Rational Agent

An agent is an entity that perceives an Environment via sensors and acts on it with actuators (changing the Environment). We assume that agents can always perceive their own actions but maybe not the consequences of the actions.

A rational agent is not perfect as it only optimizes the expectation. If the agent does not know everything it cant optimize the actual Performance Measure. A rational agent might not perceive everything, thus exploration (learning, autonomy) might be needed in order to find hidden dangers. A rational agent might not be successfull as action outcomes may not be as expected (very unlikely events can still happen) → still take actions to learn.

Summary of why a rational agent is not perfect

• rational is not omiscient
• rational is not clairvoyant (hellseherisch)
• rational is not successfull

A rational agent exhibits Rational Behavior for any given class of environments and tasks. Want to find the optimal/best rational agent.

Autonomy

Percept Action Agent Function, Agent Program, Agent Architecture Performance Measure

PEAS

For example:

• humans
• robots
• softbots
• thermostats

Example: Vacuum-Cleaner World and Agent

Science Question: What is the Agent Function? AI Question: What is the Agent Program and Agent Architecture?

Agent Program for a vacuum cleaner might look like this:

procedure Reflex−Vacuum−Agent [location,status] returns an action
	if status = Dirty then return Suck
	else if location = A then return Right
	else if location = B then return Left

Basically Table-Driven Agent