Mechanism and Machine Theory 185 (2023) 105311

Contents lists available at ScienceDirect

Mechanism and Machine Theory

journal homepage: www.elsevier.com/locate/mechmt

Research paper

Kinematic Analysis of three degrees of freedom planar parallel

continuum mechanisms
Oscar Altuzarra ∗, Mónica Urizar, Massimo Cichella, Víctor Petuya
Department of Mechanical Engineering, University of the Basque Country UPV/EHU, Bilbao, Spain

ARTICLE INFO ABSTRACT

Keywords: Parallel continuum mechanisms are a type of compliant mechanisms comprising a rigid end-
Parallel continuum mechanism effector connected to a fixed frame by flexible and slender elements in a closed-kinematic
Kirchhoff rod morphology, being their flexibility the source of mobility. Solving the position problem involves
Kinematics
the static equilibrium of the whole device considering the nonlinear large deformations
Deformation modes
of slender elements. Numerical iterative solutions are already available to solve real-time
Instability
simulations. However, an assembled solution, which meets both geometric conditions and static
equilibrium, is needed to start the process. A comprehensive Kinematic Analysis in the classical
sense is not fully established yet. The aim of the paper is to propose a procedure that solves the
full inverse and forward kinematic problems for planar three degrees of freedom systems, hence
obtaining the multiple solutions arising from the closed-loop structure and the multiplicity of
the deformation mode of slender links. From those solutions, finding of the different aspects
that comprise the workspace is possible, as well as determining singularities and instability loci.

1. Introduction

Parallel Continuum Mechanisms (PCMs) are flexible mechanical devices with a rigid end-effector that is connected to a fixed
frame using flexible slender links whose deformation is the cause of its mobility [1]. These are derived from continuum mechanisms,
where parallel assemblies of tendons, rigid parts, and flexible backbones are common ways of generating slender systems motivated
by the problem of manipulation in confined, hard to reach workspaces [2], and suitable for minimally invasive procedures due to
their dexterity and ease of miniaturization [3]. Aforementioned mechanisms belong to the group of Compliant mechanisms, those
that acquire mobility thanks to the relative flexibility of some of their parts [4,5].
Kinematic Analysis of PCMs aims to solve position problems, define singularities, obtain those aspects that build up workspaces,
and evaluate instability poses. The latter is a distinctive property of flexible mechanisms, i.e. as force equilibrium involves flexible
elements, such equilibrium poses should be stable for the mechanism to withstand a certain motion under given load [6].
Regarding position problems in PCMs of slender elements, an appropriate model for the nonlinear deformation of the rods used as
elements is needed. Pseudo-rigid-body model (PRBM) method [4,7] is the simplest approach but has important limitations because
it does not allow changes of deformation mode (commonly known in the literature as buckling mode), which are common along
any path in most systems. Another group of kinetostatic modelling methods are of the type of discretization-based approaches that
use structural compliance matrices to model each discretized element that build up an equivalent mechanism [8–10]. A consequent
natural step in the field of discretization approaches is the use of Finite Element models [11], which can be implemented for precise,
ad hoc analysis of nearly final designs, but are of more limited use in early stages of design. Regarding continuous models, Cosserat

∗ Corresponding author.
E-mail address: [email protected] (O. Altuzarra).

https://doi.org/10.1016/j.mechmachtheory.2023.105311
Received 11 October 2022; Received in revised form 20 February 2023; Accepted 20 February 2023
Available online 7 March 2023
0094-114X/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 1. Deformed shape of a clamped-free rod under load 𝑅 at end-tip.

