JoeCraft

Progress Update

by on Jul.06, 2014, under Battle for Polaris, Personal

It’s been a few months, so I thought I’d post a progress update. I’ve been rather busy on the personal and work side of things, but have managed to work on some behind-the-scenes improvements in the meantime. Here’s a summary of what I’ve done over the last few months:

Battle for Polaris

I’ve finished the first set of major restructures and refactors I wanted to do. This included updating the codebase to work with latest OGRE and CEGUI, which was a job and a half in itself. I’ve also done some general tidy-up, and changed the project’s folder structure to separate header and source files. Since I’m trying to move towards having BfPEngine be a reusable framework, this made sense to me.

The main thing left to do in terms of refactor is the phasing out of OGRE code. I’m not particularly happy about depending on OGRE for simple things such as singletons, logging, string/real types, and so forth; I’ve worked quite a bit with logging systems – particularly Python’s nice logging system – since I did this, so I think I can do much better now. These plans are, however, on short-to-medium-term hold, since it feels like it’s been too long since I’ve worked on improving the meat of the gameplay and AI.

I’ve been reading back on my original plans from this project, from way back in my University days. I’ll be revisiting the old roadmap and writing up specific documentation as I go along, but the high-level summary of where I want to take the game AI is as follows:

  • General improvements to gameplay, and unit/group AI. Already summarised in my older (recently-updated) roadmap.
  • Use of probabilistic reasoning with Bayesian Networks in the high-level AI. Initially this would be to predict the player’s ship and sector positions, as a form of inferred influence map.
  • High-level AI that learns a player model over time, e.g. learning the player’s ship preferences. This already has a design, but may need revisiting after gameplay changes.
  • Investigate reinforcement learning AI. This is of academic interest, but in terms of practicality may need a more pragmatic hybrid approach.

I would like to further document this on the blog, so will be putting up posts with diagrams and overviews as I revisit and improve each section of the game code.

Interactive Fiction

This is still in the ideas stage. I would like to do something visual and gamey with this, but it has to work around my lack of artistic talent. I hope to draw up a proper design of this when I have a chance, though BfP is the priority at the moment.


Comments are closed.