Classical AI Examples

This repository provides practical examples illustrating key concepts in classical AI (also known as symbolic AI or knowledge-based AI), such as Markov Decision Processes, Constraint Satisfaction Problems, Rational Agents, and Propositional Logic.

📚 Topics covered include:

• Reinforcement Learning – Markov Decision Processes and policy iteration
• Constraint Satisfaction – Backtracking and constraint propagation
• Rational Agents – Goal-directed behavior and resource management
• Knowledge Representation – Propositional logic and inference

📁 Repository Contents

🤖 AgentMDP.java

• Concept: Demonstrates how to find optimal policies for agents in uncertain environments
• Algorithm: Policy iteration (alternating policy evaluation and improvement)
• Environment: 3x3 grid world with obstacles, fire (negative reward), and goal (positive reward)

🤖 AgentMDP_with_GUI.py

• Concept: Demonstrates how to find optimal policies for agents in uncertain environments with visual feedback
• Algorithm: Value iteration (iteratively updating state values until convergence)
• Environment: 4x4 grid world with obstacles, fire (negative reward), and goal (positive reward)
• Visualization: Interactive turtle graphics showing the agent's path, grid states, and outcome messages

🎨 BacktrackingCSP.java

• Concept: Solves the graph coloring problem using backtracking
• Algorithm: Backtracking with constraint propagation
• Problem: Assign colors to vertices such that adjacent vertices have different colors

🧹 rational_cleaner_agent.py

• Concept: Demonstrates rational agent behavior in a cleaning environment
• Model: PEAS (Performance, Environment, Actuators, Sensors) model
• Environment: 4x4 grid with dirt locations, energy constraints, and bag capacity

🔍 WeatherLogicTruthTable.java

• Concept: Demonstrates logical reasoning using truth tables
• Operations: Conjunction (AND), Implication (IF-THEN)
• Analysis: Evaluates complex logical expressions and knowledge bases

🔧 Customization

Modifying the MDP Environment

• Change the grid layout in AgentMDP.java
• Adjust rewards and discount factor
• Add new obstacles or goals

Extending the CSP Solver

• Modify color constraints
• Add new constraint types
• Implement different variable ordering heuristics

Enhancing the Rational Agent

• Adjust energy consumption rates
• Modify bag capacity constraints
• Add new environmental obstacles

🧠 Classical AI vs. Modern AI

Classical AI differs significantly from modern AI that uses machine learning approaches:

Classical AI (This Repository):

• Symbolic Reasoning: Uses logic, rules, and knowledge representation
• Explicit Knowledge: Knowledge is explicitly programmed and interpretable
• Deterministic: Predictable behavior based on logical rules
• Problem-Solving: Focuses on search, planning, and constraint satisfaction
• Interpretable: You can understand exactly how decisions are made

Modern AI (Machine Learning):

• Statistical Learning: Learns patterns from data
• Implicit Knowledge: Knowledge is learned and often not interpretable
• Probabilistic: Makes decisions based on statistical patterns
• Data-Driven: Requires large datasets for training
• Black Box: Often difficult to understand how decisions are made

📖 References

• Poole & Mackworth: Artificial Intelligence: Foundations of Computational Agents, 3rd Edition - Comprehensive online textbook covering all concepts in this repository

📄 License

This project is part of coursework and is intended for educational purposes.