rod model [12] is used extensively in modelling systems with rods subjected to large nonlinear deformations. This model can be
mathematically expressed in different ways, with a nonlinear system of differential equations (ODEs) the most accurate one. The
closed-loop system of flexible rods and rigid elements generates a coupling of large deformation modelling equations and force
equilibrium ones to accommodate the deformation’s internal forces and moments to the load. Based on this approach, a sustained
line of research on hexapod-like flexible devices has been performed, solving a variety of problems such as real-time position control
and elastic analysis [13–16] and dynamics [17,18]. The main drawback is that, to start the analysis, it requires a known pose, so
that close positions can be solved using a direct integration of the system of ODEs from previous values of state parameters. This
is plausible for path analysis, but it does not allow for the solving of the multiplicity of solutions, and hence a comprehensive
Kinematic Analysis.
Planar PCMs considered here have slender elements that are straight rods in their free-stress reference configuration; they are
usually subjected to planar Dynamics conditions, i.e. force loads are on the plane and applied moments are perpendicular to the
plane; and therefore, rods can be considered to be subjected to neither shear nor extension. Also, the material constitutive law
used is linear. Under such circumstances, the Cosserat nonlinear differential system of equations leads to a simplified model called
the Kirchhoff rod model. The key point of such model is that the resultant system of differential equations can be manipulated to
allow an analytical integration through elliptic integrals functions, a technique widely used in the field of continuum robots and
manipulators [19]. Moreover, each deformed shape of a rod is univocally defined by the deformation mode (d-mode), and two
parameters (𝑘 and 𝜓) whose space is enclosed in a certain range of real values [4]. Hence, when looking for the solution of the
position problem of these mechanisms the objective will be to get those sets of parameters for the flexible rods of the system such
that equilibrium of the whole exists. As the space of those parameters is limited, it is possible to search for all multiple solutions
even without an initial given pose. Authors presented a procedure to solve such problem in planar mechanisms of 2 degrees of
freedom (DoF) with punctual end-effector in [20] using numerical optimization. Later, we presented a certified solution method –
interval analysis – [21] to solve the same problem, demonstrating the feasibility to perform comprehensive Kinematic Analysis on
such closed-loop flexible mechanisms. Our recent research efforts have been focused on finding efficient ways to get such solutions
so that Kinematic analysis equivalent to those performed in rigid-link devices can be carried out [22,23]. Now, we are extending
these approaches to: more complex PCMs with other actuation systems and additional degrees of freedom, including singularity and
instability evaluation, and aiming to analogous Kinematic Analysis to those performed in classical Parallel Kinematics.
The paper is organized as follows: Section 2 is a short review of Kirchhoff rod model in planar systems solved using elliptic
integrals [4,24]; Section 3 contains the definition of the type of mechanisms analysed; Sections 4 and 5 present the procedures
to get all multiple solutions of the Inverse and Direct position problems; Section 6 recalls the use of iterative position analysis in
conjunction with singularity analysis and instability analysis to get all aspects of the workspace performing a classical Kinematic
Analysis; and Section 7 discuss the experimental application of these results. Finally, in Section 8, the conclusions of this paper are
drawn.

2. Fundamentals of Kirchhoff rod model

In this section, the reader will find a summary of the fundamentals on approaching the nonlinear analysis of the deformation
of elastic planar rods with the elliptic integrals method that are treated in detail in [4,12]. As planar kinematics is considered, no
out of plane action is considered, and then no out of plane bending or torsion applies. The following two mathematical structures
define the shape of a deformed planar rod: a parametric Cartesian curve, 𝐩(𝑠) ∈ R2 , identifying the centroid coordinates 𝑥 and 𝑦 of
each cross-section, and the angle 𝜃(𝑠) which establishes the orientation of a local frame attached to the cross-section, as shown in
Fig. 1. Position 𝐩(𝑠) and orientation 𝜃(𝑠) depend on the scalar reference arc-length 𝑠 defined over a finite interval, 𝑠 ∈ [0, 𝐿], where
𝐿 is the length of the rod.

