IP Library Granted Patent US 12709013
Granted Patent B2
US 12709013 · App. 18/715,990 · Granted Aug 18, 2026

Robot calibration method and device based on equivalent kinematic model

Inventors: Tian Huang (Tianjin, CN); Yongbin Song (Tianjin, CN); Wenjie Tian (Tianjin, CN); Haitao Liu (Tianjin, CN); Juliang Xiao (Tianjin, CN); Kongfeng Zhu (Tianjin, CN); Shiwei Zhou (Tianjin, CN)
Assignees: Tianjin University; GSK CNC Equipment Co., Ltd.
B25J9/1692B25J9/1605
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12709013
App. No.
18/715,990
Granted
Aug 18, 2026
Kind
B2
Abstract

The present disclosure relates to a robot calibration method based on an equivalent kinematic model and a device thereof. The method includes: calibrating a robot real system with a degree of freedom of f≤6 and containing multiple types of time-invariant error sources as a kinematic equivalent system with six degrees of freedom; wherein 6-dimensional pose vectors y of output members of the equivalent system and the real system are the same, and the equivalent system includes a forward position solution model: y=f(ρ(q,α),β 0 ); constructing a function relationship between ρ and α; acquiring a pose measurement value of the real system; calculating an estimator {circumflex over (α)} i , of the adjustable constant parameter vector α in a decoupling manner in a joint space; modifying a driving joint variable corresponding to a received motion trajectory instruction set; and driving the robot to move according to the modified driving joint variable.

Claims (579)

1 . A robot calibration method based on an equivalent kinematic model, comprising:

calibrating a robot real system with a degree of freedom of f≤6 and containing multiple types of time-invariant error sources as a kinematic equivalent system with six degrees of freedom; wherein 6-dimensional pose vectors y of output members of the equivalent system and the real system are the same, and the equivalent system comprises a forward position solution model: y=f(ρ(q,α),β 0 ); where β 0 is a nominal value of a known kinematic parameter of the real system, ρ is a total joint variable of the equivalent system, q is a driving joint variable of the real system, and α is an adjustable constant parameter vector;

constructing a function relationship between ρ and α:

ρ

(

q

,

α

)

=

ρ

0

+

Z

(

q

)

α

,

ρ

0

=

(

q

0

)

,

Z

(

q

)

=

diaq

[

z

(

q

)

]

6

;

where z(q) indicates a function vector with the same dimension as α in which q is an independent variable;

acquiring a pose measurement value of the real system, and based on an iterative algorithm, calculating an estimator {circumflex over (α)} i of the adjustable constant parameter vector α in a decoupling manner in a joint space;

modifying a driving joint variable corresponding to a received motion trajectory instruction set according to z(q) and {circumflex over (α)} i ; and

driving the robot to move according to the modified driving joint variable.

2 . The robot calibration method based on the equivalent kinematic model according to claim 1 , wherein acquiring a pose measurement value of the real system, and based on an iterative algorithm, calculating an estimator {circumflex over (α)} i of the adjustable constant parameter vector α in a decoupling manner in a joint space comprises:

acquiring a pose measurement value y j of an output member at a configuration j (j=1~N) of the real system;

assuming that

α

^

i

k

(

i

=

1

6

,

k

=

0

,

1

,

)

 is a k-th iteration value of α i , and assuming that

α

^

i

0

=

0

;

performing first-order Taylor expansion on f(ρ(q j ,{circumflex over (α)}),β 0 ) in a neighborhood of

α

^

i

0

,

and expressing a residual of y j and f(ρ(q j ,{circumflex over (α)}),β 0 ) as:

e

j

k

=

Δ

y

j

k

-

T

j

k

Z

j

Δ

α

k

Δ

y

j

k

=

y

j

-

f

(

ρ

(

q

j

,

α

^

k

)

,

β

0

)

,

T

j

k

=

f

(

ρ

(

q

j

,

α

)

,

β

0

)

ρ

"\[LeftBracketingBar]"

α

-

α

^

k

,

Z

j

=

Z

(

q

j

)

,

Δ

α

k

=

α

-

α

^

k

;

realizing decoupling in the joint space by simultaneous left multiplication of

(

W

j

k

)

T

=

(

T

j

k

)

-

1

 at both ends of

e

j

k

=

Δ

y

j

k

-

T

j

k

Z

j

Δ

α

k

,

 and grouping and integrating a multivariate linear regression equation;

Δ

ρ

i

k

=

H

Δα

i

k

+

ε

i

k

,

Δ

ρ

i

k

=

(

(

w

1

,

i

k

)

T

Δ

y

1

k

(

w

N

,

i

k

)

T

Δ

y

N

k

)

,

H

=

(

z

1

T

z

N

T

)

,

ε

i

k

=

(

(

w

1

,

i

k

)

T

e

1

k

(

w

N

,

i

k

)

T

e

N

k

)

;

wherein

w

j

,

i

k

indicates an i-th column of a matrix

W

j

k

;

obtaining

Δ

a

i

k

by calculation based on a preset algorithm;

updating

α

^

i

k

=:

α

^

i

k

+

Δ

α

^

i

k

,

executing iterative loop until convergence, and taking a convergence value as the estimator {circumflex over (α)} i .

