CtrlK
A cycloidal drive (also called a cycloidal speed reducer, cyclo drive, cycloidal gearbox, or, in industrial robotics, an RV reducer) is a compact mechanical speed reducer that converts a high-speed, low-torque input into a low-speed, high-torque output. Internally, an eccentric input shaft pushes one or more lobed cycloidal disks against a ring of stationary roller pins, and a set of output rollers passing through enlarged holes in the disks picks up only the disk's slow rotation, ignoring the wobble. The result is a single-stage transmission that can reach reduction ratios above 100:1 with low backlash, high torque density, and excellent shock tolerance.
The principle was patented by German engineer Lorenz Braren in 1925 and commercialized in the 1930s under the trade name Cyclo. After a 1937 license to Sumitomo Heavy Industries, the technology spread into Japanese industry, then into industrial robot joints in the 1980s through Teijin Seiki's RV reducer. Today the cycloidal drive sits inside the heavy joints of nearly every major six-axis articulated arm, including machines from FANUC, ABB, KUKA, and Yaskawa. It is also used, alongside strain wave gearing, in the hip and knee joints of modern humanoid robots such as Boston Dynamics' Atlas and Unitree's H1, where the high shock tolerance matters more than the last gram of weight.
The market for precision cycloidal reducers has been dominated for three decades by Nabtesco, the Japanese supplier formed when Teijin Seiki merged with Nabco in 2003. Nabtesco's near-monopoly on the high-load joint reducer (around 60 percent of global supply) is one of the more concentrated bottlenecks in the robotics supply chain, and the rise of Chinese makers since the late 2010s has been driven largely by an explicit national effort to break that hold.
| Property | Value |
|---|---|
| Inventor | Lorenz Konrad Braren (Germany) |
| Original German patents | DE441812 and DE439484 (filed 1925, granted 1927) |
| US patent | US 1,694,031 (granted December 4, 1928) |
| First commercial production | Cyclo GmbH, Munich, 1931 |
| First Japanese license | Sumitomo / former Cyclo Getriebebau, 1937 |
| First robotic RV reducer | Teijin Seiki RV series, mass production 1986 |
| Major manufacturers | Nabtesco, Sumitomo Drive Technologies, Spinea, Onvio, Nidec-Shimpo |
| Typical single-stage ratio | 10:1 to 119:1 |
| Typical compound (RV) ratio | 30:1 to 200:1 |
| Typical efficiency | 75 to 93 percent |
| Typical backlash (precision grade) | Below 1 arcminute |
| Typical shock tolerance | Up to 500 percent of rated torque |
Lorenz Konrad Braren was chief designer at the Friedrich Deckel machine-tool factory in Munich when he sketched the layout. He filed his original German patents in 1925, with the two foundational filings published as DE441812 ("Übersetzungsgetriebe") and the related DE439484 covering the hypotrochoid disc geometry; both issued in 1927. U.S. Patent 1,694,031, "Gear Transmission," was granted on December 4, 1928. In 1931 Braren founded Cyclo GmbH in Munich and began serial production of gearboxes for conveyors, mills, and other constant-duty industrial drives, where the cycloidal reducer's compact form and shock tolerance gave it an edge over worm gears and multi-stage spur reducers.
Braren's geometry was not new mathematics. Cycloidal curves had been studied since the seventeenth century, and cycloidal tooth profiles already appeared in clockwork. What was new was the gearbox layout: an eccentric on the input shaft, two cycloidal disks running 180 degrees out of phase to balance the wobble, a ring of fixed pins, and an output carrier with rollers pinned through enlarged holes in the disks. That arrangement allows a single-stage cycloidal drive to hit reductions in the tens or hundreds with rolling, not sliding, contact at every interface.
In 1937 Cyclo GmbH signed a license agreement with the Japanese gear maker that later became part of Sumitomo Heavy Industries. The first Japanese-built Cyclo reducer rolled off the line at the Niihama Works in 1939, and Sumitomo made the Cyclo a flagship product line. Today Sumitomo Drive Technologies still markets the Cyclo brand (and its bevel-geared Cyclo BBB and BBB1 derivatives) almost a century after the original patent. The Cyclo dominated heavy industrial gearing through the postwar decades, but it was the next leap, repackaging the principle for precision robotics, that made cycloidal drives a household name.