2
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

The position and orientation of each part of the rod evolve along the arc length according to rates of change, linear 𝐯(𝑠) ∈ R2 and
angular 𝐮(𝑠) ∈ R. These rates of change are defined in the local frame, and related to the derivatives of position 𝐩(𝑠) and orientation
𝜃(𝑠) with respect to 𝑠 as:
⎧ 𝑑𝑥 ⎫
⎪ 𝑑𝑠 ⎪ ⎧𝑣 cos(𝜃) − 𝑣 sin(𝜃)⎫
⎪ 𝑑𝑦 ⎪ ⎪ 𝑥 𝑦 ⎪
⎨ 𝑑𝑠 ⎬ = ⎨𝑣𝑥 sin(𝜃) + 𝑣𝑦 cos(𝜃)⎬ (1)
⎪ 𝑑𝜃 ⎪ ⎪ 𝑢𝑧 ⎪
⎪ 𝑑𝑠 ⎪ ⎩ ⎭
⎩ ⎭
Linear 𝐯(𝑠) and angular 𝐮(𝑠) rates of change are also related to internal forces 𝐧(𝑠) [N] and moments 𝐦(𝑠) [Nm] (defined in a
fixed frame) through the material constitutive equations 𝐯 = 𝐊−1 𝑆𝐸
𝐑𝑇 𝐧 + 𝐯◦ and 𝐮 = 𝐊−1𝐵
𝐑𝑇 𝐦 + 𝐮◦ , where 𝐊𝑆𝐸 and 𝐊𝐵 are stiffness
matrices for shear–extension and bending respectively, where for initially straight rods 𝐯◦ = [1 0 0]𝑇 and 𝐮◦ = 𝟎, and where 𝐑 is the
rotation matrix that transforms magnitudes between a fixed frame and the local frame. For linear constitutive material law, stiffness
matrices are diagonal and, hence, we get
1 [ ]
𝑣𝑥 = 𝑛 cos(𝜃) + 𝑛𝑦 sin(𝜃) + 1 (2)
𝐸𝐴 𝑥
1 [ ]
𝑣𝑦 = −𝑛𝑥 sin(𝜃) + 𝑛𝑦 cos(𝜃) (3)
𝐺𝐴
𝑚𝑧
𝑢𝑧 = (4)
𝐸𝐼
where 𝐸 is the elastic modulus, 𝐴 the cross section area, 𝐺 the shear modulus, and 𝐼 the cross section’s area moment of inertia.
Note that some materials may require more complex material laws.
The simplified model of deformation, known as Kirchhoff rod model, is used. Rod’s shear and extension are neglected because
for slender rods the relative magnitude of internal forces to stiffness justifies the approximation of 𝑣𝑥 = 1 and 𝑣𝑦 = 0 and then the
deformed shape is considered to be caused only by bending, namely
⎧ 𝑑𝑥 ⎫
⎪ 𝑑𝑠 ⎪ ⎧cos 𝜃 ⎫
⎪ 𝑑𝑦 ⎪ ⎪ ⎪
⎨ 𝑑𝑠 ⎬ = ⎨ sin 𝜃 ⎬ (5)
⎪ 𝑑𝜃 ⎪ ⎪ 𝑚 𝑧 ⎪
⎪ 𝑑𝑠 ⎪ ⎩ 𝐸𝐼 ⎭
⎩ ⎭
To establish the equilibrium of forces and moments for every portion of the deformed rod, we consider an infinitesimal piece
of the rod between sections located at 𝑠 and 𝑠 + 𝑑𝑠. Equilibrium nonlinear differential equations describing the evolution of the
internal force 𝐧(𝑠) and moment 𝐦(𝑠) along rod’s arc length, subjected to distributed forces per arc length 𝐟(𝑠) [N/m] and moments
𝐥(𝑠) [Nm/m] applied on the rod, arises:
𝑑𝐧(𝑠)
+ 𝐟(𝑠) = 𝟎
𝑑𝑠 (6)
𝑑𝐦(𝑠) 𝑑𝐩(𝑠)
+ × 𝐧(𝑠) + 𝐥(𝑠) = 𝟎
𝑑𝑠 𝑑𝑠
If neither 𝐟 nor 𝐥 is applied, we get the following components of Eq. (6):
⎧ 𝑑𝑛𝑥 ⎫
⎪ 𝑑𝑠 ⎪ ⎧ 0 ⎫
⎪ 𝑑𝑛𝑦 ⎪ ⎪ ⎪
⎨ 𝑑𝑠 ⎬ = ⎨ 0 ⎬ (7)
⎪ 𝑑𝑚 ⎪ ⎪𝑛𝑥 sin 𝜃 − 𝑛𝑦 cos 𝜃 ⎪
⎪ 𝑑𝑠 ⎪ ⎩
𝑧 ⎭
⎩ ⎭
This results in a constant internal force 𝐧. If only end-point load is applied (𝑅 and 𝜓 being its magnitude and orientation,
respectively, as depicted in Fig. 1), for constant 𝐸 and 𝐼, assuming a stress-free reference straight rod, considering the last row in
Eq. (5), and upon substitution into Eq. (7), we get a single ODE in 𝜃:
𝑑2𝜃 1 𝑅
= (𝑅 cos 𝜓 sin 𝜃 − 𝑅 sin 𝜓 cos 𝜃) = sin (𝜃 − 𝜓) (8)
𝑑𝑠2 𝐸𝐼 𝐸𝐼
Integration of Eq. (8), by making use of some change of variables [24], yields:
√
√ 𝐸𝐼 [ ]
𝑅= 𝐹 (𝑘, 𝜙2 ) − 𝐹 (𝑘, 𝜙1 ) (9)
𝐿
where 𝐹 (𝑘, 𝜙) is the incomplete elliptic integral of the first kind, 𝑘 is a constant 𝑘 ∈ [−1, 1], whereas 𝜙 is a transformed variable
running continuously along the rod as 𝑠 and 𝜃 do. Angle 𝜙 varies between extremes 𝜙1 and 𝜙2 , which are defined by the boundary
conditions of the rod. For example, let us consider a rod clamped horizontally at the first extreme whereas the other end is free, as
represented in Fig. 1. Then, on the one hand, 𝜙1 is
( ( 𝜓 ))
1
𝜙1 = arcsin cos (10)
𝑘 2

3
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 2. Parametric definition of a planar flexible rod.

𝑑𝜃
while on the other end there is an inflection point of the curve, because 𝑚𝑧 and so 𝑑𝑠
is null, yielding 𝜙2

𝜙2 = (2𝑞 − 1)𝜋∕2 for 𝑞 = 1, 2, … (11)

