The inertia tensor describes how a robot's mass is distributed, and therefore how hard it is to spin about each axis — the rotational counterpart of mass, and a must-have for accurate dynamics and control.
The inertia tensor captures how a robot's weight is spread out, which decides how hard it is to start or stop spinning around each direction. A figure skater pulling their arms in spins faster — that's the inertia tensor changing.
Mass tells you how hard it is to push something into motion. But how hard is it to make a robot spin? That depends on how its mass is spread out — captured by the inertia tensor.
What it is
The inertia tensor is a 3×3 matrix that relates a body's angular velocity to its angular momentum. Its diagonal entries say how hard the body is to spin about each axis; its off-diagonal entries (products of inertia) capture how rotation about one axis couples into another. Mass concentrated far from an axis makes it hard to spin about that axis — the same reason a long pole is harder to twirl than a short one.
How mass is spread sets rotational resistance
Mass far from an axis raises the inertia about it. The tensor packages this for all three axes, so controllers know the torque needed to angularly accelerate the body.
Why robots need it accurately
The inertia tensor appears directly in the robot's equations of motion: to compute the torque that produces a desired angular acceleration, you need it. Get it wrong and model-based controllers — computed-torque, whole-body control, drone attitude — mis-predict the motion. This is why every link in a URDF carries an inertial block (mass, center of mass, and inertia tensor), and why bad inertia values make a physics simulation behave strangely or explode.
Handy facts
Principal axes. Every body has three special axes about which the tensor is diagonal (no coupling) — spin about them and rotation is clean and stable.
Parallel axis theorem. Shift the reference point away from the CoM and the inertia grows in a predictable way — how you combine link inertias into a whole robot.
Spinning-skater effect. Pull mass inward and inertia drops, so for fixed angular momentum the spin speeds up — the tensor changing in real time.
Why it matters
The inertia tensor is to rotation what mass is to translation — indispensable for dynamics, model-based control, and faithful simulation. Any robot whose motion is computed rather than just sensed depends on getting it right.
