Sizing for the Job: The Power Budget
Choosing an actuator is not about finding the strongest motor you can afford. It is about matching the machine to the task without waste. The first thing to pin down is the power and energy budget: how much mechanical power the joint must deliver, and how long the battery can keep delivering it. Power is force times speed — pushing hard while moving fast — so a joint that must lift a heavy arm quickly demands far more than one that nudges a lightweight gripper.
Three numbers anchor the choice: the torque (turning force) the joint needs at its hardest moment, the speed it must reach, and the energy stored in the battery. The motor must supply the torque; a gearbox can trade speed for torque if the motor alone falls short; and the battery sets how long the whole machine can run before it must rest or recharge.
An honest budget treats this as a chain, not a single part. The battery feeds a motor driver (the electronics that switch power into the windings), which feeds the motor, which feeds the gears, which finally moves the load. Every link loses a little energy as heat, and the weakest link caps the whole. Size the chain end to end, and you avoid the classic mistake of a powerful motor starved by an undersized driver or a tired battery.
The Heat Ceiling: Duty Cycle and Stall
A motor can almost always produce more force than it can sustain. The limit is not strength — it is heat. Current flowing through the windings warms the copper, and if it warms faster than the motor can shed heat to the air, the insulation eventually cooks. This is why every actuator carries a duty cycle and thermal limit: the fraction of time it may run hard, and the temperature it must never cross.
Here is the trap that catches beginners. A motor that easily lifts a load can quietly burn out while merely holding it. Lifting is brief: the motor spins, the moving parts fan air past the windings, and the moment passes. Holding is forever: the motor sits still, pushing hard against gravity, drawing heavy current while turning at zero speed — so it does no useful mechanical work, and nearly all that electrical power turns straight into heat, with no motion to help carry it away.
That worst case has a name. When a motor is fully blocked and cannot turn, it produces its maximum twisting force — its stall torque — but at the price of maximum current and maximum heat. Stall torque looks tempting on a spec sheet, yet it is a danger zone, not an operating point. Designers leave generous margin below it and avoid lingering there.
Going Gearless: Direct Drive and Backdrivability
Most robots multiply a fast, weak motor into a slow, strong joint with gears. Gears are wonderful for raw force, but they have a cost. They add backlash — a tiny dead zone of slack when motion reverses — they lose energy to friction, and, crucially, they make the joint hard to push backward. With direct drive, the joint is bolted straight to the motor with no gears at all, accepting less torque in exchange for a different set of virtues.
The headline virtue is backdrivability and under-actuation: you can grab the joint and turn it by hand, and the motor turns with you instead of resisting like a locked screw. A high-ratio gearbox is the opposite — push on the output and the gears barely budge, because the leverage that multiplies the motor's force also multiplies any force you apply against it. Direct drive keeps that path open in both directions, so the joint stays sensitive to the outside world.
Why does that matter? A backdrivable joint can feel what it touches. Push against a wall and the motor current rises in a way the controller can read, so the robot senses contact without a dedicated sensor. It is also gentler in a collision: with no rigid gear train to absorb the blow, the joint yields and gives, the way your own arm relaxes when bumped. This natural give is a form of mechanical compliance — built into the drivetrain instead of bolted on.
The Frontier: Quasi-Direct-Drive Legs
Pure direct drive is too weak for a robot that must carry its own weight on its legs. The frontier compromise is quasi-direct drive: a powerful, low-ratio brushless motor paired with a single light gear stage — often a ratio near six-to-one instead of a hundred-to-one. The small gearing supplies enough torque to stand and leap, while staying gentle enough that the leg remains backdrivable and still feels the ground.
This is why the new generation of quadrupedal robots and other legged machines can trot across gravel, recover from a shove, and absorb a hard landing without shattering. A backdrivable leg lets the robot run force control — commanding how hard each foot pushes rather than exactly where it goes — so the limb negotiates with rough terrain instead of fighting it. The actuator's mechanical sensitivity is doing work the software would otherwise struggle to do.
An older, complementary idea reaches the same goal from the other side. A series-elastic actuator deliberately inserts a soft spring between the motor and the load. The spring's deflection, measured precisely, reveals the exact force flowing through the joint — and shields the gears from sudden shocks. Quasi-direct drive softens the drivetrain; series elasticity adds a spring; both chase the same prize: an actuator that does not merely move, but feels.
With this, the actuator chapter comes full circle. We have moved from raw force, through the discipline of power and heat budgets, to actuators sensitive enough to feel the world through their own motors. What the robot does with that feeling — how it balances, steps, and recovers in real time — is the story of control and locomotion, where these sensitive drives finally come alive.