where 𝑞 identifies the so-called d-mode (also known as buckling mode). Each d-mode 𝑞 has a 𝑞 number of inflection points, i.e., mode
1 has one inflection point (precisely at the free end), mode 2 has two, and so on.
Upon integration of Eq. (5), we can obtain now coordinates 𝑥 and 𝑦 of points on the rod along the arc length corresponding to
parameter 𝜙𝑖 , defining the deformed shape of the rod:
√
𝐸𝐼 [ ]
𝑥(𝜙𝑖 ) = − cos 𝜓 2𝐸(𝑘, 𝜙𝑖 ) − 2𝐸(𝑘, 𝜙1 ) − 𝐹 (𝑘, 𝜙𝑖 ) + 𝐹 (𝑘, 𝜙1 )
𝑅
√
𝐸𝐼 [ ]
− 2𝑘 sin 𝜓 cos 𝜙𝑖 − cos 𝜙1 (12)
𝑅
√
𝐸𝐼 [ ]
𝑦(𝜙𝑖 ) = − sin 𝜓 2𝐸(𝑘, 𝜙𝑖 ) − 2𝐸(𝑘, 𝜙1 ) − 𝐹 (𝑘, 𝜙𝑖 ) + 𝐹 (𝑘, 𝜙1 )
𝑅
√
𝐸𝐼 [ ]
+ 2𝑘 cos 𝜓 cos 𝜙𝑖 − cos 𝜙1 (13)
𝑅
where 𝐸(𝑘, 𝜙) is the incomplete elliptic integral of the second kind.
Hence, both Eqs. (5) and (7) have been integrated to yield Eqs. (9), (12), and (13), i.e., a parametric system that uniquely defines
a deformed rod in terms of 𝑘 and 𝜓 parameters, for known boundary conditions and a certain d-mode 𝑞. Note that feasible values
of 𝑘 depend on parameter 𝜓, so that 𝜙1 , in Eq. (10), is real, i.e. between |𝑘𝑚𝑖𝑛 | = cos 𝜓∕2 and |𝑘| = 1. This condition restricts the
parameter space (𝑘, 𝜓) for these boundary conditions as shown in Fig. 2(a). As it can be seen in Fig. 2, if a duple of values 𝑘 and
𝜓 inside the feasible range is selected, Fig. 2(a), we will univocally get the deformed shape of the rod for each d-mode desired, as
depicted in Fig. 2(b) for d-modes 1 and 2.
The aforementioned parametrization, allowed by this Kirchhoff planar deformation model integrable using elliptic integrals, is
of capital importance in the following procedures because solutions to the position problems of flexible mechanisms will be defined
by a limited number of parameters (𝑘𝑖 , 𝜓 𝑖 per rod 𝑖) whose possible values are in a closed space.

3. Mechanisms under study

The planar parallel continuum mechanisms considered in this work are 3 degrees of freedom (DoF) fully parallel mechanisms
formed by closed-loop assemblies of three flexible rods connected to a rigid-body moving platform. The end-effector’s moving frame
(, 𝑥𝑦𝑧) is located at reference point  which corresponds to the centroid of the triangular platform 𝐵1 𝐵2 𝐵3 . Flexible rods’ end-tips
are attached to the moving frame at points 𝐵𝑖 via revolute joints. Flexible rods are connected to the fixed element and actuated in
three different ways described below. A fixed reference system (, 𝑋𝑌 𝑍) is defined on the fixed element.
In Fig. 3, the three types of PCMs under study are represented. The three different types of attachment and actuation of the
flexible rods chosen are the most feasible ones. The following notation is used to define each mechanism: the underline identifies
the actuated variable, 𝑃 corresponding to a prismatic joint, 𝐹 denotes the flexible rod (only bending about an axis perpendicular
to the plane) and 𝑅 the revolute joint.

4
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 3. PCMs of 3 DoF.

Fig. 3(a) shows the 3𝑃 𝐹 𝑅 mechanism, where each rod is actuated by a prismatic slider, the flexible rod has a constant length 𝐿
connecting to the platform with a revolute joint. Flexible rod’s first-tip 𝐴𝑖 (subscript 𝑖 identifies the rod) is clamped to an actuated
linear table of fixed direction. A local frame (𝑖 , 𝑥𝑖 𝑦𝑖 ) is attached to the clamping, 𝑥𝑖 axis oriented with the clamped slope. Distance
to point 𝐴𝑖 on the actuated prismatic joint from a fixed reference point 𝑂𝑖 on the linear guide defines input length 𝜌𝑖 .
Fig. 3(b) represents the 3𝑅𝐹 𝑅 mechanism, each rod of length 𝐿 being actuated by a revolute joint. Flexible rod’s first-tip 𝐴𝑖 is
clamped to the actuated revolute joint. A local frame (𝑖 , 𝑥𝑖 𝑦𝑖 ) is attached to the clamping, 𝑥𝑖 axis being oriented with the clamped
slope, and such rotating axis defines the input angle 𝜙𝑖 with reference to fixed frame’s 𝑋 axis.
Finally, in Fig. 3(c) the 3𝑃 𝐹 𝑅 mechanism is shown, where the length of the flexible rod is not constant and indeed it corresponds
to the actuated variable 𝑠𝑖 . Flexible rods slide through a fixed prismatic pair at 𝐴𝑖 . A local frame (𝑖 , 𝑥𝑖 𝑦𝑖 ) is attached to the prismatic
pair, 𝑥𝑖 axis oriented with the slope of sliding. Length of the rod free to bend from 𝐴𝑖 to 𝐵𝑖 , 𝑠𝑖 , is the input variable of the mechanism.
With above definitions, all flexible rods have the same boundary conditions as in Fig. 1, with coordinates defined in the
corresponding local frames (𝑖 , 𝑥𝑖 𝑦𝑖 ) given by Eqs. (12) and (13). Based on the fundamentals described in Section 2, flexible rods 𝑖
are ruled by Eqs. (12), (13) and (9), and uniquely defined by parameters 𝑘𝑖 and 𝜓 𝑖 , once a d-mode is chosen for them (see Fig. 2).
The quasi-static position problem of the deformed mechanism is solved by finding the set of parameters 𝑘𝑖 and 𝜓 𝑖 for each rod
subject to the corresponding boundary conditions, hence six unknowns. Load applied at the end-effector is an input (3 parameters)
alongside either the three input values at actuators for the Forward Kinematics (FK) problem, or the three pose parameters of the
end-effector for the Inverse Kinematic (IK) problem.
The equivalent rigid parallel manipulators to these PCMs are shown in Fig. 4. For the flexible 3𝑃 𝐹 𝑅, the rigid 3𝑃 𝑅𝑅 would
perform comparable motions; the manipulator analogous to the 3𝑅𝐹 𝑅 would be the 3𝑅𝑅𝑅; and finally, the 3𝑃 𝐹 𝑅 shares similarities
with the 3𝑅𝑃 𝑅.
It must be stressed that flexible PCMs behave kinematically in a more or less similar way to these equivalent rigid manipulators
when rods’ d-mode is 1, diverging dramatically as higher d-modes appear (which is not rare for d-mode 2 but requires a very

