Intelligent Agents (Introduction)
• An agent: perceives environment, makes
decisions, performs actions
– Goal: act intelligently to achieve objectives
– Used in: self-driving cars, chatbots, robots, smart
assistants
Why Agents Matter
• Agents form the foundation of AI systems
– They automate tasks and solve complex problems
– Used in fields like healthcare, transport, gaming,
finance, robotics
Characteristics of Intelligent Agents
• Autonomy: Operates without human control
– Reactivity: Responds quickly to environmental
changes
– Proactivity: Takes goal oriented actions
‑
– Social Ability: Communicates with humans/other
agents
Agent Interaction Cycle
• Perception: Collect data using sensors
– Reasoning: Process information and decide action
– Action: Execute using actuators
– Cycle repeats continuously for intelligent behavior
Understanding Rationality
• Rationality = choosing the best decision based
on available information
– Considers performance measure, knowledge,
history, and possible actions
– A rational agent aims for maximum expected
success
Factors Influencing Rationality
• Performance Measure: Defines success
criteria
– Percept Sequence: Everything sensed so far
– Knowledge: What the agent knows about
environment
– Available Actions: Options the agent can take
Types of Rationality
• Perfect Rationality: Always chooses the
optimal action
– Bounded Rationality: Makes good-enough
decisions with limitations
– Instrumental: Selects actions that best achieve
goals
– Epistemic: Builds accurate beliefs using evidence
– Practical: Balances goals, risks, and resources
Task Environment Properties (1)
• Fully vs Partially Observable: Complete vs
limited information
– Deterministic vs Stochastic: Predictable vs random
outcomes
– Episodic vs Sequential: Independent vs dependent
actions
Task Environment Properties (2)
• Static vs Dynamic: Environment changes or
stays still
– Discrete vs Continuous: Limited vs infinite states
and actions
– Single vs Multi-Agent: One vs multiple interacting
agents
PEAS Framework
• Performance Measure: Defines success for the
agent
– Environment: Where the agent operates
– Actuators: How the agent performs actions
– Sensors: How the agent gathers data
Types of Agents
• Simple Reflex: Acts based only on current
input
– Model-Based: Uses memory + world model
– Goal-Based: Chooses actions that achieve goals
– Utility-Based: Selects most beneficial outcome
– Learning Agent: Improves performance over time
Example: Taxi Driver Agent
• Partially observable: Can't see everything at
once
– Dynamic: Traffic keeps changing
– Multi-Agent: Other cars and pedestrians
– Continuous: Speed, road conditions, distances
– Sequential: Each decision affects future states
Example: Chess with Clock
• Fully observable: Board is fully visible
– Multi-agent: Two players compete
– Deterministic: No randomness, predictable moves
– Discrete: Limited number of moves
– Sequential: Each move affects future game state
Summary
• Agents sense, reason, and act to achieve goals
– Rationality = best action given information and
limits
– Task environments vary in observability, dynamics,
and structure
– Agent types range from simple reflex to advanced
learning agents

Presentation AI_Agents_Detailed_Slides.pptx

  • 1.
    Intelligent Agents (Introduction) •An agent: perceives environment, makes decisions, performs actions – Goal: act intelligently to achieve objectives – Used in: self-driving cars, chatbots, robots, smart assistants
  • 2.
    Why Agents Matter •Agents form the foundation of AI systems – They automate tasks and solve complex problems – Used in fields like healthcare, transport, gaming, finance, robotics
  • 3.
    Characteristics of IntelligentAgents • Autonomy: Operates without human control – Reactivity: Responds quickly to environmental changes – Proactivity: Takes goal oriented actions ‑ – Social Ability: Communicates with humans/other agents
  • 4.
    Agent Interaction Cycle •Perception: Collect data using sensors – Reasoning: Process information and decide action – Action: Execute using actuators – Cycle repeats continuously for intelligent behavior
  • 5.
    Understanding Rationality • Rationality= choosing the best decision based on available information – Considers performance measure, knowledge, history, and possible actions – A rational agent aims for maximum expected success
  • 6.
    Factors Influencing Rationality •Performance Measure: Defines success criteria – Percept Sequence: Everything sensed so far – Knowledge: What the agent knows about environment – Available Actions: Options the agent can take
  • 7.
    Types of Rationality •Perfect Rationality: Always chooses the optimal action – Bounded Rationality: Makes good-enough decisions with limitations – Instrumental: Selects actions that best achieve goals – Epistemic: Builds accurate beliefs using evidence – Practical: Balances goals, risks, and resources
  • 8.
    Task Environment Properties(1) • Fully vs Partially Observable: Complete vs limited information – Deterministic vs Stochastic: Predictable vs random outcomes – Episodic vs Sequential: Independent vs dependent actions
  • 9.
    Task Environment Properties(2) • Static vs Dynamic: Environment changes or stays still – Discrete vs Continuous: Limited vs infinite states and actions – Single vs Multi-Agent: One vs multiple interacting agents
  • 10.
    PEAS Framework • PerformanceMeasure: Defines success for the agent – Environment: Where the agent operates – Actuators: How the agent performs actions – Sensors: How the agent gathers data
  • 11.
    Types of Agents •Simple Reflex: Acts based only on current input – Model-Based: Uses memory + world model – Goal-Based: Chooses actions that achieve goals – Utility-Based: Selects most beneficial outcome – Learning Agent: Improves performance over time
  • 12.
    Example: Taxi DriverAgent • Partially observable: Can't see everything at once – Dynamic: Traffic keeps changing – Multi-Agent: Other cars and pedestrians – Continuous: Speed, road conditions, distances – Sequential: Each decision affects future states
  • 13.
    Example: Chess withClock • Fully observable: Board is fully visible – Multi-agent: Two players compete – Deterministic: No randomness, predictable moves – Discrete: Limited number of moves – Sequential: Each move affects future game state
  • 14.
    Summary • Agents sense,reason, and act to achieve goals – Rationality = best action given information and limits – Task environments vary in observability, dynamics, and structure – Agent types range from simple reflex to advanced learning agents