You are on page 1of 15

Journal of Mechanisms and Robotics.

Received September 28, 2017;


Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME

A General Approach to the Large Deflection


Problems of Spatial Flexible Rods Using Principal
Axes Decomposition of Compliance Matrices

Genliang Chen Zhuang Zhang


State Key Laboratory of Mechanical Shanghai Key Laboratory of Digital

d
System and Vibration, Manufacture for Thin-walled Structures,

ite
Shanghai Jiao Tong University, Shanghai Jiao Tong University,
800 Dongchuan Rd, Shanghai, China 800 Dongchuan Rd, Shanghai, China
Email: leungchan@sjtu.edu.cn Email: z.zhang@sjtu.edu.cn

ed
Hao Wang

py
State Key Laboratory of Mechanical System and Vibration, and Shanghai Key Laboratory
of Digital Manufacture for Thin-walled Structures, Shanghai Jiao Tong University,

Co
A611 Mechanical Building, Dongchuan Road, Shanghai, 200240, China
Email: wanghao@sjtu.edu.cn

ot
tN
This paper presents a general discretization-based approach of design, analysis and control of those mechanisms dramati-
to the large deflection problems of spatial flexible links in cally increases over their traditional counterparts. Therefore,
compliant mechanisms. Based on the principal axes de- establishment of effective and efficient solutions to the large
rip

composition of structural compliance matrices, a particular deflection problems of flexible links becomes one of the most
type of elements which relate to spatial 6-degree-of-freedom fundamental issues.
(DOF) serial mechanisms with passive elastic joints, is de-
To this end, many researchers have made their great con-
sc

veloped to characterize the force-deflection behavior of the


tributions during the past decades and several approaches
discretized small segments. Hence, the large deflection prob-
are now available to solve the large deflection problems in
lems of spatial flexible rods can be transformed to the deter-
nu

compliant mechanisms. Classical elliptic integral [8–10]


mination of static equilibrium configurations of their equiv-
was used to derive the closed-form solution to the second-
alent hyper-redundant mechanisms. The main advantage of
order, nonlinear differential equations under the Bernoulli-
Ma

the proposed method comes from the use of robot kinemat-


Euler beam assumption. Although the elliptic integral solu-
ics/statics, rather than structural mechanics. Thus, a closed-
tions are both analytical and accurate, the derivation is cum-
form solution to the system overall stiffness can be derived
bersome and only available for relatively simple geometric
straightforwardly for efficient gradient-based searching al-
and/or loading condition. To conquer the limitation, Howell
ed

gorithms. Two kinds of typical equilibrium problems are in-


and Midha [11] introduced the useful concept of ‘pseudo-
tensively discussed and the correctness has been verified by
rigid-body model’ (PRBM) to approximate path coordinates
means of physical experiments. In addition, a 2-DOF planar
pt

of flexible beams in terms of only one single parameter. This


compliant parallel manipulator is provided as a case study
method has been extended to several variations, including
to demonstrate the potential applications.
ce

3R [12, 13], 4R [14], and 5R [15] models, which were suc-


cessfully applied to the analysis and synthesis of compliant
mechanisms [16–19].
Ac

1 Introduction
Owing to their inherent advantages, like compactness in In addition, discretization-based techniques are widely
structure, reduction of wear and backlash, and conservation used for solving the large deflection problems in compliant
of energy [1], compliant mechanisms have benefited a wide mechanisms. Frame finite element methods (FEM) [20] were
range of applications, such as micro-electromechanical sys- utilized for the synthesis of compliant mechanisms involving
tems (MEMS) [2], biomimetic robots [3], minimally invasive nonlinear force-deflection characteristics. Although the non-
surgical devices [4] and even DNA origami mechanism [5]. linear FEM is general and can deal with various large deflec-
However, unlike the traditional rigid-body ones, the compli- tion problems, it suffers from extensive computational cost
ant mechanisms usually undergo large deflection problems and numerical instability. To overcome this drawback, the
in the flexible members [6, 7]. Consequently, the complexity chain algorithm [21] adopted a recursive method to analyze

ASME JMR-17-1331, Chen 1

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c)
the2018 by ASME
discretized small deflection elements, so that the inver- ness and effectiveness of the proposed method. In addition,
sion of overall stiffness matrix can be avoided. Based on this, a 2-DOF planar compliant parallel manipulator is studied as
different types of elements, such as circle-arc method [22], an example to demonstrate the potential applications.
PRBM [23, 24] and chain beam constraint models [25] have The rest of this paper is organized as follows. In Sec.
been proposed to improve the precision and efficiency. 2, the principal axes decomposition of compliance matrices
Apart from the aforementioned works, some others also is reviewed, based on which a particular type of elements is
proposed their own methods to the large deflection problems developed to characterize the force-deflection relationship of
of flexible rods. Adomian decomposition [26] was employed discretized small segments. Then, Sec. 3 demonstrates the
to determine the large deflections of a cantilever beam under establishment of the proposed finite-element approach to the
arbitrary load condition. In [27], an incremental linearization large deflection problems of spatial flexible rods. Two kinds
approach was presented to simplify the nonlinear problem to of typical equilibrium conditions are analyzed in the context

d
a sequence of linear ones. The smooth curvature model [28] of robot kinematic/statics. To validate the correctness of the
was developed as an efficient approach to predict the large proposed method, Sec. 4 conducts a large deflection exper-

ite
deflections of flexure-jointed robots. And recently, the beam- iment on a specialized apparatus. In Sec. 5, a planar com-
constraint model (BCM) [29] was improved based on Timo- pliant parallel manipulator is exemplified as a case study to

ed
shenko beam theory, which results in an analytical model to exhibit the effectiveness of the proposed general approach.
identify the bistable behavior of compliant mechanisms. In the end, some conclusions are drawn in Sec. 6.
Among the existing methods, most of them focus on the

py
load-deflection relation of flexible rods at their tip points.
As a consequence, they are not applicable to complex com- 2 Stiffness Modeling of Small Segment Elements
pliant mechanisms in which flexible links undergo coupling

Co
In this section, a brief introduction to the ‘principal axes
constraints, such as Festo’s delicate Bionic Tripod 3.0 [30]. decomposition’ of spatial compliance matrices is presented.
On the other hand, most of the aforementioned approaches Based on this idea, an approximation approach is established
were particularly developed for planar compliant mecha- to characterize the force-deflection relation of small segment

ot
nisms. And only few of them [24] can deal with the large elements discretized from spatial flexible rods. The main fea-
deflection problems of spatial flexible links. ture of the proposed stiffness modeling method is the use of
tN
In this paper, a general discretization-based approach is a mechanism/robotics way, rather than the mechanics ones,
proposed for the large deflection problems of spatial flexi- to solve the large deflection problems.
ble rods. Based on the principal axes decomposition of spa-
rip

tial compliance matrices [31], a particular type of elements


is established to approximate the force-deflection behavior 2.1 Introduction to the Principal Axes Decomposition
of discretized small segments via 6-DOF serial mechanisms. The principal axes decomposition [31] is inspired by the
sc

