A while ago a buddy of mine was working on a program that weighted certain tiles of a map as "good" or "bad", using neural networks. Basically, areas around monsters were bad, and areas around food or warps to other areas were good. What is this called, so I can look up some tutorials on how to do it right?

I don't think there is a name for this. It sounds like a combination of pathfinding, neural networks, and either supervised or reinforcement learning depending on if your friend trained the network to values he determined or if he used an algorithm to let the network find its own values. The inputs to the network might correspond to the things around a given position, to evaluate that position's cost. Most graph search algorithms can handle variable edge weights.

Or at the same time, you could use an ant method to implement a dynamic version of this. Basically the units would leave trails indicating usage amount. If an ant is killed going out on that trail, he's not going to increase its weight coming back, thus it will be lower than the other, safer, trail.

What it sounds to me like that was doing was using these weights as inputs to the neural net, or in other words those were how the neural net perceived the world, and then were allowed to learn/use previous knowledge and respond appropriately.

While not exactly the same thing as you describe, "Influence Maps" are closely related.

Sounds like Influence Mapping and Value Function learning in bed together!

Quote: Original post by Extrarius
While not exactly the same thing as you describe, "Influence Maps" are closely related.

That's it--I think :D; just need to find a lot of good information and examples on it.

Hi,
You may use A* for weighting and using weighted calculations. Say I have a unit at Point 1 trying to reach Point 2 over a map having float values for passibility. If the unit is killed on the way, reduce passibility value. This will make A* algorithm (slightly modified for this purpose, less passible tiles having longer time to pass as it is used in variable terrain calculations) to take another route. And I dont think neural networks should be involved in this system.

The OP didn't say anything about how the AI learned the weights - just that they were there. It could be a miscommunication. It sounds simply like influence maps to me.

AI Game Programming Wisdom 2 had an article about NN and Influence maps.

Quote: Original post by tonyg
AI Game Programming Wisdom 2 had an article about NN and Influence maps.

Wow... indeed it does. Complete with code, too. Buy the book!

Quote: Strategic Decision-Making with Neural Networks and Influence Maps

Penny Sweetser (School of ITEE, University of Queensland)
AI Game Programming Wisdom 2, 2003.
Topics: Strategy, Tactical; Genres: Strategy, RTS
Abstract: Influence maps provide a strategic perspective in games that allows strategic assessment and decisions to be made based on the current game state. Influence maps consist of several layers, each representing different variables in the game, layered over a geographical representation of the game map. When a decision needs to be made by the AI player, some or all of these layers are combined via a weighted sum to provide an overall idea of the suitability of each area on the map for the current decision. However, the use of a weighted sum has certain limitations.

This article explains how a neural network can be used in place of a weighted sum, to analyze the data from the influence map and make a strategic decision. First, this article will summarize influence maps, describe the current application of a weighted sum and outline the associated advantages and disadvantages. Following this, it will explain how a neural network can be used in place of a weighted sum and the benefits and drawbacks associated with this alternative. Additionally, it will go into detail about how a neural network can be implemented for this application, illustrated with diagrams.