Euler 3D in Practice: Converting Between Euler, Quaternions, and Matrices

Euler 3D: A Beginner’s Guide to 3D Rotations and Orientation

Understanding 3D rotations is essential in graphics, robotics, simulation, and aerospace. This guide introduces Euler angles and their use for representing orientation in three dimensions, explains common conventions and pitfalls, and shows how to convert between Euler angles, rotation matrices, and quaternions.

What are Euler angles?

Euler angles describe a 3D orientation as three sequential rotations about axes. Each angle represents rotation around one axis; the sequence defines the overall orientation. Common sequences include Z–Y–X (yaw, pitch, roll) and Z–X–Z (used in some physics contexts). Euler angles are intuitive because they relate directly to yaw, pitch, and roll motions.

Rotation order and conventions

• Intrinsic vs. extrinsic: Intrinsic rotations apply each rotation in the body’s moving frame; extrinsic rotations apply each in the fixed world frame. The same numeric angles with intrinsic vs. extrinsic interpretation produce different orientations.
• Axis sequence matters: Z–Y–X means rotate about Z, then the (new) Y, then the (new) X. Always state your sequence (e.g., ZYX / yaw-pitch-roll).
• Right-hand rule: Positive rotation direction follows the right-hand rule unless otherwise specified.

Mathematical representation

• Rotation matrix: Each elementary rotation has a 3×3 matrix. Compose matrices in the specified order (matrix multiplication is noncommutative). For Z–Y–X (angles α, β, γ):

  Code
  Copy CodeCopied
  R = R_x(γ)R_y(β) * R_z(α) 

  where R_z(α) = [[cosα,-sinα,0],[sinα,cosα,0],[0,0,1]], etc. (Order depends on intrinsic/extrinsic choice.)

• From Euler to matrix: Multiply the three elementary rotation matrices in correct order to get the full rotation matrix.
• From matrix to Euler: Extract angles using inverse trigonometric functions and careful handling of singularities (see gimbal lock).

Gimbal lock and singularities

Gimbal lock occurs when two rotation axes align, causing a loss of one degree of freedom and making some angles indeterminate. For example, in ZYX (yaw-pitch-roll), pitch = ±90° causes yaw and roll to become coupled. Practical consequences:

• Numeric instability when extracting angles from matrices.
• Erratic interpolation between orientations.

Workarounds:

• Use quaternions for interpolation (slerp) and to avoid singularities.
• Detect near-singular configurations and switch parameterization or apply small offsets.

Euler ↔ Quaternion conversions

• Why quaternions: They represent rotations without singularities, are compact (4 numbers), and efficient for interpolation and composition.
• Convert Euler → quaternion: Compute half-angles and combine using quaternion multiplication formula matching the rotation order.
• Convert quaternion → Euler: Derive rotation matrix from quaternion and extract Euler angles using the same formulas as matrix→Euler, handling singularities.

Example (conceptual, for ZYX yaw(ψ), pitch(θ), roll(φ)):

• Build quaternion q = q_z(ψ) * q_y(θ) * q_x(φ) (order matters).
• For quaternion → matrix, use standard quaternion-to-matrix formula, then extract angles.

Practical tips for developers

• Always document your convention (order, intrinsic/extrinsic, axis directions). Mismatched conventions cause subtle bugs.
• Prefer radians in code; convert only for UI.
• Normalize angles to a consistent range (e.g., (-π, π] or [0, 2π)).
• Use quaternions for animation and interpolation.
• Test edge cases where pitch ≈ ±90° and verify behavior.
• Use robust libraries (Eigen, GLM, scipy.spatial.transform.Rotation) to avoid reimplementing fragile conversions.

Worked example (ZYX / yaw-pitch-roll)

1. Given yaw ψ, pitch θ, roll φ:
   • Compute Rz(ψ), Ry(θ), Rx(φ)
   • Full rotation R = Rx(φ) * Ry(θ) * Rz(ψ) if using intrinsic ZYX interpreted as yaw then pitch then roll applied in body frame (confirm convention).
2. To get Euler from R:
   • θ = asin(-R[2,0]) (careful with value clamping)
   • If |cosθ| > ε then ψ = atan2(R[1,0], R[0,0]) and φ = atan2(R[2,1], R[2,2]); else handle singular case.

Quick reference

• Euler angles: intuitive but can suffer gimbal lock.
• Rotation matrices: robust, easy to compose, 9 values.
• Quaternions: compact, no singularities, best for interpolation.

Summary

Euler angles are a straightforward way to describe 3D orientation and are useful for human-readable controls (yaw/pitch/roll). However, be explicit about conventions, watch for gimbal lock, and prefer quaternions for interpolation and numerical stability. Use well-tested libraries for conversions and always handle singular cases.