Robots Atlas>ROBOTS ATLAS
Robotics

Reflected inertia

ActivePublished: 21 October 2026Updated: 21 October 2026Published
Key innovation
Captures the inertia of one drivetrain element as it is “seen” through the gearbox by the other and shows that it scales with the square of the gear ratio (N²) — making the reduction ratio the dominant parameter governing drive dynamics, backdrivability and impact safety of a robot joint.
Category
Robotics
Abstraction level
Primitive
Operation level
Robot control
Use cases
Motor and gear-ratio sizing for robot jointsDesigning quasi-direct-drive (QDD) actuatorsAssessing joint backdrivabilityImpact-force and collision-safety analysisTuning torque- and speed-control bandwidthProprioceptive torque control without a force sensor

How it works

1) A gearbox of ratio N couples the speeds: ω_motor = N·ω_output, and the torques: τ_output = N·τ_motor·η (η — efficiency). 2) From conservation of rotational kinetic energy ½·J·ω², the equivalent inertia referred across the gearbox changes with the square of the speed ratio, i.e. with N². 3) Load inertia reflected to the motor shaft: J_reflected = J_load / N² (large N → the motor “feels” a light load). 4) Rotor inertia reflected to the output/joint: J_reflected = J_motor · N² (large N → the joint “feels” a heavy, hard-to-backdrive drive). 5) Total joint output inertia ≈ J_motor·N² + J_link; at high N the first term dominates, which ruins backdrivability and increases impact forces. 6) Quasi-direct-drive (QDD) actuators deliberately use low N (on the order of 6:1–10:1) to keep reflected inertia small, achieving backdrivability, proprioceptive torque control and impact resilience.

Problem solved

It lets engineers compute and design the real inertia “felt” on each side of a gearbox instead of assuming the motor drives the load directly. Without this quantity one cannot correctly size the motor and gear ratio, predict torque-control bandwidth, backdrivability, or the peak forces that arise when a robot joint takes an impact.

Implementation

Implementation pitfalls
Ignoring reflected inertia in joint designHigh

Assuming the motor drives the link directly and omitting the J_motor·N² term leads to poor motor sizing, unexpectedly high output inertia and incorrect dynamics models.

Fix:Always include the reflected rotor inertia (J_motor·N²) in the joint output inertia and in the controller’s dynamics model.
N² dominance at high gear ratiosHigh

At high ratios the reflected rotor inertia grows quadratically and starts to dominate the link inertia, drastically increasing peak impact forces and degrading backdrivability.

Fix:Use low gear ratios (QDD actuators), low-rotor-inertia motors or compliant elements (SEA) to limit the effect of the N² term.

Evolution

2017
MIT Cheetah proprioceptive actuators — design for low reflected inertia
Inflection point

Sangbae Kim’s group formalized design principles for low-ratio quasi-direct-drive actuators, showing that minimizing reflected inertia (J_motor·N²) is key to impact mitigation and high-bandwidth physical interaction in dynamic legged robots.

Hyperparameters (configurable axes)

Gear ratio (N)Critical

The gearbox reduction ratio. Rotor inertia reflected to the output grows with N², so this is the single most important parameter shaping joint output inertia, backdrivability and impact forces.

6:1 – 10:1 (QDD)Low ratio: small reflected inertia, good backdrivability and impact resilience.
100:1 (przekładnia falowa)High ratio: large reflected inertia (J_motor·N²), poor backdrivability, high impact forces.
Rotor inertia (J_motor)High

The moment of inertia of the motor rotor. Multiplied by N² it gives the reflected-inertia term at the output; low-rotor-inertia motors reduce reflected inertia.