simply put it was not competitive.

In some of the generated maps, I tried to minimize any positional advantage held by one opponent or another. However, I found the numbering system of the asteroids introduced an imbalance which can drastically influence the outcome of battle unfairly. To overcome this maps can be saved and replayed with the starting points reversed.

I’ll enjoy reading of your progress and can hardly wait to try out your new AI. Perhaps you might want to start a dedicated thread to its development. I hope you find Eclipse-B helpful in your testing.

Lost Seedling:
Aino, did you give up already on the new AI? The silence is deafening!

Aino:
Nope, encountered some problems with the logic in the AI, need to rethink my approach. The silence is because of exams, sorry... :)

To be clear; I want to finish the AI, but it's not a high priority in my life at the moment, I hope that's understandable...

Lost Seedling:
Totally understandable. I'll soon be traveling for a while anyway so I'll check back in around November for an update. Good luck with the exams!

Navigation

[0] Message Index

[#] Next page

[*] Previous page