To some extent, such an element can be regarded as a spe- synthesis of spatial robot compliance using springy elements.
cial kind of ‘pseudo-rigid-body model’, in which the pas- In these approaches, parallel or serial elastic mechanisms are
sive elastic joints are associated with the segments’ physi- utilized to realize a given spatial stiffness/compliance matrix.
nu

cal appreciations, such as the concepts of ‘center of compli- The motivation of the principal axes decomposition is to in-
ance’, ‘wrench-compliant axes’ and so on. As a result, serial troduce such kind of technique for the structural compliance
Ma

hyper-redundant rigid-body mechanisms with passive elas- modeling of flexible links.


tic joints can be constructed for spatial flexible rods. Then, For a general spatial elastic system, as shown in Fig. 1,
the large deflection problems can be transformed to the de- the linear approximation to its force-deflection behavior can
termination of equilibrium configurations of the equivalent be characterized by its 6 × 6 compliance matrix which maps
ed

hyper-redundant mechanisms under various load/geometric an applied wrench into a twist deformation as
constraint conditions.
The benefit of such an approach is that the large deflec-
pt

" # " #" #


tion problems of flexible links can be solved in a more mech- ω A BT m
X = CF ⇒ = (1)
anism/robotics way, rather than elastic mechanics. Thus, the v B D f
ce

knowledge of robot kinematics/statics can be used to iden-


tify the equilibrium configurations of the equivalent hyper-
Ac

redundant mechanisms under external load conditions and/or where F ∈ R6×1 denotes the applied wrench (a force screw)
geometric constraints. To make the derivation concise, the whose moment m ∈ R3×1 and force f ∈ R3×1 are expressed
product-of-exponential (POE) formula is adopted to model in Plücker’s axis coordinates. X ∈ R6×1 is the corresponding
the motion/force transmission, such that the closed-form so- twist deformation with its angular and linear components,
lution to the mechanism’s overall stiffness matrix can be ob- ω ∈ R3×1 and v ∈ R3×1 , assembled in the form of Plücker’s
tained in a straightforward manner. Hence, the correspond- ray coordinates.
ing equilibrium problems can be solved efficiently by means In relation (1), C ∈ R6×6 denotes the spatial compliance
of the gradient-based searching algorithms. Experiments are matrix which is symmetric and positive semi-definite (PSD).
conducted on an apparatus for measuring the force-deflection The 3 × 3 blocks A, D and B are its rotational, translational
relation of flexible links. And the results verified the correct- and coupling aspects, respectively. It is easy to validate that

ASME JMR-17-1331, Chen 2

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME According to reduced form (3), the spatial compliance
matrix can be decomposed into the sum of a set of rank-one
components as

   T    T 6
I3 I3 03 03
C= A + Λ = ∑ ci t i tTi (4)
B B U U i=1

where D = U Λ UT = ∑3j=1 λ j u j uTj corresponds to the eigen


decomposition of the symmetric PSD aspect D. λ j and u j are
the corresponding eigenvalue and eigenvector, respectively.

d
And t i ∈ R6×1 are unit joint twists, given by

ite
  
ei
, ci = ai , i = 1, 2, 3

ed
Fig. 1. Principal axes decomposition of a general compliance. 
 b

i
ti =   (5)
 03 , ci = λi−3 , i = 4, 5, 6

A ∈ R3×3

both the rotational aspect and the translational one

py
u i−3
D ∈ R3×3 are symmetric and PSD matrices.
It should be noted that, in the linearized force-deflection

Co
relation (1), the twist X, wrench F and compliance matrix C where bi denotes the i th column of B. ai is the i th diagonal
are all represented with respect to the same coordinate frame. entry of A. e1 = [1, 0, 0]T , e2 = [0, 1, 0]T and e3 = [0, 0, 1]T
In other words, their components are all related to the same are unit vectors.
reference point and expressed in the same frame. Based on the rank-one decomposition (4) of compliance

ot
Since Plücker’s axis and ray coordinates are employed matrix, the force-deflection relation of the elastic system can
to represent the twist and wrench quantities, respectively, the be approximately characterized via a serial 6-DOF passive
tN
congruence transformation [32] should be utilized to map the mechanism, as illustrated in Fig. 1, by means of successively
6 × 6 spatial compliance matrix from one coordinate frame connecting the joints t i with compliance ci .
to another as It is known that the first set of decomposed twists relate
rip

to three orthogonal helical joints parallel to the coordinate


X0 = C0 F0 ⇒ C0 = Hg C HTg (2) axes of {P}, respectively. The positions and pitches of these
sc

joints can be determined as

where X0 = Hg X and F0 = H∗g F correspond to the change-


nu

of-coordinates of twists and wrenches, respectively, between r i = ei × b i , hi = eTi b i (6)


frames. Here, Hg ∈ R6×6 and H∗g ∈ R6×6 are the adjoint and
co-adjoint operators associated with g ∈ SE(3), given by
Ma

On the other hand, the second set corresponds to three


      orthogonal prismatic joints parallel to the eigenvectors of D,
R p R 0 R p̂R namely u i . It should be noted that all these vector quantities
g= ⇒ Hg = , H∗g =
03 1 p̂R R 03 R are expressed with respect to the principal coordinate frame
ed

{P}, because the reduced compliance matrix C (3) has been


mapped to it via the congruence transformation.
where R ∈ SO(3) and p ∈ R3×1 represent the relative orien-
pt

tation and position of the two frames. p̂ ∈ so(3) is the 3 × 3


skew-symmetric matrix associated with p. It is easy to verify 2.2 Mechanism Approximation of Segment Elements
ce

HTg = (H∗g )−1 . For more detail, please refer to [32, 33]. The proposed approach to the large deflection problems
In [31], it has been proved that, for a given compliance, of spatial flexible rods is a discretization-based methodology.
there is a unique coordinate frame, termed {P}, in which the
Ac

To some extent, it can be considered as a particular kind of


compliance matrix can be rewritten in the form of finite-element analysis (FEA) method, in which the segment
elements are modeled as serial elastic mechanisms.
" # " #
A AB 03 03 As shown in Fig. 2, in the proposed approach, a spatial
C = Hg C HTg = + (3) flexible rod is discretized into small segments with the length
BA BAB 03 D
of δ . In each segment, two local frames, denoted {Li } and
{Li+1 } respectively, are constructed at the proximal and dis-
where the reduced A becomes a diagonal PSD matrix. The tal ends. Since the length of these segment elements is small,
corresponding B turns into a symmetric one. The reduced the force-deflection behavior can be modeled according to
translational aspect D still keeps symmetric. linear elasticity [34]. Then, the structural compliance matrix

ASME JMR-17-1331, Chen 3

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018
of the by ASME
segment elements yields to the form of

 δ δ 2 
EIxx 0 0 0 − 2EIxx
0
δ2
 
 0 δ
 EIyy 0 2EIyy 0 0 

 
 0 δ
0 GIzz 0 0 0 
Ci =  (7)
 
