Counting the freedoms a mechanism really has
Before you build anything that moves, you want to know one number: how many independent ways can it move? That count is its degrees of freedom (often written DOF), and it is the most basic fact about a machine's mobility. A door on a hinge has one DOF — it can only swing. Your shoulder has three — it can pivot in three independent ways. A free-floating object in space has six: it can slide along three axes and rotate around three axes.
A robot arm is built from stiff bars called links, connected by joints. Each joint adds motion but also takes some away — a joint constrains the two links it connects to move only in a permitted way. The two workhorse joints are the revolute joint (a hinge that rotates, one DOF) and the prismatic joint (a slider that extends, one DOF). String several links and joints in a row and you get a kinematic chain — the skeleton of any arm.
The Grübler–Kutzbach criterion: bookkeeping for motion
Counting DOF by eye works for a simple arm, but it gets tricky once links loop back and connect to each other. Engineers use a tidy bookkeeping formula called the Grübler–Kutzbach criterion. The idea is simple: start by giving every link its full freedom to float, then subtract the freedom each joint takes away. What's left over is the mechanism's mobility.
Planar (flat) mechanisms: M = 3 (N - 1) - 2 J1 - J2 M = mobility (degrees of freedom of the whole mechanism) N = number of links, counting the fixed ground as one J1 = number of 1-DOF joints (revolute, prismatic) J2 = number of 2-DOF joints Example - a simple two-link arm bolted to a table: N = 3 (ground + 2 links), J1 = 2 hinges, J2 = 0 M = 3 (3 - 1) - 2(2) - 0 = 6 - 4 = 2 -> 2 DOF, as expected
This bookkeeping also explains a deep split in arm design. A serial manipulator is one open chain — links in a single line, like a human arm — so it is easy to count and flexible to position, but each motor must hold up everything beyond it. A parallel manipulator instead supports its platform with several closed loops at once (think of a flight simulator on six legs); the loops share the load and stay stiff, but they trade away some workspace and make the math harder.
Gears: trading speed for torque between motor and joint
Counting freedoms tells you what can move; now you need something to do the moving with force. Electric motors are fast but weak — they spin happily at thousands of turns per minute, yet by themselves they cannot push a heavy load. A transmission sits between the motor and the joint and reshapes that output, much like the gears on a bicycle let your legs climb a hill they could never push in top gear.
The key number is the gear ratio: how many times the motor spins for each turn of the joint. A 100:1 ratio means the motor turns one hundred times to rotate the joint once. The trade is beautifully simple — the joint turns 100 times slower, but it pushes with roughly 100 times more torque. Speed and force are two ends of the same lever; gears slide you along it.
Robot joints often need huge reductions in a tiny package, so they reach for special gearboxes like the harmonic drive, which can hit 100:1 in one slim stage with almost no backlash — the tiny slop where gear teeth don't quite touch. Backlash matters because every bit of lost motion shows up as positioning error at the end-effector, the tool or hand at the tip of the arm.
Compliance and back-drive: why a little give is safer
Here is the twist that beginners rarely expect: a perfectly rigid robot is often a dangerous robot. Compliance is the deliberate springiness an engineer builds into a joint — its willingness to bend a little when pushed. A stiff arm that bumps a person keeps right on pushing until something breaks. A compliant arm yields, soaking up the impact like a handshake rather than a punch.
Compliance also helps the arm grip and assemble. When you slot a peg into a hole, a rigid arm jams the moment it is a hair off-center; a compliant wrist lets the peg wiggle and settle into place on its own. This same give is what lets a collaborative robot (a "cobot") share a workbench with people safely, without a cage around it.
Closely related is backdrivability — whether you can push the joint and make the motor spin in reverse. A back-drivable joint gives way when you nudge it by hand, so it can sense and respond to outside forces. Big gear ratios fight against this: that 100:1 harmonic drive is hard to back-drive, which is exactly why high-force industrial arms feel like immovable steel. Designers must choose where on the rigid-to-springy spectrum each joint should live.
Frontier: springs in the loop and bodies made soft
The most direct way to build in give is the series-elastic actuator: place a real spring between the motor's gearbox and the joint. Measuring how far that spring stretches tells you exactly how hard the joint is pushing, turning a crude motor into a gentle, force-sensitive one. Legged robots that must absorb the shock of each footfall lean heavily on this idea.
Push further and the rigid-link picture dissolves entirely. Soft robotics replaces metal bars and discrete joints with continuously bending bodies — silicone tentacles, inflatable fingers, pneumatic artificial muscles that contract like real ones. Here, compliance isn't an add-on; it is the whole material. Such robots can squeeze through gaps and cradle a delicate fruit without ever computing a grip force.
Notice how the whole chapter ties together: you count freedoms to know what a body can do, you gear those freedoms to give them force, and you tune their stiffness so the machine is safe and capable when it finally touches the world. Mechanism design is the art of balancing those three.