Intelligent Agents Introduction Rationality Nature of the Environment Structure of Agents Summary

Introduction What is an agent? a.Perceives environment through sensors (cameras, keystrokes, etc.) percept: single input percept sequence: sequence of inputs b. Acts upon environment through actuators (motors, displays info.)

Figure 2.1

Agent Functions and Programs Function.- Maps a percept sequence into an action. Program.- Actual implementation of the function. Example: Vacuum cleaner

Intelligent Agents Introduction Rationality Nature of the Environment Structure of Agents Summary

Rationality Four elements to define rationality: Performance measure Agent’s prior knowledge Actions that can be performed Agent’s percept sequence A performance measure defines if the agent’s behavior is successful or not.

Definition A rational agent selects an action that maximizes its performance measure given a.The percept sequence b. Built-in knowledge Rational agents should be autonomous. (learn under incomplete knowledge).

Intelligent Agents Introduction Rationality Nature of the Environment Structure of Agents Summary

Nature of the Environment How do we define a task environment? PEAS: Performance Environment Actuators Sensors

Properties of Environments How do we define an environment? Fully or partially observable (Do sensors capture all relevant info.?) Deterministic or stochastic (Is next state completely determined by the current state and action?) Episodic or Sequential (Is experience divided into atomic episodes?)

Properties of Environments Static vs Dynamic (Can environment change while taking action?) Discrete vs Continuous (Is there a finite or infinite number of states?) Single Agent vs Multiagent (single-player game? or two-player game?)

Intelligent Agents Introduction Rationality Nature of the Environment Structure of Agents Summary

Structure of Agents Agent’s Program. Simplest Program: Keep track of percept sequences. Use percept-action table to decide on an action.

Limitations Let P be the no. of possible percepts Let T be the lifetime of the agent The lookup table would have size of Σ t=1 |P| t T

Simple Reflex Agents Select action based on current percept. Ignore the rest of the percept sequence. We need a set of condition-action rules

Model-based Reflex Agents The agent stores a representation of the percept history. We need two kinds of information: a.How the world evolves b.How actions affect the world (a) and (b) give us a model of the world

Goal-based Agents We need some goal pointing to desirable situations. We need to incorporate tasks like a.search b.planning Considers the future.

Utility-Based Agents Don’t just reach a goal. Generate high-quality behavior. We need a utility function mapping states to a real number.

Learning Agents A learning agent is divided in 4 components: 1.Performance Element (selects actions) 2.Learning Element (makes improvements) 3.Critic (how well the agent is doing) 4.Problem Generator (suggest new experiences)

Intelligent Agents Introduction Rationality Nature of the Environment Structure of Agents Summary

Agents perceive and act on environments. Agents need performance measures. Tasks environments include performance measures, environment, actuators, sensors. Types of agents: simple reflex, model-based, goal-based, and utility-based. Learning is used to improve performance.