δ2 δ3

 0 0 0 0 
 2EIyy 3EIyy 
 
 δ2 δ3
 − 2EIxx 0 0 0 0 

3EIxx
δ
0 0 0 0 0 EA

d
ite
where A represents the area of cross section, Ixx , Iyy , and Izz
are the three principal area moments. While E and G are the
elastic and shear modulus, respectively.

ed
Based on the principal axes decomposition presented in
Sec. 2.1, the principal frame, termed {Pi }, of the element’s
Fig. 2. Mechanism approximation of the segment elements.

py
structural compliance can be specified readily in the middle
of segment, as shown in Fig. 2. Then, (7) can be simplified
to a diagonal matrix with respect to {Pi } as
Particularly, in the case of spatial thin and flexible rods, the

Co
 effects of shearing and elongation can be neglected [1, 10],
"
Ai 03
#
 Ai = diag( EIδ , EIδ , GIδ ) so that only the three orthogonal revolute joints are required
xx yy zz
Ci = ⇒ (8) for the mechanism approximation of the discretized segment
03 Di  D = diag( δ 3 , δ 3 , δ )

ot
i 3EIyy 3EIxx EA elements.
tN
As a result, the decomposed joint twists and compliance 2.3 Force-Deflection Relation of Segments
coefficients of the discretized segment elements can be given Since the segment elements have been modeled as serial
directly as elastic mechanisms, the calculation of deformations can be
rip

transformed to determining the equilibrium configurations


        of their equivalent mechanisms under specific load condi-
 t i,2 = e2
e1 e3
i,3 =
t
i,1 =
t
  
tions and/or geometric constraints. In such a scenario, tech-
sc

, 03 03 , 03

c = δ c = δ
  δ
niques in robot kinematics/statics can be employed to charac-
i,1 EIxx i,2 EIyy
c
i,3 = GIzz terize the force-deflection relation of those small segment el-
nu

       (9) ements. In this paper, the POE formula [32,35,36] is utilized


 t i,4 = 03  t i,5 = 03
 
03
to establish the motion/force transmission for those passive
  
i,6 =
t
e1 , e2 , e3 elastic mechanisms.
Ma

 c = δ3
  δ3
 δ
ci,5 = 3EI ci,6 = Using the POE formula, the relative pose of {Li+1 } with
 
i,4 3EIyy xx EA
respect to {Li } can be represented as
From (9), it is known that the pitches of t i, j , j = 1, 2, 3
ed

vanish, so the helical joints in the serial elastic mechanisms gi (θθ i ) = exp(t̂ i,1 θ i,1 ) · · · exp(t̂ i,6 θ i,6 ) g i,0 (10)
degenerate into revolute ones. Moreover, all these revolute
joints pass through the origin of {Pi } and keep parallel to the
pt

corresponding coordinate axes, so do those prismatic joints where t̂ i, j ∈ se(3), j = 1, · · · , 6 denotes the standard represen-
t i, j , j = 4, 5, 6. Further, this property of joint distribution is tation of the joint twists with respect to {L i }. Here, the joint
ce

independent of the segment length. In other words, using the variables θ i = [θ i,1 , · · · , θ i,6 ]T correspond to the deflections
proposed approach, a segment of an arbitrary length can be of the elastic joints. g i,0 = gi (0) represents the initial pose of
{Li+1 } with respect to {L i } in the mechanism’s undeformed
Ac

approximately replaced by an elastic mechanism consisting


of three revolute and three prismatic joints, respectively. And configuration.
all these joints pass through the mid-point of the segment and Then, the Jacobian matrix of the elastic mechanism can
parallel to its principal axes of area moments. be derived readily as
Further, these elastic joints could be related to effects of
particular deformations of the segment elements. Obviously, Ẋi = (ġ i g−1 ∨
 0
θ i ) θ̇θ i ⇒ Ji = ti,1 0
, · · · , ti,6

i ) = Ji (θ (11)
the three revolute joints correspond to the bending about the
section plane and the torsion along the axis, respectively. On
the other hand, the prismatic joints are associated with the where Ẋi ∈ R6×1 denotes a twist deformation of {Li+1 } with
shear and extension/compression deformations of segments. respect to {Li } induced by the deviation of joint variable θ̇θ i .

ASME JMR-17-1331, Chen 4

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME be constructed by connecting the adjacent segments in serial.
It should be noted that all the joints in this hyper-redundant
mechanism are passive and of invariant compliances.
Using the POE formula, the pose of the hyper-redundant
mechanism’s tip frame, termed {T}, can be represented as

gst (θθ ) = g0 g1 (θθ 1 ) · · · gn (θθ n )


n  (14)
= g0,0 ∏ exp(t̂ i,1 θ i,1 ) · · · exp(t̂ i,6 θ i,6 ) g i,0
i=1

d
where g 0,0 ∈ SE(3) is the constant pose of {L1 } with respect
to the system’s inertial frame, denoted by {S}.

ite
In (14), the joint twists ti, j are represented in their local
frames {Li }. Using adjoint transformation, they can be uni-

ed
Fig. 3. Hyper-redundant mechanism approximation of flexible rods. formly mapped to {S}. Then, the local POE formula can be
rewritten in the form of global one as
The components of Jacobian, ti,0 j , j = 1, · · · , 6 corresponds

py
N
to the joint twists in the current configuration, which can be
gst (θθ ) = ∏ exp(ξˆ k θ k ) g st,0 (15)
derived as ti,0 j = Hexp(t̂i,1 θi,1 )··· exp(t̂i, j−1 θi, j−1 ) ti, j . Here, H(·) is

Co
k=1
the adjoint operator, defined in (2), associated with the POE:
exp(t̂i,1 θi,1 ) · · · exp(t̂i, j−1 θi, j−1 ) ∈ SE(3).
where θ = [θ1 , · · · , θN ]T ∈ RN×1 denotes the joint variables
As it is known, the static equilibrium configuration can
of the hyper-redundant mechanism. ξ k = Hg0,0 ··· gi−1,0 ti, j are

ot
be achieved once the load-induced reaction torques/forces in
the joint twists represented in {S}, obtained by the identity
the elastic joints are balanced by their deflections, namely
ξˆ k ≡ g0,0 · · · gi−1,0 t̂i, j g−1 −1
i−1,0 · · · g0,0 . g st,0 = g0,0 g1,0 · · · gn,0
tN
relates to the initial pose of {T} with respect to {S}. Here,
τ i = Kθ i θ i − JTi Fi → 0 (12) the subscript ‘k’ is determined by i and j as k = 6(i − 1) + j
i = 1, · · · , n, j = 1, · · · , 6. N = 6n is the total number of joints.
rip

where Fi is the applied wrench with respect to {Li }. And the Comparing (15) to (10), it is known that the hyper-
transpose of Ji corresponds to the force Jacobian which maps redundant mechanism equivalent to the whole rod shares the
the wrench Fi to the joint reactions. Kθ i = diag( c1i,1 , · · · , c1 )
sc

i,6
same form in kinematics as discretized segments. Therefore,
relates to the stiffness of elastic joints. the same approach can be employed for the overall stiffness
By differentiating (12) with respect to θ i , the overall modeling of the whole equivalent mechanism.
nu

joint stiffness matrix of the equivalent elastic mechanism can Analogous to (11), the system’s Jacobian matrix of the
be derived as hyper-redundant elastic mechanism can be represented as
Ma

