Skip to main content

Capstone Project: The Autonomous Humanoid. A final project where a simulated robot receives a voice command, plans a path, navigates obstacles, identifies an object using computer vision, and manipulates it.

Heading Breakdown​

Capstone Project: The Autonomous Humanoid is the summation of the entire curriculum. Autonomous means the robot makes its own decisions. Humanoid means it does so in a human-shaped body. A final project... details the scope: Voice command (Module 4, Whisper), Plans a path (Module 3, Nav2), Navigates obstacles (Module 2, Gazebo/Physics), Identifies an object (Module 3, Isaac Perception), and Manipulates it (Module 5, VLA). The importance is integration; demonstrating that you can build a system greater than the sum of its parts. Real usage is the Unitree G1 standing in a kitchen, hearing "Pour me some water," and executing the task autonomously. This is key for upgradable high-DoF humanoids; it proves the architecture is robust enough for real-world service.

(Note: Sidebar refers to HRI Design, but per mapping, we cover the Capstone here).

Training Focus: System Integration​

We focus on glue.

  • State Machines: Using Smach or BehaviorTree.CPP to manage the overall flow.
  • Error Handling: What happens if the cup slips?

Detailed Content​

The Mission​

  1. Start: Robot is at charging dock.
  2. Trigger: User says "Find the blue cube."
  3. Search: Robot explores the house (Nav2).
  4. Detect: Camera sees blue pixels (YOLO).
  5. Approach: Robot walks to table.
  6. Grasp: Arm reaches out (MoveIt).
  7. Return: Robot brings cube to user.

Evaluation Criteria​

  • Robustness: Does it work 10/10 times?
  • Speed: Is it faster than a snail?
  • Safety: Did it hit the wall?

Industry Vocab​

  • System of Systems: A complex entity built of smaller complex entities.
  • Full Stack Robotics: From soldering iron to neural network.
  • Deployment: Getting the code off your laptop and onto the robot.

Code Example: The Main Loop​

# Defensive Main Loop
def main():
    rclpy.init()
    robot = RobotInterface()
    
    try:
        # 1. Wait for Voice
        command = robot.listen_for_command()
        
        # 2. Plan
        plan = robot.ask_llm(command)
        
        # 3. Execute with Watchdog
        for action in plan:
            status = robot.execute(action, timeout=10.0)
            if status == 'failed':
                robot.speak("I'm sorry, I failed to " + action.name)
                robot.go_home()
                return

        robot.speak("Task complete.")
        
    except KeyboardInterrupt:
        robot.emergency_stop()
    finally:
        robot.shutdown()

Real-World Use Case: The Demo​

This project is your portfolio. It demonstrates to employers that you understand the entire pipeline. You aren't just a "vision guy" or a "control guy"; you are a Roboticist.