Agent, PEAS, and Types of Environment

1. Definition of Agent (2 marks)

An agent is anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators.

"An agent is an entity that perceives its environment through sensors and acts upon that environment through actuators." -- Russell & Norvig

• A human agent has eyes, ears (sensors) and hands, legs, mouth (actuators).
• A software agent (robot, game-playing program) has file inputs, network packets (sensors) and screen output, motors (actuators).

The agent's behavior is described by an agent function that maps any given percept sequence to an action:

f : P* --> A

where P* is the set of all possible percept sequences and A is the set of actions.

2. PEAS Description (3 marks)

PEAS stands for:

PEAS is used to specify the task environment of an agent before designing it.

Example 1: Automated Taxi Driver

Example 2: Medical Diagnosis System

3. Types of Environment (5 marks)

The environment in which an agent operates can be classified along several dimensions:

3.1 Fully Observable vs. Partially Observable

• Fully Observable: The agent's sensors give access to the complete state of the environment at each point in time.
  • Example: Chess game (entire board is visible)
• Partially Observable: The agent cannot observe the full state due to noisy or limited sensors.
  • Example: Poker game (opponent's cards are hidden), self-driving car (blind spots)

3.2 Single Agent vs. Multi-Agent

• Single Agent: Only one agent operates in the environment.
  • Example: Crossword puzzle solver
• Multi-Agent: Multiple agents interact in the same environment.
  • Competitive: Agents compete (e.g., Chess -- one agent wins, other loses)
  • Cooperative: Agents work together (e.g., multi-robot warehouse system)

3.3 Deterministic vs. Stochastic

• Deterministic: The next state of the environment is completely determined by the current state and the agent's action. No uncertainty.
  • Example: Chess
• Stochastic: The next state is not fully determined; there is uncertainty.
  • Example: Taxi driving (uncertain traffic), card games

Note: If the environment is deterministic except for the actions of other agents, it is called strategic.

3.4 Episodic vs. Sequential

• Episodic: The agent's experience is divided into independent episodes. Each episode consists of perceiving and acting; the next episode does not depend on previous ones.
  • Example: Spam email classifier (each email is independent)
• Sequential: The current decision affects future decisions. Past actions matter.
  • Example: Chess, taxi driving

3.5 Static vs. Dynamic

• Static: The environment does not change while the agent is deliberating.
  • Example: Crossword puzzle
• Dynamic: The environment can change while the agent is thinking.
  • Example: Taxi driving, stock market
• Semi-dynamic: The environment does not change but the agent's performance score changes with time.
  • Example: Chess with a clock

3.6 Discrete vs. Continuous

• Discrete: A finite number of distinct states, percepts, and actions.
  • Example: Chess (finite board positions)
• Continuous: States, time, and actions are continuous.
  • Example: Self-driving car (continuous speed, position)

3.7 Known vs. Unknown

• Known: The agent knows the rules/outcomes of all actions.
  • Example: Chess (rules are known)
• Unknown: The agent does not know how the environment works and must learn.
  • Example: A new video game with unknown rules

Summary Table

Note: The real world is generally partially observable, stochastic, sequential, dynamic, continuous, and multi-agent -- making it the most challenging type of environment for an AI agent.

Backward Chaining vs Forward Chaining

Forward Chaining (Data-Driven Reasoning)

Forward chaining starts from known facts and applies inference rules to derive new facts, moving forward toward the goal.

• It is a bottom-up approach
• Begins with available data/facts
• Fires rules whose conditions are satisfied
• Continues until the goal is reached or no more rules can fire
• Used when: all data is available and we want to derive conclusions

Backward Chaining (Goal-Driven Reasoning)

Backward chaining starts from the goal and works backward to find the facts that support it.

• It is a top-down approach
• Begins with the goal to be proved
• Breaks the goal into sub-goals
• Continues until all sub-goals are satisfied by known facts
• Used when: the goal is clear and we want to verify it

Key Differences

Example

Given Rules and Facts:

• Fact 1: It is raining
• Fact 2: I have an umbrella
• Rule 1: If it is raining AND I have an umbrella → I will not get wet
• Rule 2: If I will not get wet → I will go outside
• Goal: Will I go outside?

Forward Chaining Solution:

Start with known facts: {raining, have umbrella}

Step 1: Rule 1 fires (raining + have umbrella)
        → Derive: "will not get wet"

Step 2: Rule 2 fires (will not get wet)
        → Derive: "will go outside"

Goal REACHED: Yes, I will go outside.

Backward Chaining Solution:

Start with Goal: "Will I go outside?"

Step 1: To prove "go outside", need Rule 2
        → Sub-goal: Prove "will not get wet"

Step 2: To prove "will not get wet", need Rule 1
        → Sub-goals: Prove "raining" AND "have umbrella"

Step 3: Check facts:
        "raining" → TRUE (known fact)
        "have umbrella" → TRUE (known fact)

Goal PROVED: Yes, I will go outside.

Summary

Forward chaining is suitable when we have a set of facts and want to derive all possible conclusions. Backward chaining is suitable when we have a specific goal and want to verify whether it can be supported by existing facts. Both are fundamental inference strategies used in expert systems and AI rule-based reasoning.

Game and State Space Representation

Definition of Game (2 marks)

A game in the context of Artificial Intelligence is a type of search problem that involves two or more competing agents (players) who take turns making moves, each trying to achieve their own goal (usually to win) while the opponent tries to prevent it.

Key characteristics of a game:

• There are two players (usually called MAX and MIN)
• Players take turns alternately
• The game has a terminal state (win, lose, or draw)
• Each player tries to maximize their own utility and minimize the opponent's utility
• The outcome depends on the actions of both players

Examples: Chess, Tic-Tac-Toe, Checkers, Go

State Space Representation (3 marks)

A state space is a way of representing a problem as a set of states, operators (actions), an initial state, and a goal state. It forms a graph where nodes represent states and edges represent transitions.

Components of State Space Representation:

Example: 8-Puzzle Problem

The 8-Puzzle consists of a 3x3 grid with 8 numbered tiles and one blank space. The goal is to reach a target arrangement by sliding tiles.

Initial State:

2 | 8 | 3
---------
1 | 6 | 4
---------
7 |   | 5

Goal State:

1 | 2 | 3
---------
8 |   | 4
---------
7 | 6 | 5

State Space Representation:

• State: The arrangement of all 8 tiles and the blank in the 3x3 grid
• Initial State: The given starting arrangement
• Operators: Move blank UP, DOWN, LEFT, or RIGHT
• Goal State: The target arrangement of tiles
• Path Cost: Each move costs 1 (total number of moves)

Partial State Space Tree:

        [2,8,3,1,6,4,7,_,5]   <- Initial State
              |
    __________|__________
    |                   |
Move LEFT           Move UP
    |                   |
[2,8,3,1,6,4,7,5,_] [2,8,3,1,_,4,7,6,5]
                         |
                    Move LEFT / RIGHT ...
                         |
                   ... (Goal State)

Total number of possible states = 9! / 2 = 181,440 reachable states

This state space is searched using algorithms like BFS, DFS, A*, or Hill Climbing to find the path from the initial state to the goal state.