A fairly productive weekend, all things considered. I spent pretty much all day Saturday coding, and probably most of Friday evening. I took a siesta on Sunday, putting forth the last steps to shucking Lavos out of its shell in Chrono Trigger DS. I've pretty much exhausted all the side quests, and now I've little doubt my level 55ish heroes can pull it off.

Wasps

My communal-level defenders of the planets have undergone a radical restructuring of concept over the weekend. Instead, they're colonies of insects of sorts whose main focus is in redistributing nutrients. I'm fairly reminded of the Zerg from Starcraft. I've even implemented a Creep of sorts, this insect-like race actually changes the surface of the planet to suit their needs.

The behavior was surprisingly hive-like. The "Queen" lays an egg which hatches a "worker." The worker finds nutrients (such as grassland), consumes it, and returns the nutrients to the Queen. The Queen uses these nutrients to lay more eggs. Before you know it, there's hundreds of workers swarming around, looking for nutrients.



I found the CPU tends to bottleneck around the 700th worker. It's not a BYOND problem, it's a game design problem: I really shouldn't need 700 moving NPCs for an effective swarm implementation. What's going on right now is the Queen is getting glutted with more nutrients, outpacing her egg-production capacity, as she has nothing good to spend them on.

I think I'm ready to make the game a little more advanced now - introducing more specialized roles to my workers. Unless I decide to make these guys as little more than cannon fodder, I need to make them interesting for players to play. Perpetually feeding the queen is probably not most players' idea of a good time. (Granted, there is a working model of a game that has found a way to make that interesting.)

This is primarily a game involving conflict between multiple factions, after all. Before I'm done, these little virtual bugs are going to be building actual hives, complete with specialized chambers and deadly defenses.

Smart Guns

I mentioned in earlier comments the concept of having a weapon contain also the AI in using it. I tested that principal this weekend and found that there's a certain degree in which it works well.

At one point, I had code that would work a bit like a chess AI. It would look at every weapon in the mob's contents and get those weapons to return a score reflecting its viability as the best action. That was going too far. The mob carrying the weapon usually has a primary goal in mind that the weapon isn't able to be made aware of. Besides, having the AI on the weapons perpetually re-assessing the situation of the mob carrying them is just too bloaty.

Instead, I've adopted an understanding that each weapon in a mob's inventory addresses a specific need. The mob handles the state-level choice of which need to address and chooses the weapon which addresses that need. Then, the weapon contains the AI to place itself in a state in which it can activate. That works rather well.

In case it isn't clear by now, when I'm saying "weapon," I'm actually referring to actual abilities that mob carries in its inventory, which do things more diverse than merely attacking. These same abilities are actually usable by players controlling those mobs, using the same procedures that the AI utilizes.

Next Milestone

I can really put it off no longer: if it is indeed my plan to have these units drive like vehicles, I'm going to need to add movement code that takes the place (or rather harnesses) step_towards and other like procs. This basically breaks down to plotting out moves ahead of time and sequencing them at an appropriate speed.