A kinematic chain is an assembly of rigid links connected by joints that constrain their relative motion in a predictable way. Industrial six-axis robots like the KUKA KR 210 run as 6-DOF open kinematic chains achieving ±0.06 mm repeatability across a 2.7 m reach. The chain exists to convert simple input motion — a motor turning, a cylinder extending — into a useful, controlled output path. Without it, you have parts. With it, you have a mechanism: a four-bar linkage, a robot arm, a Stewart platform, or the suspension under your car.
Kinematic Chain Interactive Calculator
Vary the number of links and 1-DOF joints to see planar chain mobility and a live four-bar style teaching diagram.
Equation Used
The Grubler-Kutzbach equation estimates planar mechanism mobility. Each moving planar link contributes three possible degrees of freedom, while each 1-DOF joint removes two. For the worked four-bar example, n = 4 and j1 = 4, so M = 3(4 - 1) - 2(4) = 1 DOF.
- Planar mechanism with rigid links.
- Only 1-DOF lower-pair joints are counted.
- Ground is included in the link count n.
- Positive M means available motion; negative M indicates over-constraint.
Inside the Kinematic Chain
A kinematic chain is the skeleton of every mechanism. You take rigid links — bars, plates, shafts — and connect them with joints that allow specific motions while blocking others. A revolute joint (a pin) allows 1 degree of rotational freedom. A prismatic joint (a slider) allows 1 degree of translation. A spherical joint (ball-and-socket) gives 3 rotational degrees. The arrangement and count of these lower-pair joints determines what the chain can do, and you predict that behaviour mathematically using mobility analysis before you cut a single piece of metal.
Chains split into two families. An open kinematic chain — also called a serial chain — has links connected end to end with one free tip, like a robot arm. A closed kinematic chain forms a loop, like a four-bar linkage in an oil pump or the connecting-rod-piston-crank loop inside an engine. Closed chains trade workspace for stiffness and accuracy; open chains trade stiffness for reach. A Stewart platform stacks 6 closed loops in parallel and hits sub-micron positioning, but its working volume is a fraction of what a serial manipulator covers.
If the joint clearances or link lengths are wrong, the chain binds or develops slop. A four-bar linkage with link lengths off by 0.5 mm on a 100 mm coupler will trace a path that drifts 1-2 mm at the output — fine for a windshield wiper, fatal for a surgical positioner. Common failure modes are joint wear opening up backlash, bent links shifting the geometry permanently, and over-constraint where the designer accidentally specified more constraints than degrees of freedom, which causes the chain to lock or to load itself in bending even with no external force applied.
Key Components
- Rigid Link: The structural element between joints. Treated as infinitely stiff for kinematic analysis, though in practice deflection under load matters above roughly 1% of link length. Steel links in industrial robots run 5-50 mm wall thickness depending on payload.
- Revolute Joint (R): A pin joint allowing 1 rotational DOF. Bore tolerance of H7/g6 is standard for low-backlash assemblies — clearance above 50 µm on a 20 mm pin shows up as visible wobble at the chain's free end.
- Prismatic Joint (P): A linear slider allowing 1 translational DOF. Common in Linear Actuator-driven chains. Side-load capacity drops sharply if the bearing length is less than 1.5× the stroke.
- Spherical Joint (S): A ball-and-socket allowing 3 rotational DOF. Used in Stewart platforms and tie-rod ends. Heim-style rod ends typically rate ±20° before the ball contacts the housing.
- Cylindrical Joint (C): Combines rotation and translation along the same axis — 2 DOF. Found in steering columns and some drilling-machine chains.
- Ground Link: The fixed reference frame. Every kinematic chain needs one ground link by definition, otherwise you have a free-floating system and mobility math gives nonsense results.
Where the Kinematic Chain Is Used
Kinematic chains describe nearly every machine that moves on purpose. The same DOF and mobility math applies whether you are designing a 50 g micro-robot or the boom of an 800-tonne dragline excavator. What changes is which joint types you pick and how you arrange them. Builders search for this topic when they need to know if a proposed linkage will actually move the way they want, or when they're trying to figure out why a prototype binds.
- Industrial Robotics: The Fanuc M-710iC/50 is a 6-DOF open kinematic chain — six revolute joints in series — handling 50 kg payloads in automotive welding cells at Toyota plants.
- Automotive Suspension: A double-wishbone suspension on the Porsche 911 GT3 is a closed kinematic chain — the upper arm, lower arm, knuckle, and chassis form a 4-bar loop that controls camber through travel.
- Construction Equipment: The boom-stick-bucket of a Caterpillar 336 excavator is a 3-DOF open chain driven by hydraulic prismatic joints, reaching 10.6 m dig depth.
- Surgical Robotics: The Intuitive da Vinci Xi instrument arm uses a redundant 7-DOF serial chain to allow tool repositioning without moving the entry port through the patient.