∂τi  0 0 0 
Ki = = Kθ i + KJi + KFi (13) Jθ = ξ 1 , ξ 2 , · · · , ξ N ∈ R6×N (16)
∂θi
ed

0
∂ JT where the components ξ k = Hexp(ξˆ θ ) ··· exp(ξˆ θ ) ξ k are
where KJi = − ∂ θii Fi ∈ R6×6 and KFi = −JTi ∂∂ θFii ∈ R6×6 1 1 k−1 k−1

are two configuration-dependent items [37]. The details of the joint twists in the current configuration and represented
with respect to {S}. Similarly, JTθ is just the force Jacobian
pt

derivation are given in Appendix A.


relating the external wrench to joint reactions.
Likewise, the overall joint stiffness of the whole elastic
ce

3 Force-Deflection Modeling of Spatial Flexible Rods mechanism can be derived as


In this section, the force-deflection relation of the whole
Ac

flexible rods is modeled using the mechanism approximation ∂τ


K= = Kθ + KJ + KF (17)
method presented in Sec. 2. By means of replacing the dis- ∂θ
cretized segments with their equivalent elastic mechanisms,
a serial hyper-redundant mechanism, as shown in Fig. 3, can
where τ = Kθ θ − JTθ F ∈ RN×1 corresponds to the deviation
be constructed to model the large deflection problems.
of the joints’ deflection forces/torques from the reaction ones
induced by external loads. F is the external wrench applied
3.1 Kinematics and Stiffness Modeling at the tip frame {T}.
Let n be the number of discretized elements of a spatial In the overall stiffness (17), the configuration-dependent
flexible rod. Then, a serial 6n-jointed passive mechanism can item KJ ∈ RN×N can be derived in the same way as those in

ASME JMR-17-1331, Chen 5

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by
segment ASME following the derivation in Appendix A,
elements,
given by

 
0 0 ··· 0
 FT Z 0 ξ 0 0
 .. .. 
∂ JTθ ξ1 2
. . 
KJ = F=  (18)
 
∂θ .. .. ..

 . . . 0 

0 0
F Zξ 0 ξ N · · · FT Zξ 0
T
ξN 0
1 N−1

where Zξ 0 ∈ R6×6 is the adjoint representation of joint twist

d
i
with respect to {S}.

ite
On the other hand, the situation of KF = JTθ ∂∂ θF ∈ RN×N
will be different in terms of various external load conditions.
If a non-following wrench is applied at the tip-frame of the

ed
Fig. 4. Experimental setup for the large deflection problems.
flexible rod, this item will vanish since F is independent of
end configuration, namely ∂∂ θF ≡ 0. Otherwise, it should be

py
taken into account during the determination of mechanism’s a gradient-based searching algorithm is established based
equilibrium configurations. And the concrete situations will on the motion/force transmission property of the equivalent
be discussed in the following sections. hyper-redundant mechanism.

Co
Based on the kinematics/statics presented in this section, In this case, a multi-objective function which combines
two kinds of static equilibrium problems which correspond both the tip-frame pose and the joint reaction forces/torques,
to typical load/constraint conditions of spatial flexible rods, is defined for the iterative searching algorithm as

ot
will be investigated in the rest of this section. The first prob-
lem relates to the condition that a specific external wrench log(g−1
" #
θ ))∨
t g(θ
is exerted at the tip frame of the equivalent hyper-redundant min c(θθ , F) = ∈ R(6n+6)×1 (20)
tN
mechanism. In the second case, on the contrary, the tip frame Kθ θ − JTθ F
of the flexible rod is restricted to some particular geometric
constraints.
rip

where g t ∈ SE(3) represents the target pose for the tip-frame.


And (log(g−1 t g(θθ )))∨ ∈ R6×1 denotes the twist deviation of
3.2 Equilibrium Upon External Loads current pose from the target one.
sc

In the case of applying an external load, the equilibrium The objective function (20) requires to identify a proper
problem of the flexible rod corresponds to statics analysis of combination of the joint variable θ and external wrench F,
such that the tip frame approaches its target pose, meanwhile,
nu

the corresponding hyper-redundant elastic mechanism. And


according to the overall joint stiffness (17), a gradient-based the joints’ reaction forces/torques are balanced by their elas-
searching algorithm can be utilized to iteratively identify the tic deflections.
Ma

equilibrium configuration, as Thus, the gradient of the objective function (20) can be
derived as

θ j+1 = θ j + (K j )−1 τ j (19)    


∂c ∂c Jθ 0
ed

∇= , = ∈ R(N+6)×(N+6) (21)
∂θ ∂F Kθ + KJ −JTθ
where τ j is the bias of joint deflection forces/torques from
pt

their reaction ones induced by the external load in the current


iterative step j. where KJ and Kθ are defined in (17).
Accordingly, the update theme of the joint variables and
ce

According to the update theme (19), the value of joint


external load can be represented as
variable θ will be repeatedly updated until it converges to
some stable magnitude and the joints’ bias forces/torques τ
Ac

(
approaches zero. And the resultant configuration is simply 
δθ

θ j+1 = θ j + δ θ j
−1
the solution to the corresponding equilibrium problem. =∇ c⇒ (22)
δF F j+1 = F j + δ F j

3.3 Equilibrium Under Geometric Constraints


In such a large deflection problem of flexible rod, the tip 4 Experiment Verification
frame is restricted to a prescribed pose. Thus, not only the In this section, an experimental apparatus is established
static equilibrium configuration, but also the corresponding to validate the correctness and effectiveness of the propose
external load of the hyper-redundant elastic mechanism are approximation approach to the large deflection problems of
required to be determined. And analogous to the above case, spatial flexible rods.

ASME JMR-17-1331, Chen 6

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME
Table 1. Specifications of the tested spatial flexible rod.

parameters values

radius r = 1.0 mm
length L = 950 mm
elastic modulus E = 56.0 GPa
shear modulus G = 21.5 GPa
second moments of area Ixx = Iyy = 41 π mm4
polar moment of inertia Izz = 12 π mm4

d
Fig. 5. Experimental result of the flexible rod with large deflection.

ite
As shown in Fig. 4, to apply/measure external loads and

ed
geometric constraints to the spatial flexible rod, the experi-
mental setup includes the following parts:

py
1) A spatial flexible rod: a long and thin glass-fiber link is
adopted as the spatial flexible rod, whose specifications
are given in Table 1;

Co
2) A manipulation robot: a KUKA-KR10-R900-sixx robot
is used to apply external loads or geometric constraints
to the tip-frame of the flexible rod. The payload of the

ot
Fig. 6. Accuracy evaluation of the proposed method.
robot is 10 kg and its repeated positioning precision is
±0.03 mm;
tN
3) A force sensor: a six-axis force-torque sensor (the ATI On the other hand, the measured 1 and predicted reaction
Gamma SI-130-10) is mounted between the robot and torques/forces are given by
flexible rod to measure the reaction forces/torques. The
rip

   
sensing ranges are 130 N for Fx , Fy , 400 N for Fz and 0.085 0.089
 −0.002   0.003 
10 Nm for Tx , Ty , Tz . While the resolution is 1/40 N for    
 −0.022   −0.030 
Fx , Fy , 1/20 N for Fz and 1/800 Nm for Tx , Ty , Tz .
sc