5
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 4. Equivalent rigid manipulators: (a) 3𝑃 𝑅𝑅, (b) 3𝑅𝑅𝑅 and (c) 3𝑅𝑃 𝑅.

high deformation energy and involves an important risk of self collisions for higher ones). Also, it must be noted that whereas the
number of IK solutions, for the rigid manipulators, only depends on the geometric conditions that the legs must satisfy when the
end effector is located at a certain pose, in the case of the flexible manipulators additional conditions given by the fulfilment of
static equilibrium must be satisfied. Besides, the possibility of working in different buckling modes also affects the number of IK
solutions. As a consequence, the use of pseudo rigid body models to analyse the full solution of position problems has a very limited
success, and a more elaborated model is required.

4. Solving the full inverse kinematic problem

Solving the inverse kinematics problem (IK) implies obtaining the unknown input values of the actuators once the pose of the
end effector is given (3 in-data). In addition, in these flexible mechanisms the load must be considered as an input to the problem (3
additional in-data), and the complete definition of the mechanism configuration (the deformed shape) is the objective to be solved.
Obtaining the deformed shape of the flexible rods means getting the 6 parameters 𝑘𝑖 and 𝜓 𝑖 . Out of these, actuators’ values can
be obtained geometrically defining the output of the IK problem. Additionally, a full inverse solution means obtaining all multiple
solutions to a given end-effector’s pose and load, these arising from the non linearity of closed-loop equations as well as from the
multiple d-modes that a flexible rod may reach.
As mentioned, in flexible mechanisms, apart form the geometric conditions that have to be fulfilled, we have to stress that a force
equilibrium must be reached between the load and forces generated in the deformation process. This double condition is exploited
in the design of the proposed procedure.
The first step is to choose a certain d-mode for each flexible rod, this delimits the search procedure to those mechanism
configurations with the corresponding d-modes chosen, e.g. commonly the first search will be with d-mode 1 for all rods. Other
combinations of d-modes will be found by running the procedure with the new choice of d-modes bearing in mind that higher
d-modes require higher energy levels and then are less feasible.
Second, the procedure will use some distinct geometric boundary conditions that can be derived for each of the leg types described
in Section 3 to reduce the search in the parameter space. Such geometric conditions will be applied to determine a subset of possible
solutions in the closed parameter space of all solutions 𝑘𝑖 and 𝜓 𝑖 .

6
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 5. Geometric condition for leg type 𝑃 𝐹 𝑅.

The third step is to check the conditions of static equilibrium on the latter subset of solutions fulfilling the geometric conditions,
reducing them to a finite number of IK solutions.
Fourth, each of those solutions are then analysed to determine if they are regular solutions in a double sense. First, we check if
they are stable equilibrium configurations of the flexible mechanism using a variational approach. And second, Kinematic Jacobians
are obtained numerically to check closeness to singularity locus.
Finally, the resultant solutions have to be checked to verify if they are feasible with respect to other practical criteria. For
example, they are to be checked for collision between rods or/and with the end-effector, rods’ loop shapes, stress failure, and so
on. Some of these effects can be detected using geometric functions inside a last verification part of the algorithm. In the following,
we have left this last issue apart so that the full potential of the method can be appreciated showing all solutions found, even when
some of them could be undesirable or impractical in a mechanical device.

