Lecture_11_-_Decision_Trees_-_State_Machines_ppt

CSD1130 GAME Implementation Techniques

Lecture 11: Decision Trees & State Machines

Overview

  • Decision Making

    • Decision Trees

    • State Machines

      • Finite State Machines (FSM)

      • Hierarchical State Machines

  • FSM Scripting Language

  • AI Behavior in Platformer Outline

Decision Making

Knowledge Types

  • Internal Knowledge: Knowledge possessed by a character.

  • External Knowledge: Information external to the character's perception.

Process Model

  • Decision Maker: Absorbs knowledge.

  • Input Changes: Internal or external changes can alter the decision context.

  • Action Request: Translates knowledge and changes into specific actions.

Decision Trees

Basic Concept

  • A Decision Tree is a conditional structure consisting of a series of yes/no questions leading to a decision based on the current situation.

  • Each path in the tree represents potential actions (e.g., attack, hide, patrol).

Structure Example

  • Questions guiding decision making:

    • Is enemy visible?

      • If Yes: Check if enemy < 10 m.

        • If Yes: Attack.

        • If No: Move.

      • If No: Stand still.

Implementation of Decision Trees

Node Structure in C++

struct NodeTree {
   void * data;  // or int, or object
   NodeTree * left;
   NodeTree * right;
   virtual NodeTree * Execute() = 0;
   void MakeDecision();
};

Binary Decision Tree

  • Structure resembles a doubly linked list for managing states.

Example of Decision Logic

C++-Style Decision Logic

struct DecisionTreeNode {
   void * data;
   DecisionTreeNode * left;
   DecisionTreeNode * right;
   DecisionTreeNode * Execute();
   void MakeDecision();
};

  • Execution Logic: Decision-making function performs actions based on the object state.

Combinations of Decisions

  • Using logical operations in decision-making:f

    • Example:

      • If A AND B: Action 1

      • Otherwise: Action 2

      • If A OR B: Action 1

      • Otherwise: Action 2

Balancing Decision Trees

  • Unbalanced Tree vs. Balanced Tree:

    • Unbalanced trees can lead to inefficient decision paths.

    • Balancing improves performance and responsiveness.

Random Decision Trees

  • Decision nodes can introduce randomness:

    • Example:

      • If under attack: Defend

      • Else: Patrol or Stand Still.

Finite State Machines (FSM)

Concept

  • Visual representation of different states a character can be in during gameplay.

  • States include:

    • Climbing

    • Running

    • Swimming

    • Hitting

State Transitioning

  • State Manager: Tracks current states and transitions triggered by conditions.

  • Each update checks for eligible transitions.

Implementation of FSM

Generic Structure

  • Basic FSM class structure managing states and transitions:

class FSM {
   m_states;   // Holds states
   m_initialState;
   m_currentState;
   Update();   // Checks transitions
};

Hard-Coded Examples

  • Example illustrating hard-coded state transitions based on character conditions:

class FSM {
   enum State { PATROL, DEFEND, SLEEP };
   Update() {
       if (myState == PATROL) { ... }
   }
};

Pros and Cons of FSM

Pros

  • Easy to implement for simple behaviors.

Cons

  • Difficult to maintain as complexity increases.

FSM with Macros

Structure

  • FSMs can leverage macros for readability and maintainability:

bool FSM::States(...) {
   BeginStateMachine
   State(...) { ... }
   EndStateMachine
};

Hierarchical State Machines

Concept

  • Enhances FSM with layered states, managing states within other states for complex behaviors.

Conclusion and References

  • Essential Reading:

    • "Artificial Intelligence for Games" by Ian Millington

    • "AI Game Programming Wisdom 2" by Steve Rabin


Decision Diagrams

Overview: Decision diagrams are graphical representations used to illustrate decision-making processes. They consist of nodes and arcs that connect the nodes, helping to visualize the flow of decisions.

Components of Decision Diagrams:

  • Nodes: Represent decisions, conditions, or actions.

  • Arcs: Connect nodes, indicating the direction and flow of decision-making.

  • Leaves: Endpoints of the diagram, representing final actions or outcomes.

Types:

  1. Decision Trees: A type of decision diagram where each node is a decision point leading to various outcomes based on yes/no questions.

  2. Influence Diagrams: Show the influence relationships among decisions, outcomes, and uncertainties, usually incorporating probabilities.

Applications:

  • In AI, decision diagrams can be employed for modeling complex decision-making scenarios, ensuring clarity and methodical structuring of choices and their consequences.

Example: In gaming, decision diagrams can guide a character’s behavior based on player inputs or environment changes, effectively managing state transitions.