Artificial Intelligence Problem Solving What is problem Basically

Artificial Intelligence Problem Solving

What is problem ? Basically all problems/tasks are defined according to a formal definition A formal problem definition has four parts: Ø Initial State Ø Successor Function Ø Goal Test Ø Path Cost Classical example: Finding path in a graph

What is problem Solver? Problem Solver is an agent: Problem Solving Agent. It can find a sequence of actions that achieves its goal It is required where no single action can find solution

What is an agent? Perceives its environment Acts upon the environment Agent’s structure: Architecture + Agent program Architecture makes physical organs Agent program is implementation of agent function For examples: Human Robotic System Software Program

Agent (cont. …) Percept: Percept Sequence: Mapping from the precept sequence to an action Performance Measure of Agent: History of all percepts that an agent has perceived Agent Function: Agent’s perceptual inputs at a given instance Criteria which determines how successful an agent is Behavior of Agent: Action that agent performs after any given sequence of percepts

Agent (Cont. …) Example

Agent Kinds Agent kinds: Simple Reflex Model-based Goal-based Utility-based

Simple Reflex Chooses action based on only current percept Uses Condition-Action rules Needs fully observable environment

Model-Based Reflex Modified simple reflex Keeps internal state of the world Has to update the state How things happen in the world How actions affect the world

Goal-Based Needs description of desirable situations Chooses action looking at the goal More flexible as compared to reflex

Utility-Based Uses preferences/priorities (utility function) Utility function gives cost Tries to achieve optimized solution for goal

Environment Nature of Environment Properties of Environment Accessible vs. inaccessible. Deterministic vs. nondeterministic. Episodic vs. non-episodic. Static vs. dynamic. Discrete vs. continuous. Observable / Partially Observable.