Part 1: The Geometry of Motion

Welcome to the first installment of our "Robotics Zero to Hero" series. Before a robot can interact with the world, it needs to understand its own place within it. This fundamental understanding is rooted in the Geometry of Motion.

In this post, we will explore how to mathematically describe the position and orientation of a rigid body in 3D space.

The Math: SO(3) and Rigid Body Transformations

To describe an object in 3D space, we need two pieces of information:

1. Translation: A 3D vector $\mathbf{t} \in \mathbb{R}^3$ describing the position of the object's origin.
2. Rotation: A mathematical representation of the object's orientation.

The set of all valid rotation matrices in 3D forms the Special Orthogonal group, denoted as $SO(3)$. A matrix $R \in SO(3)$ must satisfy two conditions:

1. It is orthogonal: $R^T R = I$ (where $I$ is the identity matrix).
2. Its determinant is 1: $\det(R) = 1$.

A complete Rigid Body Transformation (or Homogeneous Transformation) combines rotation and translation into a single $4 \times 4$ matrix belonging to $SE(3)$:

Representing Rotations: Euler Angles vs. Quaternions

While rotation matrices are great for computation, they are over-parameterized (9 elements for 3 degrees of freedom). Two common alternatives are:

• Euler Angles $(\phi, \theta, \psi)$: Describe rotation as a sequence of three elemental rotations (e.g., Roll-Pitch-Yaw). They are intuitive but suffer from Gimbal Lock, where a degree of freedom is lost when axes align.
• Quaternions $\mathbf{q} = q_w + q_x i + q_y j + q_z k$: A 4D hypercomplex number system that avoids Gimbal Lock and is computationally efficient for interpolating rotations (using SLERP).

Python Implementation: Rotation Matrices

Let's look at how to construct basic rotation matrices around the X, Y, and Z axes using Python and NumPy.

Loading Interactive Python Environment...

Focus on the Octopus: Central Body to Tentacle Mapping

At Ensemble Control, we are building a metallic continuum octopus robot. The octopus has a central body (the head/base) and multiple flexible tentacles extending outward.

To control a specific tentacle, we must first map the central body's orientation to the base frame of that specific tentacle. If the central body is tilted, the reference frame for every tentacle must be rotated accordingly.

Let $T_{world}^{body}$ be the transformation from the world to the octopus body, and $T_{body}^{tentacle\_i}$ be the fixed transformation from the body center to the base of tentacle $i$. The absolute pose of the tentacle base is simply:

Understanding this transformation chain is the first step toward bringing our metallic octopus to life. In the next post, we will explore Forward Kinematics to calculate exactly where the tip of the tentacle is located in space.