4.1. Geometric conditions

Once the end-effector is located at a certain desired pose (IK inputs), coordinates of the hinged joint 𝐵𝑖 of each leg are known,
as well as the fixed locations of sliding guides or revolute or prismatic input joints. These conditions were enough for getting a
feasible solution in the case of rigid links parallel mechanisms, but not in flexible ones.
For leg type 𝑃 𝐹 𝑅, see Fig. 5(a), as 𝐵𝑖 location is given, the flexible rod can be deformed into many different shapes corresponding
to different input values 𝜌𝑖 , and different forces and deformation energy values. They will correspond to an infinite subset of the
𝑘𝑖 − 𝜓 𝑖 space of all possible deformations of the rod. Conversely, in the rod’s local frame (𝑖 , 𝑥𝑖 𝑦𝑖 ) such geometric condition
corresponds to all possible deformed shapes of the rod with a fixed 𝐴𝑖 extreme and a given value of 𝐵𝑖 end-tip’s 𝑦 coordinate,
i.e. 𝑦𝑖 = ℎ, as shown in Fig. 5(b).
As the space of all deformed solutions of a rod 𝑘𝑖 − 𝜓 𝑖 can be represented by Fig. 2(a), those fulfilling the above-mentioned
condition will be a subset of 𝑘𝑖 − 𝜓 𝑖 for a chosen d-mode 𝑞, so that Eq. (13) particularized with Eq. (11) equals ℎ. Those duples in
the subset correspond to different end-tip forces at 𝐵𝑖 , and they should be checked for overall equilibrium later on. Note that IK
solutions provide actuators’ values 𝜌𝑖 , which are derived from the local 𝑥𝑖 coordinate of the end-tip and the given coordinates of
points 𝐵𝑖 .
Finding the subset 𝑘𝑖 −𝜓 𝑖 is developed further on here. As explained in Section 2, elliptic integrals functions provide an analytical
integration of Kirchhoff model equations, each deformed shape depending on parameters 𝑘 and 𝜓 defined in closed ranges [−1, 1] and
[0, 2𝜋] respectively. However, boundary conditions of the rod at extremes reduce this space further, for example for clamped-hinged
rods Eq. (10) applies and a minimum value of 𝑘 is required, depending on 𝜓, to get a real number. Hence, as shown in Fig. 6(a),
[ ] [ ]
the values of 𝑘𝑖 are in the restricted ranges −1, −𝑘𝑖𝑚𝑖𝑛 and 𝑘𝑖𝑚𝑖𝑛 , 1 .
To sample values of 𝑘𝑖 and 𝜓 𝑖 such that a variation in the conditions is evenly represented, a transformation of parameter 𝑘𝑖 is
proposed as 𝑘𝑖𝑟 , providing a rectangular range as shown in Fig. 6(b):
|𝑘𝑖 | − |𝑘𝑖𝑚𝑖𝑛 |
𝑘𝑖𝑟 = 𝑠𝑖𝑔𝑛(𝑘𝑖 ) ⋅ (14)
1 − |𝑘𝑖𝑚𝑖𝑛 |
Moreover, grid density for sampling can be adjusted with a higher density in areas where we observe that the results change rapidly.
Then, applying the geometric condition defined before, we get a subset of 𝑘𝑖𝑟 − 𝜓 𝑖 that can be represented with the curves shown
in Fig. 7.
In the case of a 𝑅𝐹 𝑅 leg type (see Fig. 8(a)), for given IK location of 𝐵𝑖 and being 𝐴𝑖 fixed, all possible deformed shapes for
rod 𝑖 are those with a known distance 𝐴𝑖 − 𝐵𝑖 between end-tips. Conversely, in the local frame, they will be those whose end-tips
are located on a circumference of radius 𝑟 (see Fig. 8(b)). The corresponding subset of 𝑘𝑖 − 𝜓 𝑖 can be found as before, imposing the
condition 𝑥2 + 𝑦2 = 𝑟2 with Eqs. (12) and (13) particularized with Eq. (11).

7
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 6. Comparison of Parameter Spaces alternatives.

Fig. 7. Curves in 𝑘𝑖𝑟 − 𝜓 𝑖 verifying the geometric condition.

8
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 8. Geometric condition for leg type 𝑅𝐹 𝑅.

Fig. 9. Geometric condition for leg type 𝑃 𝐹 𝑅.

Finally, for leg type 𝑃 𝐹 𝑅 (see Fig. 9(a)), the subset of solutions will be those deformed shapes of rod with variable flexible leg
lengths 𝑠𝑖 that place end-tips 𝐴𝑖 and 𝐵𝑖 on fixed given locations as shown in Fig. 9(b). The geometric condition used is somewhat
different than in previous cases, where rod’s length was constant and known 𝐿. We need to use a unit length rod and the local
frame to define the general geometric condition for the IK problem, whereas the actual rods’ lengths will the final solution. For that
purpose, we follow this rationale: those potential solutions under the IK geometric requirement shown in Fig. 9(b), when linearly
scaled to a rod’s unit length become into the rods’ shapes shown in Fig. 9(c). Hence, the geometric condition for unit length rod
can be stated as that end-tip of rod should be located on the line of known direction defined by 𝐴𝑖 and 𝐵𝑖 , i.e. a line at a constant
angle 𝛽, as shown in Fig. 9(c). The corresponding subset of 𝑘𝑖 − 𝜓 𝑖 can be found as above, imposing the condition 𝑦∕𝑥 = tan 𝛽 with
Eqs. (12) and (13) particularized with Eq. (11). Then, all those solutions are scaled to the actual locations of 𝐴𝑖 and 𝐵𝑖 .

