In the last update, I noted that I was working on a system to make the Mantanoids (that’s the main recurring enemy of the game) appear smarter.
Today, I did a round of work on Hexoshi, and I decided to scrap the system I had in mind because it just had too many problems and “what ifs”. Instead, I’ve added adaptive machine learning to the Mantanoids’ AI.
It’s nothing too crazy. The way it works is really simple: a new segment has been added to the config file to record “A.I. data”. This is a list of entries that note an action a Mantanoid has attempted to get from one point or another, and whether or not the action “passed” (i.e. got closer to the destination than it was prior to the action) or “failed”. The A.I. then reads that list to decide in the future if those same actions in those same positions are worthwhile.
Still, it seems to be having the desired effect. While the Mantanoid A.I. is by no means intelligent with this, it is at the very least basically competent with a little training. That’s all I ever needed it to be, and I am quite happy with the result.
So now, all I have to do is add some missing sound effects, train the Mantanoid A.I. some more (the plan is to release Milestone 1 with the Mantanoids pre-trained), and make sure none of the level design causes the new Mantanoids to be unreasonable. Tentatively, I would say that Milestone 1 will probably be out next week.
Next week is also most likely when I will start the final two voting phases: one for Anneroy’s last name, and one for the name of the planet Hexoshi takes place on, so stay tuned for that as well.