Please enable JavaScript.

Coggle requires JavaScript to display documents.

Agent image - Coggle Diagram

- - - - Cameras
      - Infrared Sensors
    - - Motors
  - - - Keyboard
      - Data Packet
    - - Screen display
      - Data Transmission
      - File write
  - - - Ears
      - Organs
      - Eyes
    - - Leg
      - Mouth
      - Hand
- - - - Environment: Roads, traffic.
      - Actuators: Steering, accelerator, brakes.
      - Performance: Safety, speed, fuel efficiency.
      - Sensors: Cameras, GPS, speedometer.
    - - Performance: Diagnosis accuracy, patient safety
      - Environment: Patients, medical records.
      - Actuators: Screen for diagnoses, treatments.
      - Sensors: Keyboard input, medical test results.
    - - Environment: Floors, dirt.
      - Actuators: Movement, vacuum.
      - Performance: Amount cleaned, energy use.
      - Sensors: Dirt detectors, location sensors.
- - - - Independent short episodes (inspection tasks)
      - Example: Spam email filter
        Each email is judged separately; no link to previous decisions.
    - - Actions affect future decisions (chess, driving).
      - Example: Playing a video game
        Choices in earlier levels affect future gameplay and decisions.
  - - - Environment doesn't change during decision-making (crossword puzzle).
      - Example: Crossword Puzzle Solver
        Puzzle doesn't change while solving.
    - - Environment changes continuously (driving).
      - Example: Stock Trading Agent
        Market changes continuously during decision-making.
  - - - Next state is certain (chess).
      - Example: Tic-Tac-Toe
        Each move leads to a predictable outcome; no uncertainty.
    - - Next state uncertain, probabilistic outcomes (traffic).
      - Weather prediction system
        Unpredictable weather conditions, affected by many uncertain variables.
  - - - Limited states/actions (chess moves).
      - Example: Chess
        Limited, countable number of moves and board states.
    - - Continuous states/actions (driving).
      - Example: Robot arm in a factory
        Can move in continuous space; joint angles and positions are real-valued.
  - - - Only one agent involved.
      - Example: Solving a Sudoku puzzle
        No other agent is involved; the program solves it alone.
    - - Multiple agents interacting (competitive/cooperative).
      - Example: Self-driving cars in traffic
        Each car is an agent interacting with other agents (cars, pedestrians, traffic lights).
  - - - Agent fully understands rules/environment.
      - Example: Solving a maze with a map
        Agent knows the layout and rules in advance.
    - - Agent must learn environment structure.
      - Example: Exploring a new building with no map
        Agent must explore and learn about the environment as it moves.
  - - - Agent can observe complete state.
      - Example: Chess
        The agent can see the full board and the positions of all pieces at all times.
    - - Limited or noisy sensor data
      - Example: Autonomous Car
        Sensors like cameras and radars may miss pedestrians or traffic signs due to fog, darkness, or obstruction.
- - - - Decisions based on utility function to maximize overall satisfaction.
      - Example: Self-Driving Car
        
        Picks a route not just to reach the destination but also to maximize comfort, minimize time, and reduce fuel usage.
    - - Uses a model of the world to handle partially observable environments. Maintains an internal state.
      - Example : Robot Vacuum Cleaner (like Roomba)
        
        Remembers where it's already cleaned, detects furniture, and updates its map of the room.
    - - Improves performance over time by learning from experiences.
      - Example : Online Recommendation System (like Netflix or YouTube)
        
        Learns from user behavior over time to recommend better content.
    - - Acts based solely on current percept (condition-action rules).
      - Example : Thermostat
        
        If room temperature < 24°C, then turn on heater.
        
        No memory or consideration of previous temperatures—just reacts.
    - - Makes decisions based on explicit goals (planning, search).
      - Example: Autonomous Delivery Drone
        
        Given a goal to deliver a package, it plans a path avoiding no-fly zones and traffic to reach the destination.