The Array/HashWumpus Agent

A Basic Hero for the Wumpus Environment

The ArrayWumpus agent and the HashWumpus agent are the first agents of a family of agents that develops a better strategy for succeeding in the Wumpus environment with each generation. Those two strains co-exist peacefully together and even use the same strategy for discerning their actions or representing the environment. There are only three differences in the two strains:

• The ArrayWumpus and its successors all use a static array to represent the environment while the HashWumpus and its successors generate their 'view' of the environment dynamically using a hash - list.
• All array-organized agents use the strategy of going right first then up then left and down at last while all hash list-organized agents use the strategy of going first left then down then right and up at last.
• Array-organized agents can be told to be 'daring' which means that they even venture to maybe unsave locations if they have found no gold in save location. Hash-list-organized agents always leave the cave - even without a piece of gold - without stepping in may dangerous places. As the penalty for dying is a rather big one, it seems to be an unnecessary risk to instruct an agent to be 'daring'.

Representation of the Environment

The agent deduces the following conclusions/facts from the percepts it receives:

• The agent marks its start field as save. As it always starts at the upper right corner of the maze it marks the fields to its right and above as walls.
• If the agent has moved to a field not yet visited, it marks it as a save field ( i.e. without pit or wumpus ).
• If a bump is percepted, the field that lies in the direction of the last moving action is marked as a wall.
• If a scream is percepted, every field is marked as not housing the wumpus.
• If a stench is percepted and any three out of four of the surrounding fields are marked as not housing the wumpus, the fourth field is marked as housing the wumpus.
• If a breeze is percepted and any three out of four of the surrounding fields are marked as not containing a pit, the fourth field is marked as containing a pit.
• If no stench is percepted, the surrounding fields are marked as not housing the wumpus. If no breeze is percepted, the surrounding fields are marked as not containing a pit.
• Furthermore updateField checks if a piece of gold was taken if the last round's action was a grab. If so it knows that it may leave the field if it reaches the exit.

The basic strategy for generating a new action

The agent decides for its next action by combining the percepts for the current round with his knowledge about the environment.

• If the agent percepts a glitter, it grabs for the gold indicated thereby.
• If the agent reaches the exit and has already found gold, it leaves the cave.
• If the agent reaches the exit and a distinctive number of moves have passed, it leaves the cave whether or not he carries any gold.
• If the agent knows that the Wumpus is in any of the adjacent fields, it will shoot the wumpus if it has at least one arrow left.
• If none of the above action is taken the agent moves to the next field that is save and has not yet been visited.
• If the agent can choose from more than one save field it moves right, up, left, down, respectively in the case of an array-based agent or left, down, right up, respectively in the case of a hash-list-based agent.
• If no save field that has not yet been visited exists the agent moves to the surrounding save field that has been visited the fewest times.

A demonstration shows how the ArrayWumpus and the HashWumpus act in the same fashion with only choosing different directions.

Navigation:

• Back to the agent intro.
• Go to the Wumpus main page

Last modified: Sa. June 06 14:34:54 1997 by Christian Schwaiger

email to: webmaster@kr.tuwien.ac.at