The Road Ahead, AI Update

The next update for Chopper Tank Flak will focus on improving the AI. I have mostly followed a development strategy of "make the worst aspect of the game better" and I think the artillery update highlighted some significant issues with the AI. Specifically, the AI struggles in situations where it needs to think more than 1 turn ahead.

Currently the AI uses ranked choice to evaluate what every unit could do and then chooses the best choice and moves the associated unit. It then repeats that process until it has moved all its units.

So far this has proven very effective at:

• keeping its units out of harms way
• maximizing the effect of units
• exploiting opportunities to attack

but is pretty bad at:

• moving its units in a manner that seems consistent between turns
• identifying obstacles to victory and how to overcome them
• accepting short term losses for long term rewards
• performing defensive actions

To address this the next update will overhaul the AI's method of evaluating moves to be more adaptable. Based on some research I have done it seems that I have made my AI with a "backwards" problem solving process. Currently, it looks at what it has available, and then determines what it could do with its recourses to get closer to victory. Instead, the AI should first determine what it needs to do to win, check for obstacles that prevent this, and create an ordered list of tasks with the las task being a victory condition.

I am hoping to get this done by the end of October. I have set my deadline further than I did with the artillery update because the implementation involved is more complicated. Testing will also be more complicated because of the unpredictable nature of a game's AI.