Fm = 

 , F p =  −0.423 
   (23)
4) A vision system: a FARO-Edge-ScanArm HD is used  −0.472   
 −0.385   −0.379 
to measure the exact deflection of rod in its equilibrium
nu

0.334 0.271
configuration. The measurement accuracy is ±25µm.
where the torque and force units are Nm and N, respectively.
Since both tip-frame pose and external load of the flexi-
Ma

According to (23), the deviations of reaction torque and


ble rob can be measured in the apparatus, the two equilibrium
force between the estimated and measured results can be
problems discussed in Sec. 3.2 and 3.3, namely the situations
obtained as δ m = [0.004, 0.005, −0.008]T Nm and δ f =
under geometric constraints and upon external loads, can be
[0.049, 0.006, −0.063]T N, respectively. Although the pre-
ed

validated simultaneously by one uniform experiment.


diction errors are slightly larger than the sensing resolution,
In the physical experiment, the tip frame of the flexible
they are still in an acceptable range since many aspects, such
rod is dragged to a prescribed pose via the robot, as shown in
pt

as the gravity of rod, the misalignment of force sensor and so


Fig. 5 The rod’s deflected configuration is identified by the
on, are not taken into account in the physical experiment.
ScanArm. Meanwhile, the reaction force/torque is measured
ce

by the ATI sensor. The measured equilibrium configuration


of the deflected rod is shown in Fig. 5, where the calculated 5 Application to A Planar Compliant Mechanism
Ac

result is also illustrated as a comparison. Based on the above discussion, this section presents an
From the figure, it is known that the result obtained by application of the proposed method to the kinematics/statics
the proposed approximation method is very close to the one analysis of a planar 2-DOF compliant parallel manipulator.
achieved through the physical experiment. The standard 3D As exhibited in Fig. 7, the mechanism constitutes two planar
(RMS) deviation between calculated and measured results is parallelogram linkages consisting of rigid links and flexible
2.271 × 10−4 m with the maximums of −1.737 × 10−3 m and rods, respectively.
1.376 × 10−3 m, respectively, as shown in Fig. 6. It should
be noted that the measurement is given in the form of point
cloud of the rod’s surface, so that this evaluation is based on 1 It
should be noted that the reading of ATI sensor fluctuates even when
the distances of those points to the calculated ones. the flexible rod reaches its equilibrium configuration. Hence, the result in
(23) is a mean value of the sensor reading within a time interval.
ASME JMR-17-1331, Chen 7

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME The corresponding joint twists can be given by

" #
e3
ξi = , i = 1, · · · , n (24)
− 1n (i − 21 )l e2 × e3

where l represents the rod’s effective length, which is also an


variable in the optimization problem.
For such a particular equilibrium problem, the objective
function can be represented as

d
(log(g−1 ∨
 
t g))

ite
min c(x) =  Kθ θ − JTθ F  ∈ R(n+7)×1 (25)
 

fy

ed
Fig. 7. A planar 2-DOF translational compliant mechanism. where x = [ l, θ , F ]T ∈ R(n+7)×1 denotes the variables of the

py
optimization problem, including the effective length l ∈ R,
the joint deflections θ ∈ Rn×1 and external wrench F ∈ R6×1 .
As shown in the figure, the rigid limb is actuated by a gt ∈ SE(3) is the configuration of mid-platform actuated by

Co
revolute joint, such that the mid-platform can translate along the rigid limb. Here, fy = eT5 F corresponds to the component
a circular trajectory. Meanwhile, the length of flexible limb of F in y-direction and e5 = [0, 0, 0, 0, 1, 0]T ∈ R6×1 .
is controlled by a prismatic joint. Thus, the height of end- Since the length of segments, as well as the joint twists,
effector can be actively adjusted to compensate the vertical

ot
changes during the iterative searching process, the influence
motion of rigid limb and deflection of flexible one. Hence, of the rod’s effective length should also be taken into account
this compliant parallel manipulator can perform planar pick-
tN
in the gradient of (25). Then, we have
and-place tasks by controlling the active joints properly.
It is known that one of the key issues for controlling such  
Jl Jθ 06
a compliant mechanism is to accurately calculate the large
rip

 
∂c ∂c ∂c
∇= , , =  Kl Kθ + KJ −JTθ  (26)
 
deflection of the flexible limb. In this case, the equilibrium ∂l ∂θ ∂F
problem is slightly different from those in the above section. 0 01×n eT5
sc

It relates to a hybrid situation which combines both specific


geometric constraint and applied wrench. On one hand, the
where Jl ∈ R6×1 is the coefficient relating the small change
orientation and horizontal position of the flexible linkage are
nu

of rod length to the twist deviation of pose at the tip frame.


constrained by the mid-platform once the actuation of rigid
Kl ∈ Rn×1 corresponds to the matrix of joint stiffness due to
limb is specified. On the other hand, the flexible limb can
the change of rod length. The derivation of Jl and Kl is given
freely translate along the linear guides on the mid-platform.
Ma

in Appendix B.
And this implies that the vertical external force reacted in the
It is worth noting that in planar cases three components
linear guides should be vanished when the flexible limb is in
in both (log(g−1 ∨
t g)) and F are always zeros. Therefore, the
its static equilibrium state.
associated components in the objective function c(x), as well
ed

The difficulty of this equilibrium problem is caused by as the corresponding rows and columns in ∇, are eliminated
the change of effective (deformed) length of the flexible rod. from the optimization model in order to make the searching
To overcome this issue, a variable segment length is adopted
pt

algorithm more stable.


in the proposed method during the searching process. Thus, Then, the update theme for the variables in this hybrid
the length of segments can be expressed as δ = 1n l, where l equilibrium problem can be represented as
ce

denotes the effective length of the flexible limb. As a result,


the equilibrium problem can be stated: to determine a proper
x j+1 = x j + ∇−1 c j (27)
Ac

length of flexible limb for a given pose of its tip frame, such
that in the equilibrium configuration the corresponding force
component in vertical direction vanishes. which will be iteratively repeated until the objective function
It should be noted that, in planar cases, only the bending c approaches zero and the variable x converges stably.
effect within the plane is taken into account in the deflection As a result, the effective length of the flexible limb and
of flexible rods. In other words, in each segment, only the the concrete deflection in its equilibrium configuration can
revolute joint perpendicular to the plane is taken into account be simultaneously obtained in terms of the resultant l and θ .
in the proposed approximation method. Meanwhile, the corresponding constraint force reacted in the
As a result, a planar n-R hyper-redundant mechanism is mid-platform can also be identified according to F following
obtained for large deflection calculation of the flexible limb. the proposed update theme.