- Aerospace Test Rigs: Moog Stewart platforms — 6-DOF parallel closed chains — drive flight simulators at FlightSafety International with 1.2 m heave stroke and ±25° pitch.
- Consumer Mechanisms: A Drawer Slide pair forms a 1-DOF prismatic kinematic chain. The De Sterk under-mount hardware on IKEA Maximera drawers is a closed-chain variant with a self-closing 4-bar linkage at the back.
The Formula Behind the Kinematic Chain
The Grübler-Kutzbach equation tells you how many degrees of freedom a planar or spatial chain has before you build it. At the low end of useful chains — M = 1 — you have a constrained mechanism with one input and a fully determined output, like a four-bar linkage. At M = 0 you have a structure (a truss). At M = 6 or higher you have a full general-purpose manipulator like a 6-axis robot. The sweet spot for most industrial mechanisms is M = 1 to 3 — enough freedom to do useful work, few enough joints to keep the chain stiff and cheap. Go above 6 and you're in redundant-manipulator territory where the math has multiple valid solutions and you need a control strategy to pick one.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| M | Mobility (degrees of freedom) of the chain | DOF (count) | DOF (count) |
| n | Total number of links including ground | count | count |
| j1 | Number of 1-DOF joints (revolute, prismatic) | count | count |
| j2 | Number of 2-DOF joints (cylindrical, universal) | count | count |
| j3 | Number of 3-DOF joints (spherical, planar) | count | count |
Worked Example: Kinematic Chain in a vineyard pruning robot arm
A precision-agriculture startup in Mendoza is designing a grapevine pruning robot for narrow trellis rows. They want a chain that can position cutting shears anywhere within a 1.2 m × 0.8 m work envelope and orient the blade at any angle. They need to verify mobility before committing to a CNC build — too few DOF and the shears can't reach behind a cane, too many and the controller and cost balloon. The candidate chain has 7 links (including ground), 6 revolute joints, and 1 prismatic joint extending the shear head.
Given
- n = 7 links
- j1 = 7 1-DOF joints (6 revolute + 1 prismatic)
- j2 = 0 2-DOF joints
- j3 = 0 3-DOF joints
Solution
Step 1 — apply the spatial Grübler equation at the nominal design with 7 links and 7 single-DOF joints:
That result is a problem. M = 1 means the entire chain has only one independent motion — like a slider-crank. The shears would trace a single fixed path through space, which is useless for pruning where every cane sits in a different place. The team needs more DOF.
Step 2 — at the low end of useful pruning-arm complexity, drop one joint to test sensitivity. With 6 links and 6 revolute joints in series:
Zero DOF means a structure — the arm cannot move at all. Each joint added beyond this baseline buys exactly 1 DOF in a serial chain, which is why industrial 6-axis robots have exactly 6 revolute joints feeding ground.
Step 3 — at the high end, scale up to a 7-DOF redundant arm by adding a second wrist roll joint, giving 8 links and 8 revolute joints:
Wait — that gives M = 2, not 7. The Grübler formula doesn't "see" the redundancy because the team is mistakenly counting joints in a closed loop. The original chain is open (serial), so the correct redundant configuration is 8 links + 7 joints, giving M = 6×7 − 5×7 = 7 DOF. The takeaway: for an open serial chain, M simply equals the number of 1-DOF joints, and the team needs at minimum 6 revolute joints to reach any pose in the work envelope, or 7 to do so while avoiding singularities mid-row.
Result
The candidate 7-link, 7-joint design as drawn yields M = 1 DOF — useless for general pruning. To position and orient the shears anywhere in the 1.2 × 0.8 m envelope, the team needs a 7-link, 6-joint serial chain (M = 6 DOF) at minimum, or 8 links and 7 joints (M = 7 DOF) for redundancy that lets the elbow swing around foliage without moving the shear tip. At the low end of useful complexity (M = 0) the arm is a structure; at the nominal 6-DOF point it reaches every pose but hits singularities; at 7 DOF it works around obstacles cleanly — this is why the Kuka LBR iiwa runs 7 axes. If the built arm shows fewer DOF than predicted, look for: (1) an over-constrained closed loop accidentally introduced by a parallel cable carrier or rigid harness, (2) a joint physically locked by interference between adjacent links at the extremes of travel, or (3) a coupling between two motor shafts that turns two independent revolute joints into a single 1-DOF input.
