I'm finally starting to work on the artificial intelligence component of my engine and I'm hitting some design questions that seem like they would benefit from some external input. What I'm looking at doing is some kind of rule-based optimized (probably A*, maybe a GA) bipartite matching system. This means that I would provide the user of my engine with some way to encode tasks and resources for each agent in the system as n-dimensional vectors into a potential field using scripting, instancing, and persistence systems that are already in place.
My question is, whether you understand the approach I outline above, would this be useful to you as a game programmer? Eventually I would like to release my engine as a as-close-as-possible complete open source game engine and I want to gear it as much as possible every step of the way towards what game programmers are actually looking for. This would give you a way to build agents into a simulation and give them resources that they would be able to optimally chain together in order to accomplish some task.
My other question is is there anybody on here that's interested in this kind of subject matter? If so, maybe we could work together on this. I think it's a fascinating field and I would be more than willing to share what little knowledge I have.