3 . The robot method based on the equivalent kinematic model according to claim 2 , wherein acquiring a pose measurement value y j of an output member at the configuration j (j=1~N) comprises:

evenly selecting N>n=dim(α i ) measurement configurations in a robot workspace, and determining a pose instruction set {y 0 } corresponding to the configuration;

obtaining an inverse position solution model f −1 (·) of the equivalent system according to the forward position solution model f(·) of the equivalent system;

based on the inverse position solution model f −1 (·) of the equivalent system, mapping a pose vector y 0,j of the output member at a configuration j to the total joint variable ρ 0,j of the equivalent system;

constructing a driving joint variable q j of the real system according to the total joint variable ρ 0,j of the equivalent system;

driving the robot to traverse the configuration j;

measuring coordinates of a preset measuring target point of a calibration gauge on the robot in a calibration coordinate system;

obtaining a pose measurement value y j of the output member at a configuration j (j=1~N) based on the coordinate.

4 . The robot calibration method based on the equivalent kinematic model according to claim 3 , wherein modifying a driving joint variable corresponding to a received motion trajectory instruction set according to z(q) and {circumflex over (α)} i , comprises:

based on the inverse position solution model f −1 (·) of the equivalent system, mapping the received motion trajectory instruction set to an instruction set {ρ 0 } of the equivalent system for the driving joint variable;

constructing an instruction set {q} of the real system for the driving joint variables according to {ρ 0 };

modifying {q} according to z(q) and {circumflex over (α)} i based on a pre-constructed error compensation modification formula {q m,i }={q i −z T (q){circumflex over (α)} i }, in which i=1~5, and obtaining the modified instruction set {q m,i } of the driving joint variable.

5 . The robot calibration method based on the equivalent kinematic model according to claim 4 , wherein the method further comprises:

interpolating the modified instruction set {q m,i } of the driving joint variable.

6 . The robot calibration method based on the equivalent kinematic model according to claim 1 , wherein constructing a function relationship between ρ and α comprises:

expressing each component ρ i (i=1~6) of ρ as a quadratic response surface function with each component q 1 ~q 5 of q as an independent variable:

ρ

i

=

q

i

+

μ

0

,

i

+

j

=

1

5

μ

j

,

i

q

j

+

j

=

1

5

k

=

j

5

μ

j

.

k

,

i

q

j

q

k

,

i

=

1

6

,

q

6

=

0

;

where μ *,*,j indicates coefficients of the quadratic response surface;

obtaining the following formula by variable substitution:

ρ i =q i +z T ( q )α i , where i= 1~6, q 6 =0;

wherein:

z

(

q

)

=

(

1

z

l

z

n

)

T

=

(

1

q

1

q

5

q

1

2

q

1

q

2

q

5

2

)

T

,

α

i

=

(

α

0

,

i

α

1

,

i

α

n

,

i

)

T

=

(

μ

0

,

i

μ

1

,

i

μ

5

,

i

μ

1

,

1

i

μ

1

,

2

,

i

μ

5

,

5

,

i

)

T

,

i

=

1

6.

7 . The robot calibration method based on the equivalent kinematic model according to claim 1 , wherein the function relationship between ρ and α is a linear function or a nonlinear function.

8 . The robot calibration method based on the equivalent kinematic model according to claim 2 , wherein obtaining

Δ

a

i

k

by calculation based on a preset algorithm comprises:

obtaining

Δ

a

i

k

by calculation based on a damped least square method.

9 . The robot calibration method based on the equivalent kinematic model according to claim 2 , wherein prior to acquiring a pose measurement value of the real system, and based on an iterative algorithm, calculating an estimator {circumflex over (α)} i of the adjustable constant parameter vector α in a decoupling manner in a joint space, the method further comprises:

constructing an analytical model or a numerical model of full kinematic Jacobian T=∂f/∂ρ and full dynamic Jacobian W=T −T of the equivalent system.

10 . A robot calibration device based on an equivalent kinematic model, comprising:

a processor and a memory;

wherein the processor is connected with the memory through a communication bus;

wherein the processor is configured to call and execute a program stored in the memory;

the memory is configured to store a program, and the program is at least configured to execute the robot calibration method based on the equivalent kinematic model according to claim 1 .