9
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

4.2. Static equilibrium

The following step is to find which of the different combinations of possible deformed shapes of rods given by the three subsets of
𝑘𝑖 − 𝜓 𝑖 (as those in Fig. 7) generate a static equilibrium of forces and moments at the end effector, including not only the end-forces
𝑅𝑖 due to the deformation of the rods, but the load wrench, i.e., 𝐹𝑋 , 𝐹𝑌 , and 𝑀𝑍 . In the global fixed frame, using the corresponding
rotation matrices for each rods’ local frame 𝐑𝐹𝑖 that may depend on the inputs, and taking moments about the loading point on the
end effector, we get:
{ 𝑖 } { } { }
∑3
𝑅 cos 𝜓 𝑖 𝐹𝑋 0
𝐑𝐹𝑖 ⋅ 𝑖 𝑖 + =
𝑖=1
𝑅 sin 𝜓 𝐹𝑌 0
{ 𝑖 } (15)
∑3
𝑅 cos 𝜓 𝑖
𝑖 𝐹
𝐛 × 𝐑𝑖 ⋅ + 𝑀𝑍 𝐤 = 𝟎
𝑖=1
𝑅𝑖 sin 𝜓 𝑖

where 𝐛𝑖 is the location of the end-point attachments on the end effector 𝐵𝑖 with respect to the loading point 𝑃 .
In order to find all solutions, certified methods as [21] can be used to get 𝑘𝑖 − 𝜓 𝑖 final solutions. Also, global minimization
processes on Eq. (15) from previous subsets 𝑘𝑖 − 𝜓 𝑖 generate the multiple solutions of the IK problem with no formal guarantee
of certified complete solution. In the latter case, depending on the numerical sampling, approximate solutions are found, and a
refinement procedure is advised. In such case, it is convenient to apply direct numerical integration on Eqs. (5) and (7), and a Newton
scheme to geometric and static constraints to refine those approximate solutions. Finally, if a more realistic model is required for
the material properties or nonlinear deformations, the obtained solutions of the Kirchhoff model can be used as starting guess poses
to be optimized on the improved model.

4.3. Stability of equilibrium

A configuration on equilibrium of the flexible mechanism is reached when the potential energy of the mechanical system is on a
stationary state. This can correspond to a local minima of the energy function which means that an stable equilibrium configuration
is reached, or any other situations that mean an unstable equilibrium (usually saddle points of the energy function [23]). Unstable
equilibria are hardly reached in practice because the system moves out of it under any small perturbation, and therefore are solutions
of the position problem that must be discarded in the procedure. Such instabilities of the flexible mechanism can occur at any
point of the rods in the system [25]. Therefore, a variational approach should be considered for a comprehensive analysis of this
phenomenon. In the procedure proposed here, we have included the instability analysis of the aforementioned mechanisms following
the seminal work in [16], and adding to the line of calculus presented in [23]. If the analysis performed indicates that the static
equilibrium obtained is unstable, it is discarded as a plausible solution.

4.4. Kinematic Jacobians and singularities

Once the procedure has provided the multiple stable solutions for the IK problem, these need to be evaluated regarding kinematic
properties and singularities, which requires finding the Kinematic Jacobian matrices. In rigid-link parallel manipulators, as the
Jacobian is defined analytically, we can obtain the geometric conditions for singular poses, which eventually leads to the finding of
singularity locus. However, this is not the case in continuum parallel manipulators, because the Jacobians are numerically obtained
at each pose by applying an iterative position analysis with small displacements around the studied pose of either inputs 𝛥𝝆 or
outputs 𝛥𝐱 :
{ }
𝛥𝐩
𝛥𝐱 = = 𝐉𝐼𝐾 ⋅ 𝛥𝝆
𝛥𝜽 (16)
𝛥𝝆 = 𝐉𝐹 𝐾 ⋅ 𝛥𝐱
Based on such approach, Jacobians can be evaluated and velocity ellipsoids found [23]. Singularities are here determined by
proximity to singular conditions of Jacobians, which requires the evaluation of the position problem along a finite path or area so
that relative values of the norms can be appropriately assessed to establish the singularity threshold, or, better, detecting changes
of the sign of the norm of Jacobians on those paths. An in-depth singularity analysis is presented in [26], allowing a much more
detailed classification of the singularities found.

4.5. Case studies

The Full IK solution procedure described has been applied to those flexible mechanisms declared in Section 3 with symmetric
arrangements and simple regular dimensions so that analogies between the flexible manipulators under study and their rigid
counterparts are easily appreciated. In subsequent figures, the different IK solutions found for a given output pose for each flexible
manipulator will be shown. Bearing in mind that each leg can be working in a different d-mode, the set of solutions combining
d-modes 1 and 2 will be represented. No higher d-modes have been tried because the deformation energy needed is much more
higher, instabilities are therefore more prompt, and self-collisions of deformed rods are very common in such modes. No material
limit has been considered so far, but that is another important issue preventing the possibility of having higher d-modes than two.

