Q1. Scenario: Design a vacuum cleaner robot that cleans a room. Describe its sensors, actuators, environment, and performance measure. What type of agent is it?
Environment: room with dirt. Sensors: bump sensors, dirt detection. Actuators: wheels, suction motor. Performance measure: amount of dirt cleaned, time. It is a simple reflex agent (or model-based if it maps the room).
Q2. Scenario: A self-driving car's environment includes other cars, pedestrians, and traffic lights. Classify the environment as fully/partially observable, deterministic/stochastic, episodic/sequential, static/dynamic, discrete/continuous.
Partially observable (blind spots), stochastic (other drivers unpredictable), sequential (long-term decisions), dynamic (environment changes), continuous (positions, velocities). This complexity requires sophisticated planning.
Q3. Scenario: A thermostat maintains temperature. Is it a rational agent? What is its goal and performance measure?
Yes, it's a simple reflex agent. Goal: keep temperature within desired range. Performance measure: deviation from setpoint, energy consumption, response time. Rational action: turn heater on/off accordingly.
Q4. Scenario: You are building a trading bot for cryptocurrency. What kind of environment is this? How does the agent's perception affect its rationality?
Partially observable (dependence on hidden market sentiment), stochastic (price fluctuations), sequential (actions affect future prices), dynamic (fast changes). The agent must use historical data and real-time signals to act rationally (maximize profit).
Q5. Scenario: Explain the difference between a model-based and a model-free agent using a chess-playing AI example.
Model-based: predicts opponent moves using game rules and simulates outcomes (e.g., minimax). Model-free: learns a policy directly from self-play (e.g., AlphaZero's value network). Model-free adapts without explicit simulation.