I need to write a wumpus agent which has no information about the location of the wumpus or the gold whatsoever. The agent can sense stink and breeze which means that the wumpus or a pit is nearby. It can sense glimmering as well meaning that it found the gold. I've generated a map, the map is a grid - based one (tilemap).

The thing is I have no idea how to start out. The agent can move, or rotate 90 degrees and then move. How can an agent choose a path when for example it has two perfectly fine grids (it has no knowledge about a pit, wumpus or brezze, stink of any kind) to which it could move. How should it choose between these two paths for example? How should the agent be able to choose a path in general? Is there an algorithm for it? I don't know if I could use A* because the agent does not know where the gold is, yet.

I will be thankful even if you can only tell me an algorithm or theory which could be useful for this.

Thanks!

The first approach should always be "how would YOU decide what to do"? Then see if you can put it down in code. You could also look at the situation as an optimization problem and then use some sort of Monte Carlo technique. But that might be overkill.

Oh, and there's always this secret website.