Intelligent Agents Overview

Intelligent Agents

• Agents are entities that perceive and act on their environment
• An agent's behavior is described by an agent function
• Agent = Architecture + Program
• An agent program runs in cycles of: perceive, think, and act
• Agent programs map percept histories to actions

Agent Function

• Maps percept histories to actions
• Formally represented as f: P* → A

Structure of Intelligent Agents

• Agent program: the implementation of the agent's perception-action mapping

• Function Skeleton-Agent(Percept) that returns an Action

  • memory ← UpdateMemory(memory, Percept)
  • Action ← ChooseBestAction(memory)
  • memory ← UpdateMemory(memory, Action)
  • return Action

• Architecture: a device capable of executing the agent program (e.g., computer)

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: lookup table
• Lookup table is very large for many agents

Vacuum-cleaner Agent Function

• function 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

Agent Function – Lookup Table

• A trivial agent program tracks the percept sequence to index into a table and then choose an action
• The designers create a table with the appropriate action for every percept sequence
• Drawbacks:
  • Huge table (P^T), P: set of possible percepts, T: lifetime
  • Space to store the table
  • Table takes a long time to build
  • Limited autonomy

Rational Agent

• Strives to "do the right thing" based on perception and actions
• Right action maximizes the agent's success
• Performance measure: Objective criterion for agent's success.

Rationality

• Performance measuring success
• Agents prior knowledge of environment
• Actions that agent can perform
• Agent's percept sequence to date
• Rational Agent: For each possible percept sequence, a rational agent should select an action that is expected to maximize its performance measure, given the evidence provided by the percept sequence and whatever built-in knowledge the agent has.
• Rational is different from omniscience (all knowing with infinite knowledge)
  • Percepts may not supply all relevant information.
  • E.g., in card game, don't know cards of others.
  • Rational is different from being perfect
    • Rationality maximizes expected outcome.
    • Perfection (omniscience) maximizes actual outcome.

Back to Vacuum Cleaner Agent

• Is this agent rational?
  • Depends on performance measure and environment properties
• Performance measure Awards one point for each clean square in a 1,000 time-step lifetime
• Geography of the environment is known a priori
• Dirt distribution and initial location are not known
  • agent correctly perceives its location, and whether that location contains dirt
• Under these circumstances the agent is rational: its expected performance is at least as high as any other agent.

Vacuum Cleaner Agent – Irrational

• Same agent would be irrational under different circumstances
  • Once all dirt is cleaned up, it oscillates needlessly

Autonomy in Agents

• The autonomy of an agent is determined by its own experience, rather than designer knowledge
• Extremes:
  • No autonomy—ignores environment/data
  • Complete autonomy—must act randomly/no program
  • Example—baby learning to crawl.

Specifying Task Environment (PEAS)

• Performance measure, Environment, Actuators, Sensors (PEAS)
• The first step in designing an agent is specifying the task environment as fully as possible

PEAS – Examples

• Vacuum Cleaner
• Automated Taxi Driver
• Medical Diagnosis System
• Spam Filter

Interacting Agents

• Agents can interact with each other and resolve conflicts.

Collision Avoidance Agent (CAA)

• Goals: Avoid collisions with obstacles
• Percepts: Obstacle distance, velocity, trajectory
• Sensors: Vision, proximity sensing
• Actuators: Steering wheel, accelerator, brakes, horn, headlights
• Actions: Steer, speed up, brake, blow horn, signal (headlights)
• Environment: Freeway

Lane Keeping Agent (LKA)

• Goals: Stay in current lane
• Percepts: Lane center, lane boundaries
• Sensors: Vision
• Actuators: Steering wheel, accelerator, brakes
• Actions: Steer, speed up, brake
• Environment: Freeway

Conflict Resolution by Action Selection Agents

• Arbitrate:
  • If Obstacle is Close then CAA, else LKA
• Challenges: Doing the right thing