ASME JMR-17-1331, Chen 8

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME

d
ite
ed
Fig. 8. Verification via the simulation results of FEA model. Fig. 9. Residual errors of the proposed method from FEA model.

py
According to the identified equilibrium configuration of
flexible limb, the kinematics and statics of the studied planar
compliant mechanism can be accomplished. Thus, the exact

Co
position of end-effector and the reaction forces in joints can
be predicted for a given set of actuations. On the other hand,
given a prescribed position for the end-effector, the required

ot
inputs of the mechanism’s active joints can also be achieved
by compensating the change of limb’s effective length.
tN
It is known that the total length of flexible limb does not
influence the configuration of its deflected part, but imposes
a purely vertical displacement to the end-effector. Therefore,
rip

the effective length of flexible limb is only determined by the


rigid limb’s motion, namely the position of the mid platform.
Thus, in the studied example, the total length of the flexible
sc

limb is fixed to l0 = 0.115 m. While the link length of rigid


one is set to r0 = 0.075 m, which also equals to the radius of Fig. 10. Consuming time versus computational accuracy.
the mid-platform’s circular constraint, as indicated in Fig 8.
nu

And same kind of flexible rods as the one tested in Sec. 4,


are used for this mechanism. From the results, it is known that the residual errors in
vertical direction increase as the horizontal displacement of
Ma

Figure 8 delineates the equilibrium configurations of the


flexible limb corresponding to different inputs of rigid limb, the mechanism’s end-effector increases. However, the values
the range of which is set q1 ∈ [ 0, 0.8 ] rad. As an verification, of those residual errors are relatively small, comparing to the
a corresponding FEA model is build up using the commercial end-effector displacements. When the number of discretized
ed

software Abaqus/Standard 6.13 and the simulation results are elements is larger than 50, the maximal position errors will
also drawn in Fig. 8. It should be noted that since the two decrease to less than 4.0 × 10−4 m, which is less than 5.0h
rods in the flexible limb are identical to each other, only one of the end-effector’s vertical movement.
pt

of them is illustrated in the figure. For the sake of clarity, To validate the computational efficiency of the proposed
the rod’s deflected configuration is moved to the middle of approximation approach, the consuming time corresponding
ce

symmetric mechanism. The corresponding trajectory of the to different discretization of the flexible limb is investigated.
mechanism’s end-effector is drawn in Fig. 8, where the FEA The results are drawn in Fig. 10, with the comparison of the
Ac

simulation results are also given as a comparison. corresponding computational accuracy.


From the figure, it can be seen that the results calculated From the figure, it is apparent that the consuming time
by the proposed approximation method are reasonably close increases significantly as the number of the rod’s discretized
to those obtained from the FEA simulation, even in large de- elements increases. On the other hand, the improvement of
flected configurations. The residual of the end-effector tra- computational accuracy is very limited when the discretized
jectory in y-direction is illustrated in Fig. 9. Here, the posi- elements are larger than 50. Therefore, for this particular
tion error is defined by the deviation of the results obtained example, the flexible limb can be discretized into 50 segment
by the proposed method from the FEA model. In addition, elements in the proposed approximation method, such that
the effect of different discretization of segment elements is the computational burden can be reduced to a small range,
also compared in the figure. but without influencing the calculation accuracy.

ASME JMR-17-1331, Chen 9

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME
It should be noted that the established algorithm for this According to (13), it can be represented as
studied example is implemented in MATLAB on a personal
computer with a processor of 2.7 GHz. Since the gradient of
FTi ∂ t0i,1 FTi ∂ t0i,1
 FT ∂ t0   
objective function has been explicitly derived in the model, i i,1
∂ θi,1 ··· ∂ θi,6
∂θi
the rate of convergence is relatively high. On the contrary, it    
∂ JT
 
KJi = − ∂ θii

Fi = −  .. 
 = − .. .. ..
takes ABAQUS about 842 seconds to achieve the solution to

 .   . . . 
the same large deflection problem. Moreover, the simulation    
FTi ∂ t0i,6 FTi ∂ t0i,6 FTi ∂ t0i,6
 
of FEA model is unstable due to the existence of hard contact ∂θi ∂ θi,1 ··· ∂ θi,6
between the flexible rods and the linear guides on the rigid (28)
mid-platform. where the entries can be derived as
According to the above analysis, it is known that, in the

d
studied planar compliant mechanism, the effective length of ∂
 
the flexible limb can be estimated accurately and efficiently Hexp(t̂i,1 θi,1 )··· exp(t̂i, j−1 θi, j−1 ) ti, j

ite
∂ θi,k
via the proposed method. Accordingly, the exact vertical po- = Hexp(t̂i,1 θi,1 )··· exp(t̂i,k−1 θi,k−1 ) ∂ −1
∂ θi,k Hexp(t̂i,k θi,k ) Hexp(t̂i,k θi,k )
sition of the mechanism’s end-effector can be predicted pre-

ed
cisely and compensation to the flexible limb deflection can H−1 Hexp(t̂i,1 θi,1 )··· exp(t̂i, j−1 θi, j−1 ) ti, j
exp(t̂ i,1 θi,1 )··· exp(t̂i,k−1 θi,k−1 )
be implemented for prescribed pick-and-place tasks.
= Hexp(t̂i,1 θi,1 )··· exp(t̂i,k−1 θi,k−1 ) Zti,k H−1 t0i, j

py
exp(t̂ i,1 θi,1 )··· exp(t̂i,k−1 θi,k−1 )

= Zt0 t0i, j
i,k
(29)

Co
6 Conclusions −1
where ∂
∂ θi,k Hexp(t̂i,k θi,k ) Hexp(t̂i,k θi,k ) = Zti,k ∈ R6×6 corresponds
In this paper, a general discretization-based approach is
proposed for the large deflection problems of spatial flexible to the adjoint representation of joint twist t̂i,k , which satisfies
the following identity [33] as

ot
rods in compliant mechanisms. Based on the principal axes
decomposition of compliance matrices, a specialized type of
tN
elements is established to approximate the force-deflection "
ω
# "
ω
ω̂ 03
#
relation of small segments with 6-DOF elastic mechanisms. ξ= ⇒ Zξ = ⇒ Hg Zξ H−1
g ≡ ZHg ξ ≡ Zξ 0
Using the proposed approach, the large deflection problems v v̂ ω
ω̂
rip

of flexible rods can be transformed to the kinematics/statics


analysis of hyper-redundant mechanisms with passive elastic
joints. Two kinds of typical equilibrium problems of flexible The adjoint representation of a twist is also known as the
cross product operator in screw theory [38] and Lie bracket
sc

rods have been intensively discussed in the proposed frame-


work. And the effectiveness of the proposed method to these in se(3) [32], which satisfies Zξ 1 ξ 2 = ξ 1 ×ξξ 2 = ([ξˆ 1 , ξˆ 2 ])∨
equilibrium problems has been validated via physical exper- for any two joint twists ξ 1 and ξ 2 and Zξ 1 ξ 2 = −Zξ 2 ξ 1 .
nu

iments carried out on a particularly constructed apparatus. In According to (29), it is known that the entries vanish in
the end, a planar 2-DOF compliant mechanism is designed as the case of k ≥ j. Thus, the stiffness item KJi in (13) can be
a potential application of the established solving algorithm. obtained readily as
Ma

The kinematics of the compliant mechanism can be achieved


efficiently using the proposed approach and the correctness  
of the obtained results have been verified by means of a cor- 0 0 ··· 0
ed

responding FEA model.


