5: Intelligent Agents (PEAS, Agent Types)

5: Intelligent Agents (PEAS, Agent Types)

An intelligent agent is a fundamental concept in artificial intelligence, representing any entity that perceives its environment through sensors and acts upon that environment through actuators. This Perceive-Think-Act cycle is the core of an agent's operation. The "think" component involves the agent's internal process for mapping perceptions to actions, which is defined by its agent function and implemented as its agent program.

A critical framework for understanding and designing agents is the PEAS model, which stands for Performance measure, Environment, Actuators, and Sensors. This model provides a structured way to specify the setting for an intelligent agent.

• Performance Measure: The objective standard for evaluating an agent's success. For a vacuum cleaner agent, this might be the amount of clean floor over time.
• Environment: The external world in which the agent operates. This includes everything the agent can perceive.
• Actuators: The mechanisms through which the agent exerts influence on its environment (e.g., a robotic arm, a steering wheel, a display screen).
• Sensors: The devices that provide the agent with information about the environment (e.g., cameras, microphones, lidar, keyboards).

Agents can be categorized into increasingly sophisticated types based on their capabilities:

1. Simple Reflex Agents: These are the most basic agents. They act based on the current percept, ignoring percept history. Their actions are determined by condition-action rules (if-then rules). They are efficient but fail in partially observable environments.

2. Model-Based Reflex Agents: To overcome the limitations of simple agents, these agents maintain an internal state that models the unobserved aspects of the world. This internal model is updated based on percept history, allowing the agent to operate effectively in partially observable environments.

3. Goal-Based Agents: These agents expand on the model-based design by incorporating goal information. Their decision-making is driven by achieving a specific goal, considering future actions and their outcomes. This allows them to choose actions that are beneficial in the long term.

4. Utility-Based Agents: Goals alone can be insufficient; sometimes there are multiple ways to achieve a goal, with some being better than others. A utility-based agent uses an internal utility function (a measure of happiness or success) to compare different states. It chooses the action that maximizes its expected utility, making it capable of making nuanced trade-offs.