Kinematic Chain vs Alternatives
The choice between an open serial chain, a closed parallel chain, and a hybrid is the first architectural decision in any mechanism design. Each one wins on different metrics, and getting this wrong locks in performance limits no amount of tuning can fix later.
| Property | Open Kinematic Chain (Serial) | Closed Kinematic Chain (Parallel) | Hybrid Chain |
|---|---|---|---|
| Workspace per unit footprint | Large — full sphere of reach | Small — typically 10-20% of serial reach | Medium |
| Positioning accuracy (industrial typical) | ±0.05 to ±0.2 mm | ±0.005 to ±0.02 mm | ±0.02 to ±0.1 mm |
| Stiffness (end-effector load capacity vs arm mass) | Low — errors stack down chain | High — load shared across legs | Medium |
| Maximum useful speed | 1-3 m/s tip speed | 10+ m/s (Delta robots in pick-and-place) | 2-5 m/s |
| Mobility math complexity | Trivial — M = number of joints | Requires Grübler with loop equations | Full mobility analysis required |
| Cost for equivalent payload | Lower | 2-4× higher | 1.5-2× higher |
| Best application fit | Reach-dominated tasks: welding, painting | Precision/speed tasks: CMM, flight sim, pick-place | Machine tools, surgical arms |
Frequently Asked Questions About Kinematic Chain
You've hit a singularity — a configuration where two joint axes align and the chain instantaneously loses one effective DOF even though M is still 6 on paper. The most common one is the wrist singularity, where joints 4 and 6 share an axis and the controller can't decide how to split the rotation between them.
Diagnostic check: query the joint angles at the failed pose. If joint 5 is near 0° (or 180°), you're in a wrist singularity. The fix is either a redundant 7-DOF arm like the Kuka iiwa, or path planning that routes around the singular region — which is why production cells often forbid certain orientations at certain XYZ points.
Use the planar form M = 3(n−1) − 2j1 − j2 only when every joint axis is parallel (or every motion lies in one plane), like a four-bar linkage or a windshield wiper. Use the spatial form for anything 3D — a robot arm, a Stewart platform, an excavator boom with offset joints.
If you apply the spatial equation to a planar mechanism you'll get a negative M, which falsely tells you the chain is over-constrained. A real four-bar gives Mplanar = 1 but Mspatial = −3 — yet it works fine in practice because the parallel-axis condition gives you 4 redundant constraints that the spatial formula doesn't account for.
Three numbers decide it: required workspace, required accuracy, and required dynamic bandwidth. If you need to reach 1 m+ and ±0.1 mm is acceptable, a serial 6-axis like the Fanuc M-20iD wins on cost and footprint. If you need ±0.01 mm or sub-degree pose stability under 50 kg load, the Stewart platform wins because the load is shared across 6 legs in tension/compression — no cantilever bending.
Rule of thumb: serial robots have stiffness around 1-5 N/µm at the tool point; Stewart platforms run 50-500 N/µm. If your process force divided by your allowable deflection exceeds 5 N/µm, walk away from serial.
You've built a Grashof-violation linkage, or you're at a toggle position. Grübler tells you the chain has 1 DOF in general — it doesn't tell you the input link can complete a full rotation. Grashof's law says s + l ≤ p + q (shortest + longest ≤ sum of the other two) for the input crank to rotate fully.
If your link lengths violate Grashof, the mechanism becomes a rocker-rocker and physically jams at the limit positions. Measure all 4 link lengths with calipers, plug into Grashof, and you'll find one is 1-3 mm longer than the design called for — usually because pivot-hole centres drifted during fabrication.
Errors stack geometrically along an open kinematic chain. A 50 µm clearance at joint 1 multiplied by a 1 m link gives roughly 50 µm of tip error from that joint alone. Add 5 more joints, each contributing similar slop, and the tip error doesn't add — it compounds because each downstream joint amplifies the angular error of every upstream joint by its lever arm.
Quick check: lock joints 2-6, push the tip and measure deflection — that gives you joint 1's contribution. Repeat for each joint. The biggest contributor is almost always the joint with the longest downstream lever arm, not the one with the worst bearing.
Yes, with caveats. A flexure provides approximately 1 rotational DOF over a small angular range, so Grübler treats it as j1 = 1. The trap is range — a typical thin-section flexure works over ±5° to ±15°, not the full 360° a real revolute joint allows. Beyond that range the flexure either yields plastically or buckles.
Compliant mechanisms designed by Larry Howell's group at BYU explicitly use this — they apply standard mobility analysis but constrain the range of motion separately. If your design needs more than about 30° of rotation per joint, use a real bearing.
Redundancy buys obstacle avoidance and singularity escape, but it doesn't buy accuracy — and it can hurt it. Each extra joint adds bearing clearance, encoder error, and structural compliance to the error budget. A 7-DOF arm typically has 15-20% worse repeatability than its 6-DOF equivalent at the same payload.
The reason to choose 7-DOF is when your task fails on a 6-DOF arm — pruning around branches, MRI-guided surgery — not when you need tighter tolerances. For pure accuracy at a fixed pose, fewer joints with stiffer bearings always wins.
References & Further Reading
- Wikipedia contributors. Kinematic chain. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
