Wednesday 5 June 2013

Rethinking artificial intelligence

A while back I removed AI system I had been developing for some time and merged it to a command structure. Maybe I was looking for a simpler solution, but after some thinking it's obvious that there must be a separate AI logic with the original primary/secondary task idea.

How tasks will be organized is not yet clear, but each of the creature has to keep something in mind, a goal to which it is reaching. A number of things can change the goal, but not secondary tasks which involve things like swimming or some other circumstance as a temporary block between the creature and its target or goal.

I'm too dumb to use any sophisticated AI algorithms, but I think you don't need any of that to make creatures do what you want. Also, algorithms tend to produce predictable patterns that can be exploited. I intend to create an unpredictable AI that can sometimes make huge mistakes or forget something important. If I can make it so that the player will see those mistakes then it's going to be cool.

1 comment:

  1. I think the biggest challenge with enemy behaviour is communicating it to the player. Players will assume random systems are intelligent and complicated systems are dumb if you have no way of showing the reasons for their behaviour. Smart Kobold was great for this - it very effectively communicated enemy intent and reasoning.

    ReplyDelete