Stewart Platform (6-DOF)

Six legs, six freedoms

A rigid body in space has six degrees of freedom: three translations (x, y, z) and three rotations (pitch, roll, yaw). To hold and move a body precisely you need at least six constraints. The Stewart platform supplies exactly six — one per extending leg — and arranges them so that any combination of leg lengths corresponds to a unique pose of the moving platform within its workspace.

Each leg is a linear actuator (hydraulic ram, ballscrew, or piezo stack) terminated at both ends with passive multi-axis joints — typically a U-joint at the base and a ball joint at the platform. The legs only push and pull along their axes; they don't carry bending. That means leg actuators can be slim and high-force, and the loads at the joints are pure tension/compression, not torque.

Because the load is shared among six legs in parallel, the structure is dramatically stiffer than a serial robot of comparable size. Where a 6-DOF serial arm sums each gearbox's compliance through a long lever arm, a Stewart platform places six stiff legs in parallel — the result is closer to a truss than to a chain.

Kinematics: easy in reverse, hard forward

The geometry is fully specified by six base anchor points B₁...B₆ in the base frame and six platform anchor points P₁...P₆ in the platform frame. Given a desired platform pose (rotation matrix R and translation t), each leg's length is the Euclidean distance:

L_i = ||(R · P_i + t) − B_i||

That's six independent square-root operations, computable in microseconds on any controller. The inverse problem — given six leg lengths, find the platform pose — is a coupled system of six nonlinear equations with up to 40 real solutions in general. No closed form exists for arbitrary base/platform anchor layouts; controllers either run a numerical Newton iteration starting from the previous solution (cheap, since the platform moves slowly enough that warm-starting always converges) or use a constrained Bertini polynomial solver (rarely needed in practice).

Worked example: leg length variation for a 5° tilt

Take a Stewart platform with hexagonal symmetric anchors at radius r_b = 300 mm on the base and r_p = 200 mm on the platform, separated by a nominal vertical distance h = 600 mm at neutral pose. Leg lengths at neutral are all roughly:

L₀ = √(h² + (r_b − r_p·cos(α))² + (r_p·sin(α))²) ≈ √(360000 + 10000 + 10000) ≈ 624 mm

(α encodes the angular offset between paired base and platform anchors; ≈30° in a typical layout.)

To tilt the platform 5° about the x-axis, each platform anchor moves: P_i' = R_x(5°)·P_i. The y-coordinate of an anchor at +r_p moves to r_p·cos(5°) ≈ 199.2 mm and the z-coordinate gains r_p·sin(5°) ≈ +17.4 mm. The corresponding leg becomes:

L = √((Δx)² + (Δy)² + (h + 17.4)²) ≈ √((Δx)² + (Δy)² + 381200) ≈ 638 mm

That single leg has extended ≈14 mm. Its diametrically opposite leg has shortened by approximately the same amount; the four side legs each change by a few mm. In a system with 1 m of stroke, 14 mm is well inside the envelope, but you can see how a 25° tilt rapidly consumes the available leg travel.

Stewart platform and competing manipulators

Architecture	Topology	Workspace	Stiffness	Typical use	Limit
Stewart / hexapod	6 legs parallel, 6-DOF	Small (≈leg stroke)	Very high	Flight sim, optics, machine tools	Hard forward kinematics; small workspace
Delta robot	3 parallel legs, 3-DOF translation only	Cone above base	Moderate–high	Pick-and-place, fast packaging	No rotation; ≤4-DOF in extended designs
Cartesian / gantry	3 prismatic axes serial	Rectangular box	High along axes	3D printing, CNC, large-area work	3-DOF translation only; bulky
SCARA	2 horizontal revolute + 1 prismatic	Cylindrical sector	Low vertically, high laterally	Assembly, electronics	4-DOF only; not 6-DOF
6-DOF serial arm (UR5, ABB)	6 revolute joints in a chain	Large hemisphere	Low–moderate	Welding, painting, assembly	Compliance compounds through joints; lower repeatability
Cable-driven parallel	4–8 cables to moving body	Very large (hall-sized)	Tension only, low stiffness	Camera platforms, large fabrication	Cables can only pull; needs gravity or biasing

