Intelligent Agents

Introduction


An agent can be anything that can be viewed as perceiving its environment through sensors and acting upon that environment through effectors.

For example a human agent has eyes, ears and other organs for sensors and hands, legs, mouth and other organs for effectors. A robotic agent substitutes cameras and infrared range finders for the sensors and various motors for effectors. A software agent has encoded bit strings as its percepts and actions.

An agent program maps from a percept to an action, while updating the internal state.

Figure: Agensts interact with environments through sensors and effectors.





















Rational Agent

A rational agent is an agent that does the "right" thing. As an approximation we can say that the right action is the one that will cause the agent to be most successful. That leaves us with the problem of deciding how and when to evaluate an agent's success.

Performance Measure

Performance measure is the criteria that determines how successful an agent is. Obviously there is no  one fixed measure suitable for all agents. For example, consider an agent that is supposed to vacuum a dirty floor. Performance measure here, would be the amount of dirt cleaned up in a single eight hour shift. Amount of electricity consumption, for instance, can also be included in the performance measure as a penalty. Another consideration for the performance measure in this case would be the noise generated.

Ideal Rational Agent

For each percept sequence, an ideal rational agent should perform whatever the action that is expected to maximize the performance measure, on the basis of the evidence provided by the percept sequence and whatever build in knowledge the agent has.

Autonomy

If the agent's actions are based completely on built-in knowledge, such that it needs to pay no attention to it's percepts, then we say that the agent lacks autonomy.

For example, if the clock manufacturer was was prescient enough to know that the clock owner would be going to some other country at a particular date, then a mechanism could be built in to adjust the clock (the agent) by the time difference at the right time. This would certainly be a successful behavior, but the intelligence seems to be belong to the clock's designer than the clock, the agent, itself.

Exapmle:

Agent type     Percetps           Actions          Goals                     Enviroment

Taxi Driver      Cameras,            Steer,             Safe, Fast, Legal,     Roads,
                    Speedometer,      Accelerate,      Comfortable trip,      Pedestrians,
                    Sonar,                Brake,             Maximize profit        Other traffic,
                    Microphone,                                                          Customers

Now we have to decide how to build a real program to implement the mapping from percepts to action. We will find that the different aspects of driving suggest different types of agent program. We will consider four types of agent programs:

Figure: Simple reflex agents - Respond immediately to percepts.
























Figure: Agents that keep track of the world.























Figure: Goal-based agents - Acts so that they will achieve their goal.


























FigureUtility-based agents - Try to maximize their own happiness.