Intelligent Agents Overview

• Course Name: Artificial Intelligence
• Lecturer: Amir EL-Ghamry
• Topic: Intelligent Agents

Intelligent Agents and Environments

• Agents are anything that perceives its environment through sensors and acts upon that environment through actuators.
• An agent program runs in cycles of: perceive, think, and act.
• An agent is composed of architecture and program.

PEAS (Performance measure, Environment, Actuators, Sensors)

• PEAS defines the task environment.
• Performance measure, Environment, Actuators, and Sensors are critical components to define the task environment.

Agent Examples

• Human agent: Eyes, ears, and other body parts are sensors; hands, legs, mouth, and other body parts are actuators.
• Robotic agent: Cameras and infrared range finders are sensors; various motors are actuators.
• Software agent: Keystrokes, file contents, and received network packages are sensors; displays on the screen, files, and sent network packets are actuators.

Agents vs. Other Software

• Agents are autonomous, acting on the user's behalf.
• Agents have intelligence, adapting to environment changes (fixed rules to learning engines).
• Agents have social ability, communicating with users and systems.
• Agents can cooperate with others for complex tasks.
• Agents can migrate between systems to access resources or meet other agents.

Agent and Environment

• Percept: agent's perceptual input at any given moment.
• Percept sequence: complete history of all agent perceptions.
• An agent's choice of action depends on the entire percept sequence.
• Agent function: maps from percept histories to actions (f: P* → A).
• The agent function is implemented by an agent program.

Structure of Intelligent Agents

• Agent program implements the agent's perception-action mapping. A function Skeleton-Agent(Percept) returns Action (includes steps for updating memory and choosing the best action based on memory).
• Architecture is the device that runs the agent program (e.g., general-purpose computer, specialized device).

Vacuum-cleaner world

• Percepts: location (A or B) and contents (dirt or not), e.g., [A, Dirty].
• Actions: Left, Right, Suck, NoOp.
• Agent's function: a look-up table (for each possible percept sequence maps to an action). This can be a very large table for many agents.
• Example function Vacuum-Agent([location,status]) returns an action. If the status is Dirty then return Suck. If the location is A then return Right. If the location is B then return Left.

Agent Function – Lookup table

• A trivial agent program: Keeps track of the percept sequence and uses it to access a look-up table to determine the appropriate action.
• Drawbacks of look-up tables: huge, time-consuming to build, no autonomy, even learning requires significant time to learn entries.

Rational Agent

• An agent should strive to do the right thing based on perception and actions.
• Right action: one most likely to cause success.
• Performance measure: An objective criterion for an agent's success (e.g., amount of dirt cleaned, time taken, electricity consumed, generated noise).

Rationality - Good Behavior

• Performance measuring success.
• Agents prior knowledge of environment.
• Actions the agent can perform.
• Agent's percept sequence to date.
• Rational Agent: For each possible percept sequence, maximizes agent's performance measure.

Autonomy in Agents

• Agent autonomy is the extent to which its behavior is determined by experience, not designer knowledge.
• Extremes: No autonomy (ignores environment/data), Complete autonomy (must act randomly).
• Example: baby learning to crawl.
• Ideal: design agents with some autonomy, potentially enhanced with experience.

Specifying the Task Environment (PEAS)

• PEAS: Performance measure, Environment, Actuators, and Sensors.
• Critical first step in designing an agent.

PEAS examples

• (Vacuum cleaner, Windshield Wiper, Self-driving car, Automated taxi driver, Medical diagnosis system, Spam filter, Satellite image analysis system, Part-picking robot, Interactive English tutor) Detailed descriptions of the PEAS characteristics for each example are provided in the pages.

Interacting Agents (Collision Avoidance Agent, Lane Keeping Agent)

• Agents that operate simultaneously and interact with each other, such as those in a car.
• Example: (Collision Avoidance Agent, Lane Keeping Agent, Conflict Resolution) Details of these specific types are provided in the pages.