 .. .. 
 Fi Zt0i,2 t0i,1
 T .
0 . 

KJi =   ∈ R6×6 (30)
 
.. .. ..
pt

. .
 
 . 0 
 
T 0 0
Acknowledgements Fi Zt ti,1 ··· FTi Zt0 t0i,5 0
ce

i,6 i,6

This work was jointly supported by the Natural Science


Foundation of China under Grant 51505279 and 51421092, which is a 6 × 6 lower triangular matrix.
Ac

and the National Basic Research Program of China (973 Pro- On the other hand, the item KFi is caused by the change
gram) under Grant 2014CB046600. of external loads due to the deviations of joint variables. So
it replies on the loading conditions. If the applied wrench is
independent of the mechanism configuration, this item yields
to zero, namely ∂∂ θFii = 0.
APPENDIX A Derivative of Configuration-Dependent Here, a specific case that Fi keeps constant in the distal
Stiffness Items frame {Li+1 } is provided as an example. Let Fi = H∗gi Fi,0 ,
The first item KJi represents the effects of small changes where Fi,0 denotes a constant wrench in {Li+1 } and H∗gi is
in force transmission due to the deviations of joint variables. the co-adjoint operator between {Li+1 } and {Li }. Then, this

ASME JMR-17-1331, Chen 10

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018
item can by
be ASME
represented as Meanwhile, the partial differentiation of joint stiffness
with respect to the length of rod can be derived readily in a
  straightforward manner as
−JTi ∂∂θ i H∗gi Fi,0
h ∂ H∗ ∂ H∗gi
i
= −JTi ∂ θ gi (H∗gi )−1 H∗gi Fi,0 , · · · , ∗ −1 ∂ Kθ 1
∂ θi,6 (Hgi ) H∗gi Fi,0 Kl = θ = − Kθ θ (33)
i,1
h i ∂l l
T T
= JTi Zt0i,1 Fi , · · · , Zt0i,6 Fi ∈ R6×6
(31) where Kθ = nEIl In is just proportional to the n-order identity
where the differentiation of co-adjoint operator H∗gi can be matrix In .
derived as Based on the above derivations, all components of the
objective function’s gradient (26) can be obtained in closed-

d
form in terms of the current values of variables. Then, the
∂ H∗gi ∂ H−T
T

ite

gi ∂ Hgi −1
(H∗gi )−1 = HTgi =− H = −ZTt0 equilibrium configuration can be identified by repeating the
∂ θi,k ∂ θi,k ∂ θi,k gi i,k
update them (27) until it converges.

ed
which relates to the inverse transpose of the adjoint represen-
tation of t0i,k . References

py
[1] Howell, L., 2001. Compliant Mechanisms. John Wiley
& Sons, New York, NY.
[2] Saxena, R., and Saxena, A., 2009. “Design of elec-

Co
APPENDIX B Derivation of Jacobian and Stiffness due
to the Change of Rod Length trothermally compliant mems with hexagonal cells us-
The item Jl represents the linear approximation between ing local temperature and stress constraints”. ASME
the deviation of tip frame and the small change of rod length. Journal of Mechanical Design, 131(5), p. 051006.

ot
Using the POE-based error model of serial robots [36], it can [3] Asbeck, A., and Cutkosky, M., 2012. “Designing com-
be represented as pliant spine mechanisms for climbing”. ASME Journal
tN
of Mechanisms and Robotics, 4(3), p. 031007.
[4] Edelmann, J., Petruska, A., and Nelson, B., 2017.
(∂ gst g−1
st )
∨ n+1
∂ξ “Magnetic control of continuum devices”. The Interna-
Jl = = ∑ Hexp(ξˆ θ )··· exp(ξˆ θ ) Ai i (32)
rip

∂l i=1
1 1 i−1 i−1 ∂l tional Journal of Robotics Research, 36(1), pp. 68–85.
[5] Marras, A., Zhou, L., Su, H., and Castro, C., 2015.
“Programmable motion of DNA origami mechanisms”.
sc

where ξˆ n+1 = log(g0 ) denotes the twist associated with the Proceedings of the National Academy of Sciences,
initial pose of tip frame. Ai ∈ R6×6 is a polynomial matrix 112(3), pp. 713–718.
of the joint twist Zξ i , given by [6] Midha, A., Norton, T., and Howell, L., 1994. “On the
nu

nomenclature, classification, and abstractions of com-


pliant mechanisms”. ASME Journal of Mechanical De-
(∂ exp(ξˆ i θi ) exp(−ξˆ i θi ))∨ 5
Ma

sign, 116(1), pp. 270–279.


Ai =
∂ξ i
= ∑ (ai, j Zξj−1
i
)
[7] Awtar, S., Slocum, A., and Sevincer, E., 2007. “Charac-
j=1
teristics of beam-based flexure modules”. ASME Jour-
nal of Mechanical Design, 129(6), pp. 625–639.
ed

with the coefficients [8] Frisch-Fay, R., 1962. Flexible bars. Butterworths,
ai,1 = θi Washington D.C.
ai,2 = 12 (4 − θi sin θi − 4 cos θi ) [9] Kimball, C., and Tsai, L., 2002. “Modeling of flexural
pt

beams subjected to arbitrary end loads”. ASME Journal


ai,3 = 12 (4θi − 5 sin θi + θi cos θi ) of Mechanical Design, 124(2), pp. 223–235.
ce

ai,4 = 21 (2 − θi sin θi − 2 cos θi ) [10] Zhang, A., and Chen, G., 2013. “A comprehensive el-
liptic integral solution to the large deflection problems
ai,5 = 12 (2θi − 3 sin θi + θi cos θi )
Ac

of thin beams in compliant mechanisms”. ASME Jour-


According to the definition of joint twists (24), it is easy nal of Mechanisms and Robotics, 5(2), p. 021006.
to obtain that [11] Howell, L., and Midha, A., 1995. “Parametric deflec-
tion approximations for end-loaded, large-deflection
 " # beams in compliant mechanisms”. ASME Journal of
1
03 Mechanical Design, 117(1), pp. 156–165.
 − n (i − 1 )e , i = 1, · · · , n



∂ξ i  2 1
[12] Su, H., 2009. “A pseudo-rigid-body 3R model for
= " # determining large deflection of cantilever beams sub-
∂l  03


 , i = n+1 ject to tip loads”. ASME Journal of Mechanisms and
−e2

Robotics, 1(2), p. 021008.

