피지컬 ai 6

Main ideas, concepts, and lessons

1) Actuators and why arm robots matter

• An actuator is a moving, acting device.
• The video uses arm robots as the foundation for understanding how to control an actuator.

2) Core math for arm robots: trigonometry

To describe arm movement, the video emphasizes trigonometry with a right triangle:

• sine θ = height / hypotenuse
• cosine θ = base / hypotenuse
• tangent θ = height / base

Lesson: Trigonometric relationships are fundamental for describing robot link positions.

3) Servo motors and PWM (actuation input)

• A servo motor is presented as a motor with an adjustable angle.
• Control method: adjust the rotation angle by controlling PWM (pulse-width modulation) pulse width supplied to the motor.

PWM example mapping:

• -90° via a small pulse
• 0° via a medium pulse
• 90° via a longer pulse

4) Joints, endpoints, and kinematics (robot arm movement)

• A joint is conceptualized like a human hip/elbow that enables relative motion.
• For a two-link arm (two motors/joints):
  • Joint angles relative to the ground: θ1 and θ2
  • Link lengths: L1 and L2

Endpoint position (end effector) in components:

• Endpoint y coordinate: L1 sin(θ1) + L2 sin(θ2)
• Endpoint x coordinate: L1 cos(θ1) + L2 cos(θ2)

Lesson: With link lengths and joint angles, you can compute the endpoint position.

5) Forward vs inverse kinematics (explicit comparison)

Forward Kinematics

• Inputs: fixed joint angles (θ1, θ2)
• Output: endpoint (end effector) position
• Key point:
  • Computation is relatively straightforward.
  • Endpoint placement is not intuitive—you must set angles first, then see where the endpoint lands.

Inverse Kinematics

• Inputs: desired endpoint position
• Output: required joint angles
• Key point:
  • Endpoint can be chosen intuitively.
  • Solving joint angles is more complex, and some desired endpoints may be unreachable (no solution exists).

6) Matrix / transformation viewpoint (and why it scales)

• The video introduces coordinate transformation for expressing rotations.
• For one link, you can represent the position using cos θ and sin θ, then place it into a matrix form.
• For multiple joints/links:
  • Instead of only step-by-step formulas, represent successive transformations using matrix multiplication.
  • Compose transformations from origin → first frame → next frame → end effector.

Lesson: Matrix methods make it easier to handle more than two joints, even if manual derivations get harder.

7) Inverse kinematics approach (conceptual)

• The video references geometric tools for inverse kinematics:
  • Law of Sines
  • Law of Cosines
  • Pythagorean theorem-based angle reasoning
• Lesson: Inverse solutions can be derived geometrically, though complexity increases with more joints.

---

Methodology / instruction-like structure (detailed bullets)

A) Computing a 2-link arm endpoint position (forward kinematics)

Given:

• Link lengths L1, L2
• Joint angles θ1, θ2

Compute endpoint coordinates:

• x = L1 cos(θ1) + L2 cos(θ2)
• y = L1 sin(θ1) + L2 sin(θ2)

If the endpoint is known instead and you want joint angles:

• Use inverse kinematics (may rely on geometric laws like sines/cosines and may have no solution).

B) Forward vs inverse kinematics workflow (conceptual)

Forward kinematics

• Step 1: choose θ1, θ2
• Step 2: compute (x, y) of the end effector

Inverse kinematics

• Step 1: choose desired (x, y) endpoint
• Step 2: solve for θ1, θ2 that achieve it (possibly impossible if unreachable)

C) Scaling to 3+ joints (matrix transformation method)

When step-by-step trigonometric formulas become difficult:

• Represent each joint/link motion as a transformation
• Multiply/compose transformations in sequence:
  • origin → frame after joint 1 → frame after joint 2 → … → end effector

Benefit: Provides a systematic computation structure for more degrees of freedom.

D) Two-wheel “car-like” / differential drive (wheel rolling and motion)

Rolling distance idea (no slip assumed):

• d = rθ

For differential drive:

• Two-motor robot behavior:
  • If left and right wheel speeds are equal → robot goes straight
  • If wheel speeds differ → robot follows circular motion about a center of rotation (ICC)
• Special cases:
  • If one wheel stops and the other moves → robot spins around the stationary wheel
  • If wheels rotate in opposite directions → robot spins in place

E) Predicting robot motion state (kinematics over time)

• Robot state (pose) includes (x, y, θ)
• Given wheel velocities (VL, VR) and wheel separation L:
  • Determine radius of rotation and rotational speed
  • Predict future pose by integrating velocity changes over time

F) Relationship between differentiation/integration and motion prediction

The video frames it as:

• Differentiation: “current change” (how motion is changing right now)
• Integration: accumulating changes over time to predict future position/angle

Lesson: With current motion quantities and how they change, you can forecast future states.

G) Line tracing (color-following behavior)

Setup:

• Use a downward-facing camera to observe the line (orange line mentioned).
• Determine whether the line appears on the left or right in the camera view.

Control using differential drive:

• If the robot needs to steer right:
  • Accelerate the left wheel (rotate robot toward the right)
• If it needs to steer left:
  • Accelerate the right wheel

Feedback loop concept:

• Continually adjust wheel speeds based on whether the robot drifts out of the line:
  • if it goes out, steer back in and repeat adjustments

---

Speakers / sources featured (as mentioned)

• Newton (referred to as “Mr. Newton” and Principia / Philosophiæ Naturalis Principia Mathematica implied)
• Two speakers are implied:
  • Main speaker/instructor (explaining actuators, kinematics, PWM, robots, math)
  • Another participant/respondent (brief agreement/answers such as “Right?”, “Hmm”, “Yes”, “Please give it some thought”)
• No names are provided for these participants.