Teijin Seiki, a Japanese aerospace and machinery firm, started building cycloidal reducers in the late 1970s for the swing motors on hydraulic excavators, an application that demands shock tolerance and a compact form factor. In the early 1980s the major Japanese robot makers asked Teijin Seiki to harden the design for use in the base and shoulder joints of their new generation of articulated arms. The result was the RV reducer ("rotate vector," referencing the eccentric two-stage geometry), with a first-generation product line launched in 1980 and refined volume production starting in 1986.
The RV reducer combines a planetary first stage with a cycloidal second stage in one housing. That hybrid layout multiplies ratios while keeping the rugged shock behavior of the cycloidal geometry. Within a few years it became the default solution for the high-load joints (J1, J2, J3) of six-axis industrial arms, while harmonic drives took the lighter wrist joints (J4, J5, J6). In 2003 Teijin Seiki merged with the brake and aerospace company Nabco to form Nabtesco Corporation, and the RV reducer line became Nabtesco's anchor product. By the 2010s the cycloidal drive had quietly become a component without which modern manufacturing would not look the way it does: arms welding car bodies, palletizing boxes, and loading semiconductor wafers all turn on cycloidal reducers, most from a single supplier in Japan.
A single-stage cycloidal reducer has five functional parts: a high-speed input shaft, an eccentric cam keyed to that shaft, one or two cycloidal disks with a lobed (epicycloid or hypocycloid) outer profile, a stationary ring populated with cylindrical pin rollers, and a slow-speed output carrier carrying rollers that pass through enlarged holes drilled through the disks.
In operation, the input shaft spins the eccentric, which forces each cycloidal disk to wobble in a small circular orbit. The lobed edge of the disk presses against the ring of pins, but because the disk has one fewer lobe than the ring has pins, full revolutions of the input shaft only nudge the disk forward by one tooth. The disk therefore rotates slowly in the opposite direction of the input.
The output rollers pick up that slow rotation. They sit in oversized holes in the disk, and the eccentric wobble is absorbed by the gap between roller and hole, so only the disk's true rotation about the central axis transmits to the output shaft. The radial wobble cancels out at the output. Most real units use two cycloidal disks set 180 degrees out of phase. The phase offset cancels static imbalance and reduces vibration; high-speed designs use three or more disks for the same reason.
The basic single-stage reduction follows from the lobe count. If the ring has P pins and the cycloidal disk has L lobes, with P = L + 1, the reduction is
i = L (or equivalently i = N / (N - 1) for an external configuration with N pins),
so for example a disk with 10 lobes engaging 11 ring pins gives a 10:1 reduction in a single stage. Commercial single-stage cycloidal drives have been built up to 119:1. Compound (two-stage) units, which combine two cycloidal stages or a planetary first stage with a cycloidal second stage, reach into the thousands; published designs go up to 7,569:1.
The RV reducer used in industrial robots is a two-stage cycloidal drive. The first stage is a planetary spur reduction (typically with three planet gears) that takes the input down by a modest factor. The second stage is the cycloidal stage proper, with two RV disks set 180 degrees apart, driven by eccentric cams on the carrier shafts of the first stage. This compound layout is what gives RV reducers their characteristic combination of compact diameter, high ratio (typically 30:1 to 200:1), and the very high shock tolerance robotics needs.
Several structural variants of the basic Braren layout are produced today: single-stage cycloidal (one or two disks driven by a single eccentric, used in industrial Cyclo gearmotors and most hobby builds); two-stage compound cycloidal, including the RV layout (planetary primary, cycloidal secondary) and pure stacked two-stage Braren designs that reach ratios into the thousands; twin-disk and multi-disk designs, where two or three offset disks reduce vibration from the eccentric motion (standard in Nabtesco RV and Sumitomo Fine Cyclo); the cycloidal-bearing combo of Spinea TwinSpin, which integrates a high-precision cross roller bearing into the housing; the hybrid cycloidal-segment Wittenstein Galaxie, which swaps Braren's smooth-lobed disk for separate tooth segments running on a polygonal eccentric; and mini or 3D-printable cycloidals used in hobby robotics.
