drbeco / wumpus Goto Github PK

Agent may have lifes to spare.

The same labyrinth will be restarted,

• agent goes back to (1,1)
• Wumpus keep same state (dead or alive)
• If no arrows when dying, set it to 1. Otherwise, keeps the same amount.

The game will not call init_agent/0, but restart_agent/0.

Arrows can keep going and get back to you, killing you.
This is a new type of hazard.

Its important to to remember from the original game:

• You can say the squares you want the arrow to go
• If you make a mistake, the arrow goes random
• If the arrow manages to circle back to you, you die

Also, this is important only if the wumpus move when shooted, because then you may want to shoot at some distance.

And the second bat do nothing

Add bats

The original topology from Gregory Yob!

Give the opportunity to setup a random start cave for the Agent, and random start direction

Create a debug option to repeat the exact same map randomness.

If set to 0, makes a random sized grid NxN.

• Number of arrows set by world_setup

Where can I find the code for agent 001?

Change the cave's order to allow following from cave 1 to 20.

Some predicates regarding world_setup is incorrect.

Shooting without arrows should not wake up Wumpus

They just maximize the options, and sometimes the maximum is above the target.

(inverse also true: minimizing too much).

Old bug registered here.

After shoot, if it did not kill the wumpus, it will be found near de wall it hit.

No glitter, no perception.

A blind grab will do the magic.

Wumpus follow the agent when it smells him.

Change the angle of the wumpus when initializing.

Specially important in mode walker.

Sua opinião é muito importante para nós! 😄

Pessoal,

Gostaria de receber um retorno do que acharam do modo de jogo dodeca no manual.

Também se puderem, coloquem sugestões sobre cada item do jogo (Wumpus, Morcegos, Buracos, Ouro, Agente, Flechas, Vidas, Navegação, qualquer coisa). Comentários sobre a habilidade, a quantidade, alguma outra forma criativa de alterar o jogo para melhorá-lo, torná-lo mais desafiador no que estiver fácil ou mais fácil no que estiver muito complicado. Não prometo aceitar todas sugestões, pois tento deixar o jogo com o mínimo de modificações a partir do original, mas algumas alterações talvez façam sentido, e aí várias cabeças pensam melhor que uma.

O que achou do modo dodeca?

Tem alguma sugestão geral (para todos os modos)?

An example of him running:

http://sprunge.us/JceP

Create a benchmark to evaluate an agent performance.

Bugs number #18, #6 and #5 try to define arrows behavior.

This bug is going to summarize and to define what would happen.

From #18, reutilize the arrow: arrow will be found near the wall if not killed wumpus. No glitter, no perception. Just a blind grab will get it.

From #6, arrow can get you: you may say what squares the arrow can go. If you make a mistake, the arrow goes random. If the arrow circle back to you, you die. This is a new hazard. It only makes sense in the dodeca word.

From #5, you can set the starting number of arrows. The original game start with 5. Another option is to find an arrow.

Wumpus should move a bit

After actions exhausted, agent die and loses -500.

Create a new intelligent agent for the dodecahedron map

A more clean code.

Pits - number or probability
Gold - number or probability
Bat - number or probability
Wumpus - fixed 1? or number or probability? Lets see...

Give priority to the current angle

unless... facing (1,1), than it should avoid that angle.

Agent canot use any simulator predicates except (*)

• start/0

Use:
:- module(wumpus, [start/0]).

Update this list

• +1000 for each piece of gold after agent leaves the cave
• -500 if dead:
  1. Wumpus food
  2. Falling in a pit
  3. Starved to death (too many actions)
• -1 for each action:
  1. goforward
  2. turnright
  3. turnleft
  4. grab
  5. shoot
  6. sit
  7. climb

Use the same messages as the original from Gregory Yob