8: Building a Knowledge-Based Agent

A knowledge-based agent operates by maintaining a knowledge base (KB) and using inference to make decisions. This chapter outlines the architecture, processes, and applications of such agents.

8.1 Architecture of a Knowledge-Based Agent

The architecture of a knowledge-based agent includes:

1. Knowledge Base (KB): A repository of knowledge in the form of logical sentences.

2. Inference Engine: A system that derives conclusions or decisions based on the KB.

8.1.1 Knowledge Base (KB)

The KB contains:

• Facts: Statements about the world. Example: "The sky is blue."

• Rules: Logical relationships between facts. Example: "If it is raining, the ground is wet."

8.1.2 Inference Engine

The inference engine performs reasoning by:

1. Applying rules of logic (e.g., Modus Ponens).

2. Deriving new facts from existing ones.

3. Testing the truth of queries.

---

8.2 The Agent's Cycle

A knowledge-based agent operates in a continuous loop:

1. Perceive: Gather information about the environment through sensors.

2. Update KB: Add new information to the KB.

3. Reason: Use inference to deduce actions or responses.

4. Act: Execute actions through actuators.

Example:

• Environment: A robot operating in a warehouse.

• Perception: "A package is at location A."

• KB Update: Add "Package(A)" to the KB.

• Reasoning: "If a package is at location A, move to location A."

• Action: The robot navigates to location A.

---

8.3 Knowledge Representation

8.3.1 Declarative Knowledge

Facts and rules explicitly stated in the KB. Example: "All humans are mortal: ∀x Human(x)→Mortal(x)\forall x \, Human(x) \rightarrow Mortal(x)∀xHuman(x)→Mortal(x)."

8.3.2 Procedural Knowledge

Knowledge about how to perform tasks, represented as algorithms or procedures. Example: "To move to location A, navigate along the shortest path."

---

8.4 Logical Inference

Logical inference is the process of deriving new knowledge from existing facts and rules.

8.4.1 Forward Chaining

• Starts with known facts and applies rules to derive new facts. Example:

1. Fact: "It is raining."

2. Rule: "If it is raining, the ground is wet."

3. Conclusion: "The ground is wet."

8.4.2 Backward Chaining

• Starts with a goal and works backward to determine if the goal can be satisfied. Example:

1. Goal: "Is the ground wet?"

2. Check Rule: "If it is raining, the ground is wet."

3. Verify Fact: "It is raining."

4. Conclusion: "Yes, the ground is wet."

---

8.5 Handling Uncertainty

Not all environments provide complete or reliable information. Knowledge-based agents use probabilistic reasoning to manage uncertainty.

8.5.1 Probabilistic Logic

Incorporates probabilities into logical reasoning. Example: "There is a 70% chance that it will rain tomorrow."

8.5.2 Bayesian Networks

Graphical models that represent probabilistic relationships between variables. Example: Predicting the likelihood of a traffic jam based on weather and time of day.

---

8.6 Applications of Knowledge-Based Agents

8.6.1 Automated Customer Support

Agents answer queries by reasoning about a knowledge base of FAQs and support documentation.

8.6.2 Diagnostic Systems

Medical systems use knowledge bases to infer diseases based on symptoms. Example: "If the patient has a fever and cough, they may have the flu."

8.6.3 Robotics

Robots use logical reasoning to plan and execute tasks in dynamic environments. Example: A delivery robot navigating a building to deliver packages.

---

8.7 Building a Knowledge-Based Agent: Example

Problem: Room Cleaning Robot

1. Environment: A robot operates in a 3x3 grid with dirty and clean tiles.

2. Objective: Clean all tiles while minimizing energy consumption.

Steps:

1. Knowledge Base:

   • Facts: "Tile A1 is dirty."

   • Rules: "If a tile is dirty, clean it."

2. Inference:

   • Forward chaining: "Tile A1 is dirty → Move to A1 → Clean A1."

3. Action:

   • The robot navigates to A1 and cleans the tile.

---

8.8 Summary

In this chapter, we explored:

1. The architecture and operation of knowledge-based agents.

2. Techniques for knowledge representation and logical inference.

3. Applications in areas like customer support, diagnostics, and robotics.

4. A practical example of building a room-cleaning robot.