The numbers below come from manufacturer datasheets and peer-reviewed comparisons. Real-world figures vary with size, ratio, lubrication, and load.
| Property | Typical value | Notes |
|---|---|---|
| Reduction ratio (single stage) | 10:1 to 119:1 | Higher ratios possible with non-standard geometry |
| Reduction ratio (two stage / RV) | 30:1 to 200:1 in one housing | Compound cycloidal up to 7,569:1 |
| Backlash | 0.5 to 1 arcmin (precision grade) | Some Spinea TwinSpin and Nabtesco RV-N grades quote ~0.1 to 0.2 arcmin pure backlash |
| Hysteresis loss | About 1 arcmin or below | Industry standard for precision-grade units |
| Efficiency (single stage) | Up to ~93 percent | Drops with very high ratios and very low loads |
| Efficiency (two stage) | Up to ~86 percent | Compound cycloidal or planetary-cycloidal hybrids |
| Shock load tolerance | Up to 500 percent of rated torque | Higher than harmonic drives, which sit around 300 percent |
| Service life | Typically 6,000 to 20,000 hours rated | Rolling contact wears more slowly than the flexspline of a harmonic drive |
Backlash needs a caveat. Marketing copy that describes cycloidal drives as "zero backlash" is a useful approximation but not literally true. Actual angular play under load is small (around an arcminute or less in precision grades) and very stable over the lifetime of the unit because every contact is rolling rather than sliding. That stability matters as much as the absolute number for a robot running the same path a million times.
Shock tolerance is the property that genuinely sets cycloidal drives apart from their main rival. Load is shared simultaneously across roughly half the ring pins, so an overload spike does not concentrate on one or two teeth. The 5x rated-torque shock figure is what makes cycloidal reducers the natural choice for the base joint of a heavy six-axis arm or the knee of a 100-kilogram humanoid.
The two technologies that compete for the same niche, compact high-ratio reducers for robot joints, are cycloidal drives and strain wave gearing (commonly called harmonic drives after the original Harmonic Drive SE / Harmonic Drive LLC trade name). They are different enough in geometry that each has applications where the other simply will not fit.
| Feature | Cycloidal drive | Harmonic drive (strain wave) |
|---|---|---|
| Reduction principle | Eccentric cam, lobed disks engaging ring pins | Wave generator deforms a flexspline inside a rigid circular spline |
| Single-stage ratio | ~10:1 to ~119:1 | ~30:1 to ~320:1 |
| Backlash | <1 arcmin (precision grade), very stable | Often quoted near zero, can creep up with flexspline fatigue |
| Torque density (per unit weight) | High, but heavier than harmonic | Higher torque-to-weight; very compact axially |
| Shock load tolerance | Up to ~500 percent rated | About ~300 percent rated; flexspline can fatigue from peaks |
| Efficiency | ~85 to 93 percent | ~70 to 85 percent, drops at low load |
| Stiffness | High and roughly linear | Slightly nonlinear due to flexspline elasticity |
| Backdrivability | Lower at high ratios; harder to backdrive | Higher; flexspline elasticity helps a motor sense external torque |
| Lifetime under shock duty | Long (rolling contact) | Shorter; flexspline fatigue is the limiting failure mode |
| Mass | Heavier (steel pins, disks, eccentrics) | Lighter (thin-walled flexspline) |
| Cost | Generally more expensive in small sizes | Generally cheaper in small sizes, often the only choice for the wrist |
| Typical robot use | Base, shoulder, elbow, hip, knee | Wrist, fingers, lightweight arms, cobots |
This is why most six-axis industrial arms use both. The bottom three joints carry the weight of the arm and absorb the shock of unloaded moves and crashes, so they run on RV-style cycloidal reducers. The wrist joints need light, compact, low-inertia transmissions and rarely see crash loads, so they run on harmonic drives. Humanoid robots increasingly follow the same pattern: cycloidal-class units at hip and knee, harmonic drives at shoulder, elbow, and wrist.
A 2012 IEEE paper by Sensinger and colleagues compared the two for high-ratio single-stage robotic transmissions and found cycloidal drives thinner, more efficient, and lower in reflected inertia at comparable torque, but harmonic drives lighter and quieter at comparable size. The comparison gets more nuanced once planetary gears enter the picture: planetary stages are cheaper, give moderate ratios per stage (3:1 to 10:1), and dominate in cobots and lightweight humanoids that use quasi-direct drive actuator architectures. A cycloidal stage replaces two or three planetary stages in roughly the same axial length, at the cost of more demanding manufacturing and a small but nonzero amount of backlash.
