Ninja Boss Fight said:
To be honest with you, I'm not giving the full picture.

I had figured there was more to the story. I've been on this forum for a very long time and the number of people who come in and ask a very specific question and are responding in a way that suggests that their very specific question hasnt been answered typically have something that they're not telling us… unfortunately, I can only reply to what i'm seeing.

Ninja Boss Fight said:
The problem is that NPCs tend to be predictable and take from the immersion of the player in the game.[…]

So, I want to contribute a piece of software that showcases one or several ways in which autonomous agents can be beneficial to gameplay, what these agents are capable of and how to create them.

IMO: this is a good topic, however may be deemed as too broad. How you describe it to your supervisor needs to follow on the more complicated response you've given in the last reply, than what you started this thread with - avoid oversimplifying with the supervisor as you described:

Ninja Boss Fight said:
[…] I proposed to my supervisors the idea to redevelop this technology with new features […]

Simply, while the objective may be the same, you're likely to be significantly expanding on your initial investigations, you've completed a survey approach, you've done the lit review, you can say there are shortfalls in the approaches out there, and that you've even investigated some of them, further exemplifying that there is issues in these limited approaches, that you want to compare and contrast - practically - a number of the solutions, expanding on the work that you had done previously and going into much more depth than simply “Refactoring your existing code”.

If you're read Springer publication's “Believable Bots” (https://doi.org/10.1007/978-3-642-32323-2), you might find some things to challenge and expand on. The section titles all seem very much in line with the things you're thinking about and may serve as some inspiration for a subject title.

@taby Could you elaborate more on the maze. It depends on what the purpose of solving a maze is. What will solving the maze create as an impact? Do you mean pathfinding or search algorithms? If so what would be the goal, what are we testing for?

When I say autonomous I mean a virtual entity that acts on its own without direct interaction. In the case of ants, each ant is an agent and performs tasks as a single unit regardless of the impact that work done has on he colony. It is self sufficient and in the context of a game, preserves its life as best it can and performs its duty as best it can.

@stragen Thanks for the advise. Do you mind if I stay in touch with through the development?

@Ninja Boss Fight Not a problem at all.

Artificial intelligence and Machine Learning all boil down to a base optimization problem. You describe the problem space, solution space, all constraints and try to find a global optima at fastest time. Hill climbing with random jump to get out of local optimum, for instance.

Many masters students write their first research paper (or two) and craft it into a thesis project. Getting the paper into an IEEE or ACM journal is somewhat common. Based on the work you've already done, is there anything you have learned and can identify as a gap in current published knowledge?

If you want to work on something that hasn't been done much, work on team coordination. Command a squad in a combat game, and have them do the job properly. There are lots of games that offer squad combat; see if you can run your NPCs into one.

I didn't read through all previous responses, but I skimmed and didn't see someone say the word arXiv, so I think my advice can help.

You should, always, when interested in any research project, first ask your advisor where such a project would be published and then try and find the most similar publication in that work to see how people talk about it and what you an use as a crutch. You don't need to understand it necessarily, but you should have your eyes on the first actual section once or twice.

The instruction here is that your project should be the equivalent of a full time job for three and a half months. This is a significant fraction of your life, so spending two or three hours every Monday morning scrolling through the relevant section of arXiv.org, even if you don't read anything most days, is just part of the expense report. As it is, projects like cellular automata are a rich subject in research https://arxiv.org/pdf/2006.13532.pdf,​ and probably fall closest as an academic question to what you've already described.

Don't expect it to be obvious how you can contribute, anything that is an idea that you would have upon initially reading any work should be an idea had by the author, and possibly addressed. If I had a penny for every time I've thought ‘yeah, but what about BLANK’ just for the author to go ‘So, we considered BLANK, and it’s either the next section, or a stupid idea,' I would have about a pound/dollar.

Eventually you'll either learn something your advisor didn't know, in which case you write that up and present that to them, or you'll come up with a high-level overview of the system, in which case you talk that out and submit that. Either is a great result. In the best case result you'll get something new and you could work with someone to help you make it a publishable result.

@mtlk Now to take my own advice.