ASME JMR-17-1331, Chen 11

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c)
[13]2018 by ASME
Chen, G., Xiong, B., and Huang, X., 2011. “Finding tion approach”. Mechanism and Machine Theory,
the optimal characteristic parameters for 3R pseudo- 43(5), pp. 641–658.
rigid-body model using an improved particle swarm op- [28] Odhner, L., and Dollar, A., 2012. “The smooth cur-
timizer”. Precision Engineering, 35(3), pp. 505–511. vature model: An efficient representation of Euler-
[14] Zhu, S., and Yu, Y., 2017. “Pseudo-rigid-body model Bernoulli flexures as robot joints”. IEEE Transactions
for the flexural beam with an inflection point in compli- on Robotics, 28(4), pp. 761–772.
ant mechanisms”. ASME Journal of Mechanisms and [29] Chen, G., and Ma, F., 2015. “Kinetostatic modeling
Robotics, 9(3), p. 031005. of fully compliant bistable mechanisms using Timo-
[15] Yu, Y.-Q., and Zhu, S.-K., 2017. “5r pseudo-rigid-body shenko beam constraint model”. ASME Journal of Me-
model for inflection beams in compliant mechanisms”. chanical Design, 137(2), p. 022301.
Mechanism and Machine Theory, 116, pp. 501–512. [30] Bionic tripod 3.0. https://www.festo.com/.

d
[16] Jensen, B., and Howell, L., 2004. “Identification [31] Chen, G., Wang, H., Lin, Z., and Lai, X., 2015. “The
of compliant pseudo-rigid-body four-link mechanism principal axes decomposition of spatial stiffness matri-

ite
configurations resulting in bistable behavior”. ASME ces”. IEEE Transactions on Robotics, 31(1), pp. 191–
Journal of Mechanical Design, 125(4), pp. 701–708. 207.

ed
[17] Midha, A., Bapat, S., Mavanthoor, A., and Chinta, V., [32] Murray, R. M., Sastry, S. S., and Li, Z. X., 1994. A
2012. “Analysis of a fixed-guided compliant beam with Mathematical Introduction to Robotic Manipulation.
an inflection point using the pseudo-rigid-body model CRC Press, Inc.

py
(PRBM) concept”. ASME Journal of Mechanisms and [33] Selig, J. M., 2005. Geometric Fundamentals of
Robotics(45035), pp. 351–361. Robotics. Springer, New York.
[18] Vogtmann, D., Gupta, S., and Bergbreiter, S., 2013. [34] Archer, R., and Lardner, T., 1978. An Introduction to

Co
“Characterization and modeling of elastomeric joints the Mechanics of Solids. McGraw-Hill series in engi-
in miniature compliant mechanisms”. ASME Journal neering sciences. McGraw-Hill.
of Mechanisms and Robotics, 5(4), p. 041017. [35] Chen, I., Yang, G., Tan, C., and Yeo, S., 2001. “Local

ot
[19] Zhou, L., Marras, A., Castro, C., and Su, H., 2016. POE model for robot kinematic calibration”. Mecha-
“Pseudo rigid-body models of compliant DNA origami nism and Machine Theory, 36(11-12), pp. 1215–1239.
tN
mechanisms”. ASME Journal of Mechanisms and [36] Chen, G., Wang, H., and Lin, Z., 2014. “Determination
Robotics, 8(5), p. 051013. of the identifiable parameters in robot calibration based
[20] Saxena, A., and Ananthasuresh, G., 2001. “Topology on the POE formula”. IEEE Transactions on Robotics,
rip

synthesis of compliant mechanisms for nonlinear force- 30(5), pp. 1066–1077.


deflection and curved path specifications”. ASME Jour- [37] Cutkosky, M., and Kao, I., 1989. “Computing and con-
nal of mechanical Design, 123(1), pp. 33–42. trolling compliance of a robotic hand”. IEEE Transac-
sc

[21] Her, I., 1986. “Methodology for compliant mechanisms tions on Robotics and Automation, 5(2), pp. 151–165.
design”. Ph.d. thesis, Purdue University. [38] Dai, J., 2014. Screw Algebra and Lie Groups and Lie
[22] Campanile, L., and Hasse, A., 2008. “A simple and Algebra. Higher Education Press, Beijing.
nu

effective solution of the elastica problem”. Journal of


Mechanical Engineering Science, 222(12), pp. 2513–
Ma

2516.
[23] Pauly, J., and Midha, A., 2006. “Pseudo-rigid-body
model chain algorithm: Part 1 - introduction and con-
cept development”. In International Design Engineer-
ed

ing Technical Conferences and Computers and Infor-


mation in Engineering Conference, pp. 173–181.
[24] Chase, R., Todd, R., Howell, L., and Magleby, S., 2011.
pt

“A 3-D chain algorithm with pseudo-rigid-body model


elements”. Mechanics Based Design of Structures and
ce

Machines, 39(1), pp. 142–156.


[25] Ma, F., and Chen, G., 2015. “Modeling large planar
Ac

deflections of flexible beams in compliant mechanisms


using chained beam-constraint-model1”. ASME Jour-
nal of Mechanisms and Robotics, 8(2), p. 021018.
[26] Banerjee, A., Bhattacharya, B., and Mallik, A., 2008.
“Large deflection of cantilever beams with geometric
non-linearity: Analytical and numerical approaches”.
International Journal of Non-Linear Mechanics, 43(5),
pp. 366–376.
[27] Lan, C.-C., 2008. “Analysis of large-displacement
compliant mechanisms using an incremental lineariza-

ASME JMR-17-1331, Chen 12

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 byofASME
A list figure captions
Figure 1. Principal axes decomposition of a general compliance.
Figure 2. Mechanism approximation of the segment elements.
Figure 3. Hyper-redundant mechanism approximation of flexible rods.
Figure 4. Experimental setup for the large deflection problems.
Figure 5. Experimental result of the flexible rod with large deflection.
Figure 6. Accuracy evaluation of the proposed method.
Figure 7. A planar 2-DOF translational compliant mechanism.
Figure 8. Verification via the simulation results of FEA model.
Figure 9. Residual errors of the proposed method from FEA model.
Figure 10. Consuming time versus computational accuracy.

d
ite
ed
py
Co
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac

ASME JMR-17-1331, Chen 13

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 byofASME
A list table captions
Table 1. Specifications of the tested spatial flexible rod.

d
ite
ed
py
Co
ot
tN
rip
sc
nu
Ma
ed
pt
ce
Ac

ASME JMR-17-1331, Chen 14

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Journal of Mechanisms and Robotics. Received September 28, 2017;
Accepted manuscript posted February 05, 2018. doi:10.1115/1.4039223
Copyright (c) 2018 by ASME

d
ite
ed
py
Co
Table 1. Specifications of the tested spatial flexible rod.

ot
parameters values
tN
radius r = 1.0 mm
length L = 950 mm
rip

elastic modulus E = 56.0 GPa


shear modulus G = 21.5 GPa
sc

second moments of area Ixx = Iyy = 14 π mm4


polar moment of inertia Izz = 12 π mm4
nu
Ma
ed
pt
ce
Ac

ASME JMR-17-1331, Chen 15

Downloaded From: http://mechanismsrobotics.asmedigitalcollection.asme.org/ on 02/05/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like