The cycloidal reducer is a notoriously hard precision part to make. The disk profile must be ground to micrometer-level accuracy or the drive will run with audible ripple and unacceptable backlash. Bearings, eccentrics, and pin rings all demand precision steel and careful heat treatment. For decades the practical knowledge of how to grind a good cycloidal disk was concentrated in a handful of Japanese plants, which is much of why Nabtesco's market share stayed above 50 percent.
Key manufacturing concepts include profile shifting (the mathematically pure cycloidal profile is rarely used; manufacturers shift the profile slightly to control contact pattern and give the disk small clearance to the pin ring), "second-order cycloid" formulations (Hsieh and others have published variants that improve load distribution at the cost of more complex grinding paths), and pin-tooth wear analysis (Pham, Hsieh, Yang and other mechanical-engineering authors have tied contact stress, profile shift, and lubricant film to predicted lifetime). Production lines typically measure each ground disk and matched pin ring, then sort them into bins so paired components meet a target backlash without further fitting.
The Chinese push into precision reducers since around 2015 has been an explicit industrial policy move, treating the cycloidal reducer as a strategic component on par with semiconductors. Chinese entrants include Shuanghuan, Ningbo Zhongda Leader, Suzhou Greenable, Hongfeng, and Yi Sheng. Local prices for RV-class reducers have come down 30 to 50 percent compared with Japanese imports, though Chinese makers still trail on the very high-precision and ultra-low-backlash grades.
The precision cycloidal reducer industry is concentrated, with one Japanese supplier holding the dominant position and a long tail of regional and Chinese entrants. This table covers the main producers and their flagship product lines.
| Manufacturer | Country | Key product line | Position |
|---|---|---|---|
| Nabtesco Corporation | Japan | RV-N, RV-E, RV-C, RD series | About 60 percent of global precision cycloidal supply for industrial robots; near monopoly in heavy-load joint reducers through the 2010s |
| Sumitomo Drive Technologies (Sumitomo Heavy Industries) | Japan | Cyclo 6000, Cyclo BBB, Fine Cyclo | Original Japanese licensee of the Braren patent (1937); leader in industrial Cyclo gearmotors |
| Spinea | Slovakia | TwinSpin (M, T, H, G, E series) | High-precision cycloidal-bearing combo units, 18 to 4,500 Nm range, used in machine tools and robotics |
| Onvio | Japan / United States | Onvio cycloidal reducers | Aerospace and high-precision robotics |
| Wittenstein | Germany | Galaxie | Hybrid cycloidal-segment design with sub-arcminute backlash, marketed as harmonic-class precision with cycloidal robustness |
| Ningbo Zhongda Leader | China | Leader RV / harmonic | Mass-market RV reducer for Chinese robot makers |
| Zhejiang Shuanghuan Driveline | China | RV-C, RV-E, RV-H | Listed on Shenzhen Stock Exchange; entered RV reducer market in 2014 |
| Suzhou Greenable Transmission | China | Cycloidal and harmonic reducers | Niche Chinese supplier, including humanoid robot joint modules |
| SHIMPO Drives (Nidec) | Japan | EVL, ABLE | Lightweight cycloidal reducers and inline gearmotors |
| Twin Disc | United States | Cycloidal industrial reducers | Industrial drives, marine |
The market structure is unusually lopsided. As of the early 2020s, Nabtesco supplied roughly 60 percent of the world's precision cycloidal reducers and around 90 percent of the medium-to-heavy-load RV units used in articulated robots. Top-five manufacturers controlled about two-thirds of global supply, with Sumitomo a distant second at around 19 percent. The remaining slice splits between Spinea, Onvio, Wittenstein, and a growing list of Chinese suppliers.
Outside robotics, the cycloidal drive is a workhorse of heavy-duty industrial gearing. Common applications include CNC machine tool indexers and rotary tables, wind turbine yaw and pitch drives (where shock tolerance handles gust loads), conveyor and bulk-material drives, mining and quarry drives, steel-mill roll-table drives, and marine winches.
