Scavenger World (Final Project)

Scavenger World

Create a rational agent program to explore a world, collecting interesting objects. The agent program should use a suitable combination of search and logic techniques.

Performance Measure
The agent program will be evaluated based on the total value of objects returned to base.

All agent programs will be evaluated in identical worlds. Each agent program total value will be divided by an ideal score for the world. This percentage will be multiplied by 150 to calculate the points awarded for this project.

Agents must QUIT to receive a score. The agent need not be at the base.

Environment
The agent is an autonomous rover vehicle exploring the surface of an alien world. It has an arm that is capable of picking up objects and a bin to store objects. The rover has a base where all objects need to be deposited for analysis.

The bin can store up to 100. volume units. Objects pack neatly in the bin.

The agent may only recharge and deposit at its own base. Recharging at another agent's base will cause electrical damage to the rover. Objects that have been deposited into a base for analysis are unrecoverable. Bases are indestructable.

#          title damage energy traversable is_fall

Interface: plain      0      0           1       0

Interface: rocks      1      0           1       0

Interface: mud        0      1           1       0

Interface: wall      10      1           0       0

Interface: cliff      0      0           1       1

• The world is partially observable. The agent must explore to discover the layout of the land and the location of objects. The agent will be taught by an expert regarding the objects that might be encountered.
• The world is static. Objects only move if an agent rover acts upon them.
• The exploration of the world is sequential. The agent may learn more about the world as it explores. Each world exploration is episodic in nature.
• The environment is mostly deterministic.
• The environment is discrete in time and space.
• The evaluation of agents will be in single-agent environments.

Actuators
The agent has the following actions:

• GO_NORTH
• GO_SOUTH
• GO_EAST
• GO_WEST Horizontal movement costs 1 charge unit per square. Every meter in vertical climb costs 0.001 charge unit. Vertical drops add 0.001 charge unit. Falling off cliffs does not recharge the battery. Charge cost is not impacted by the mass of objects in the hopper.
• LOOK direction
• RECHARGE
• PICKUP obj-id-str
• DROP obj-id-str
• DEPOSIT obj-id-str
• EXAMINE obj-id-str
• QUIT
• NOOP

Sensors
The agent receives the following percepts

• CHARGE number
• HP number
• X_LOC number
• Y_LOC number
• Z_LOC number
• LOOK interface-description
• EXAMINE obj-id-str color-str shape-str size-str
• HOPPER obj-id-str
• BASE base-id-str
• BASE_HOPPER obj-id-str
• OBJECT obj-id-str

Suggested points of interest

• A list of expected objects can be found here.