It actually sounds like a "Schema Architecture" that Arkins proposed in 1998 http://mitpress.mit.edu/catego.... You can implement it in about 10 lines of python because it's just that: the sum of attractive (goals) and repelling (obstacles) force vectors, weighted by the inverse of the distance squared. I was surprised OP didn't mention the Schema architecture, because it is exactly that, and since it sounds like a (simulated) robot game...
Your paper on newborn looking is really interesting. I build robotic models of the early visuo-motor system and am super interested in neonates but it's extremely hard to get any data from them and so I pretty much don't have any data from less than 8 weeks, which is really unfortunate.
Peer review "cannot" catch fraud and is not meant for it either.
Sure it is. That's the entire point, to determine if the research is valid. Just because they *do not* review it thoroughly, doesn't excuse them when they fail to catch fraud. "The reviewers do not, and cannot, replicate the results" And what *excatly* is preventing them?
The purpose of peer review is not to replicate results, it is to determine whether the methods are sound, as OP said.
What *exactly* is preventing them from replicating is: thousands of hours and millions of dollars of equipment. Not everyone has access to a trillion dollar LHC or super high tech bio lab, and even if the reviewer does, he is doing his own research and cannot spend his grant money or time to the experiment described in the paper just for the purpose of peer review.
Now, you might suggest re-vamping the system so that there is specific funding for scientists to peer review papers, but that is insane since there are literally thousands of papers published every month, and that is only counting the highest tiers of journals and proceedings.
Artificial intelligence has the same relation to intelligence as artificial flowers have to flowers. -- David Parnas