The canonical use of a cycloidal drive in 2026 is a Nabtesco RV reducer in the base or shoulder joint of an industrial robot. All Big Four robot makers (FANUC, ABB, KUKA, Yaskawa) source RV reducers from Nabtesco for the high-load joints of their arms, and Nabtesco's supply contracts have at times been exclusive enough that buying their reducers locked out competitors. Industry estimates put the share of cycloidal reducers in mid-to-large industrial robots at around 60 percent of all reduction gearing by unit, with harmonic drives taking most of the rest at the wrist.
The pattern is consistent across the major arms: J1 (base) uses an RV reducer at ratio ~100:1 to 200:1; J2 (shoulder) and J3 (elbow) use slightly smaller RV reducers; J4, J5, and J6 (wrist) use harmonic drives. Universal Robots' UR3 to UR16 cobots are an exception, sticking with planetary and harmonic stages to keep the safety-rated joint mass low.
Cycloidal and cycloidal-style reducers have become standard in the lower-body joints of modern humanoid robots, where shock tolerance from walking and falling matters more than absolute weight. Humanoid makers often build custom hybrid actuators that do not fit cleanly into one category.
The broad pattern: cycloidal-class reducers are favored where joints see both high torque and shock loads (hip, knee); harmonic drives are favored where joints see high precision and tight packaging (wrist, elbow). Most humanoid platforms use both.
Cycloidal drives also appear in robotic positioners and turntables, heavy-duty pick-and-place machines, robot welding cells (where collision shock is a real concern), mobile-robot drive trains, surgical and medical robotics (where Spinea's TwinSpin units pair low backlash with an integrated bearing), and quadruped knee and hip actuators in custom low-ratio cycloidal forms designed for backdrivability rather than maximum torque.
A cycloidal disk is one of the few precision gearbox parts that can be 3D-printed in plastic and still work, because the lobed contact spreads load over many points. The open hardware robotics community has produced several reference designs, including James Bruton's printed cycloidal actuators for his OpenDog quadruped project, the OpenTorque actuator (Gabrael Levine, MIT) which couples a single-stage cycloidal reducer with a direct drive frameless motor, and printable cycloidal designs from ROBOTBUILD, How To Mechatronics, and Aaed Musa aimed at desktop arms and small humanoids. These builds will not match a Nabtesco RV in stiffness or lifetime, but they have made the cycloidal layout the default learning vehicle for high-ratio mechanical reduction in the hobby community.
Beyond the variants section above, recent designs include cycloidal-pin gearing (rolling pins instead of fixed ones, which reduces sliding friction), the Wittenstein Galaxie hybrid (separate tooth segments on a polygonal eccentric, with measured backlash below 0.1 arcminute in the production grade), magnetically preloaded cycloidals (research designs using magnetic preload to remove residual backlash for telescopes and surgical robots), and differential cycloidals (two stages in series at different ratios, reaching ultra-high overall ratios for solar tracking and similar slow-motion applications).
Cycloidal drives have known weaknesses. A single-disk drive has an unbalanced rotating eccentric that produces noticeable torque ripple; twin-disk and triple-disk designs cancel most but not all of it. The pin ring is the long-term wear surface even with rolling contact, so lubrication and load duty heavily influence service life. Below about 30:1, harmonic drives generally show lower backlash and hysteresis at comparable size. A high-end Nabtesco RV reducer for an industrial robot joint typically lists between 1,000 and 10,000 USD depending on size and grade, which is why the humanoid robotics industry treats reducer cost as a primary line item. At high ratios cycloidal drives are difficult to backdrive, limiting their appeal in compliant or impedance-controlled actuators. A cycloidal reducer that matches a harmonic drive on torque is typically heavier, because disks and pin rings are solid steel rather than thin-walled. Cheap cycloidals often show measurable ripple and high backlash because of imperfect grinding.
The humanoid robot wave that started in 2023 changed the demand profile. Each Tesla Optimus, Figure 02, or Unitree H1 carries between 14 and 30 reducers, and projected unit volumes for mass humanoid production exceed the entire current industrial-robot reducer market. That has triggered fresh investment in cycloidal-style and harmonic-style joint modules from Japanese incumbents and Chinese challengers alike. Whether the existing supply concentration holds, or whether humanoid demand finally pries the market open, is one of the more interesting open questions in robotics hardware.