Real-world specs

• PI Hexapod H-840 — 200 mm stroke per leg, payload up to 250 kg, repeatability ±0.1 μm in translation. Used in semiconductor metrology and aerospace alignment.
• Hexel / Moog flight simulator base — multi-ton motion platforms with ≈1 m stroke, ±20° pitch and roll, ±25° yaw, payload 5–10 tonnes. Driven by hydraulic actuators with 5,000 psi supply pressure.
• Symetrie Notus — high-precision 6-DOF hexapod for telescope secondary mirror positioning, ±5 mm range, 0.5 μm repeatability, used at ESO Very Large Telescope.
• CAE Series 7000XR — full-flight simulator (FFS) Level D motion base, six 60-inch hydraulic cylinders, 12,500 lb dynamic payload, sub-50 ms latency from input to motion.
• Atari Hard Drivin' arcade base (1989) — early consumer use; full 6-DOF cockpit motion with hydraulic legs, the first arcade machine with a true Stewart platform.

Variants

• Cable-driven Stewart — replaces rigid extending legs with cables under tension. Massively larger workspace (entire room) but only one-sided forces and inevitable cable sag at low tension.
• Octahedral / 3-3 layout — base and platform anchors arranged as triangles, with each apex carrying two legs. Equivalent kinematics to standard 6-6 layout but mechanically simpler to build.
• 3-PRS / 3-RPS — three legs instead of six, with each leg containing additional internal joints. Provides 3-DOF (limited tilt + heave); simpler and cheaper than a full Stewart but less general.
• Piezo-driven nano-hexapod — sub-millimeter stroke, sub-nanometer resolution. Used in synchrotron beamline optics and AFM stage stabilization.
• Redundant 7-leg Stewart — adds one extra leg for fault tolerance and to internally pre-load the structure, reducing backlash. Hyperstatic; requires precise leg-length calibration.

Where it actually gets used

• Flight simulators. The killer application. Pilots train on a fixed visual display while the cockpit accelerates, tilts, and shakes on a 6-DOF base, fooling the inner ear into perceiving real flight. Washout filtering returns the platform slowly to neutral so it doesn't run out of stroke.
• Vehicle and ride simulators. Same architecture, smaller footprint. Theme-park motion bases, F1 driver-in-the-loop simulators, ship-bridge training.
• Telescope secondary mirror positioning. The secondary mirror must be aligned to micrometers across temperature swings; hexapods provide the active correction loop.
• Machine-tool five-axis heads. The Sprint Z3 head and similar designs replace the conventional A/C two-axis head with a parallel-kinematic structure that's stiffer and faster.
• Active vibration isolation. Floating optical tables and seismic-grade isolation use Stewart platforms with feedback control to cancel ground motion.
• Surgical robotics. Some bone-positioning and orthopedic devices (Taylor Spatial Frame, ROSA Spine) use hexapod kinematics for precise fragment manipulation.

Common failure modes

• Single leg actuator failure. Cascades immediately: the platform loses one degree of freedom and becomes unstable along that leg's axis. Hydraulic platforms incorporate fail-safe valves that lock all legs and lower the platform on accumulator pressure. Loss of all power must result in a controlled descent, not a free-fall.
• Joint binding at workspace edges. The U-joints and ball joints at leg ends have angular limits (typically ±30°). Aggressive yaw inputs at large translation offsets can drive a joint past its limit, jamming or shearing the joint. Workspace planning must respect joint angular envelopes, not just leg-length envelopes.
• Singular configurations. At certain poses the legs become coplanar or coaxial; the Jacobian's determinant goes to zero and the platform gains an unconstrained internal motion. The Stewart-Gough Type II singularity is the canonical example. Modern controllers detect approaching singularities and refuse motion plans that pass near them.
• Calibration drift. Sub-micron repeatability requires sub-micron knowledge of every anchor point. Thermal expansion, mechanical settling, and shipping-induced shift change anchor positions; periodic recalibration is mandatory for high-precision use.
• Hydraulic resonance. Long oil columns in hydraulic legs introduce compliance that resonates with the platform mass. Flight simulator legs typically have a 5–15 Hz first resonance that must be kept out of the closed-loop bandwidth — limiting how quickly the platform can respond.