Canonical Spherical-Wrist 6R Pattern

Shared subset · near-neighbours · what is and is not invariant
← Back to index

The pattern

The diagram represents a common subset of 6R industrial arms: a vertical base, parallel shoulder/elbow axes, and three concurrent wrist axes whose successive directions are perpendicular. A horizontal shoulder offset a₁ is common but may be zero. The drawing uses the representative a₃ = 0 case, so O₂ and O₃ coincide; the table below deliberately keeps a₃ variable for robots with an elbow offset. Under this frame placement, the family has the displayed shape of the DH table and α values in {0, ±π/2}. The ABB GoFa is compared later as a deliberate near-neighbour, not a member of this subset: its nonzero a₅ separates two wrist axes, so it has no spherical wrist.

z₀ z₁ z₂ z₃ z₄ z₅ z₆ x₁ x₂ x₃ O₀ O₁ O₂=O₃ O₄=O₅ O₆ d₁ a₁ a₂ d₄ d₆ representative case shown: a₃ = 0
z-axis
x-axis
common normal
displacement along z
origin

A conventional spherical-wrist table shape

With the frame placement used below, the spherical-wrist subset fills these slots. This is a useful template, not a unique table and not a template for every 6R arm.

iαiaidiθi
1π/2a₁d₁q₁
20a₂0q₂
3π/2a₃0q₃
4−π/20d₄q₄
5π/200q₅
600d₆q₆
Be precise about geometry versus frame choice. The α magnitudes are 0 or π/2 here because consecutive axes are parallel or perpendicular. In the displayed assignments, d₂ = d₃ = d₅ = 0: parallel-axis origin freedom permits d₂ = 0, and wrist concurrency permits the z₄–z₅ intersection to be used for both wrist origins, giving d₅ = 0. The value d₃ = 0 comes from the regional elbow/forearm geometry and chosen origins; it is not forced by spherical-wrist concurrency.

The α signs depend on the complete frame assignment, and a₃ is not reliably zero: it carries an elbow offset on several robots and can have either sign. The displayed spherical-wrist assignments use a₄ = a₅ = a₆ = 0. GoFa's a₅ = −80 mm is exactly why that robot falls outside the spherical-wrist subset.
The α signs are not invariant. Reorienting a z-axis can affect both adjacent twists, the positive joint-coordinate sense, and constant offsets, so the whole assignment must stay consistent. De Luca's own exams show the resulting sign variation directly: the 2019 Kawasaki S030 paper required all twists to be 0 or +π/2, and its published solution has the α column π/2, 0, π/2, π/2, π/2, 0 — the fourth entry is +π/2 rather than the RS010N assignment's −π/2. Read every frame and sign constraint before assuming a sign pattern.

Published numerical comparison — four members plus GoFa

Four spherical-wrist members and the GoFa counterexample have full numerical DH tables in De Luca's published exam solutions. Values are in millimetres and transcribed from those solutions; the S030 source's apparent row-6 “d₄ = 165” label typo is recorded here as d₆ = 165. Read a₃ to see physical variation within the pattern, then a₅ to see the family boundary.

Parameter Kawasaki RS010N
2024-09-19
Kawasaki S030
2019-06-17
Yaskawa GP7
2024-01-24
ABB GoFa
2025-06-11 · counterexample
Fanuc CR-15iA
2022-09-09
a₁ (shoulder offset)10015040075
a₂ (upper arm)650870445444640
a₃ (elbow offset)0040−110≈192.55
a₅000−800
d₁ (base column)430680330265648
d₄ (forearm)7001080440470700
d₆ (tool flange)881658010175
α column π/2, 0, π/2,
−π/2, π/2, 0
π/2, 0, π/2,
+π/2, π/2, 0
π/2, 0, π/2,
−π/2, π/2, 0
π/2, 0, π/2,
+π/2, π/2, 0
−π/2, 0, −π/2,
π/2, −π/2, 0
Constraint imposed none on twist signs all αi must be 0 or +π/2 none on twist signs all αi nonnegative none on twist signs
Two lessons sit in this table. The α signs follow the complete frame assignment and the problem's constraints, not the robot alone. The S030 and GoFa papers require nonnegative twists, whereas the RS010N paper does not impose that sign rule. The Fanuc goes further: its published α column differs from the RS010N in four entries — rows 1, 3, 4, and 5 — while retaining the same spherical-wrist architecture as the RS010N.

And a₃ is where these robots physically differ. The two Kawasakis have zero elbow offset; the GP7, GoFa and Fanuc do not. On GoFa, a₃ and a₅ are both nonzero and negative. That a₅ entry is not a small exception to a spherical-wrist template — it records the axis separation that makes GoFa non-spherical.
The Yaskawa Motoman UP500 exam (2025-02-07) has no published DH solution in the course archive. Its data sheet is not enough reason to copy this template blindly: establish each joint-axis relation, common normal, origin, and sign constraint before filling any slot.

Recognizing the spherical-wrist subset

How to decide whether the conventional template is an appropriate starting point.

1. Six revolute joints

All joints are R (no prismatic)
3 + 3 split: regional + wrist

2. Vertical base rotation

Joint 1 axis is vertical
z₀ points up along it

3. Parallel shoulder+elbow

Joints 2 and 3 both horizontal
Same direction → α₂ = 0

4. Orthogonal spherical wrist

Joints 4, 5, 6 intersect at one point
Successive wrist axes are perpendicular; O₄ = O₅ here

5. Regional offsets may vary

a₁ may be zero or nonzero
a₃ may also be zero or nonzero

6. Tool offset

d₆ from wrist center to flange
Usually small (~80-200 mm)
If a 6R robot has the parallel shoulder/elbow pair and three concurrent, successively perpendicular wrist axes, this table is a useful hypothesis. Verify every axis relation and common normal before using it. Then apply the exam's frame, sign, origin, and zero-reference constraints and read all nonzero dimensions from the data sheet — commonly a₁, a₂, a₃, d₁, d₄, and d₆, with a₁ or a₃ allowed to vanish.

Related robots and counterexamples

Most of these robots fall outside the subset; Unitree is included as a symbolic family member.

ABB GoFa CRB 15000

6R near-neighbour · no spherical wrist
a₅ = −80 mm separates two wrist axes
Official problem explicitly states no spherical wrist
All αi constrained to be nonnegative

UR5 (Universal Robots)

No spherical wrist
Joints 2, 3, 4 all parallel (three-in-a-row)
a₁ = 0 (no shoulder offset)
Different α pattern: π/2, 0, 0, −π/2, π/2, 0

Efort ECR5

Non-spherical wrist
Wrist joints have offsets between them
Negative d₄, d₅ (exam-specific convention)

Panda assignment / FR3 comparison

7R · distinct source assignments
2018 Panda solution gives symbolic signs
2024 FR3 solution uses a separate nonnegative-constant frame assignment

PAL TIAGo

Prismatic joint 1
Torso lift is prismatic (d₁ = q₁, θ₁ = 0)
1P + 7R = 8 dof ⇒ redundancy 2
z₀ ∥ z₁ (α₁ = 0)

Siemens Artis Zeego

α = π on row 6, 7 rows total
Medical C-arm geometry
6R + 1P = 7 dof
Row 6 is prismatic with α = π

Unitree Z1 Pro

exam 2026-06-10 · symbolic spherical-wrist member
Spherical wrist, so it is in the family
a₃ > 0 (elbow offset); d₃ = 0
α column −π/2, 0, −π/2, π/2, −π/2, 0 — every nonzero twist sign flipped vs the Kawasaki