10
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 10. PCM 3𝑃 𝐹 𝑅: five IK solutions for mode 111 (1st d-mode in each leg).

In relation to PCM 3𝑃 𝐹 𝑅, a set of possible IK stable solutions are represented in Figs. 10–12. In Fig. 10, the five IK solutions
associated with mode 111, that is, d-mode 1 for each leg, are shown. In Fig. 11, the two IK solutions associated with mode 222,
i.e. each leg in mode 2, are shown. Finally, in Fig. 12, one IK solution arises for the combination of modes 122, 212 and 221,
respectively. Here it can be observed that deformation shapes with self collision loops appear in those rods with mode 2. Such
occurrence can be also detected with geometric computation disregarding those IK solutions in an additional check.
Regarding the 3𝑅𝐹 𝑅, the set of possible IK solutions is represented in Fig. 13. In this case, eight solutions arise, all of them
for mode 111, meaning that each leg is working in d-mode 1. The equivalent rigid manipulator, the 3𝑅𝑅𝑅 (Fig. 4(b)), also has
eight solutions of the IK problem. This can be the most appropriate example to appraise the similarities between rigid and flexible
mechanisms.

11
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 11. PCM 3𝑃 𝐹 𝑅: two IK solutions for mode 222 (2nd buckling mode in each leg).

Fig. 12. PCM 3𝑃 𝐹 𝑅: IK solutions for mixed modes.

Finally, in relation to PCM 3𝑃 𝐹 𝑅, the set of possible IK solutions are represented in Figs. 14 and 15. In Fig. 14, the solutions for
modes 111 and 222 can be seen, having one solution for mode 111 and two for mode 222. In the case of combining the d-modes
in each leg, Fig. 15 shows the existence of one solution for each combination mode 122, 212 and 221.

5. Solving the full forward kinematic problem

The full FK problem has the objective of finding all possible poses of the end-effector for given values of the inputs at the motors
and the load to be sustained. For all above-mentioned types of mechanisms, the actuation input condition means that first extreme
of flexible rods 𝐴𝑖 is given as well as its slope, i.e. the local frame 𝑖 location is defined; and rods’ lengths are known.
Solving the FK problem consist in finding the six parameters 𝑘𝑖 − 𝜓 𝑖 that define the deformed shape of flexible rods 𝑖 of the
mechanism, and eventually the pose of the end-effector given by position of their end-tips 𝐵𝑖 . Two conditions must be fulfilled: that
the end-tips assemble at the end-effector, and that forces imposed by those end-tips due to the rods’ deformation on the end-effector
produce a force equilibrium with the specified load.
First condition means, for a rigid-body triangular platform, that the three end-points 𝐵𝑖 must fulfil rigid-body equations, i.e. that
distances 𝑑𝑖𝑗 between the attachment points 𝐵𝑖 and 𝐵𝑗 , and angles 𝛽𝑖𝑗𝑘 between sides 𝐵𝑖 𝐵𝑗 and 𝐵𝑗 𝐵𝑘 are constant. A sufficient
condition being for example a choice of: distance 𝑑12 between 𝐵1 and 𝐵2 , distance 𝑑13 between 𝐵1 and 𝐵3 , and angle 𝛽213 between
sides 𝐵1 𝐵2 and 𝐵1 𝐵3 :
|𝐩1 (𝐿1 ) − 𝐩2 (𝐿2 )| = 𝑑12
|𝐩1 (𝐿1 ) − 𝐩3 (𝐿3 )| = 𝑑13 (17)
( 1 ) ( 1 )
| 𝐩 (𝐿1 ) − 𝐩 (𝐿2 ) × 𝐩 (𝐿1 ) − 𝐩3 (𝐿3 ) | = 𝑑12 𝑑13 sin 𝛽213
2

where 𝐩𝑖 are coordinates of 𝐵𝑖 in the fixed frame, and 𝐿𝑖 are known rods’ lengths. Second condition is that a static equilibrium is
generated, as explained in Section 4.2, being expressed in Eq. (15).
The FK solving procedure starts with the definition of the sampling grid of parameters 𝑘𝑖 and 𝜓 𝑖 as in Fig. 6, and the choice of
the d-modes that will be searched. Then, a global optimization process is used with these 6 parameters (𝑘𝑖 and 𝜓 𝑖 ) defined in closed
ranges, and the 6 equations in Eqs. (15) and (17). As a result, we may get several 3-duples of parameters that provide the multiple
FK solutions with that choice of d-modes. Those solutions are then checked for stability and singularity, providing the definite FK
solutions. To complete the process, other possible combinations of d-modes at will are evaluated. As an example, the FK problem is
solved for the 3𝑅𝐹 𝑅 flexible manipulator, considering combination mode 111. In this case, four stable solutions arise, these being
represented in Fig. 16.

12
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 13. PCM 3𝑅𝐹 𝑅: eight IK solutions for mode 111.

Fig. 14. PCM 3𝑃 𝐹 𝑅 IK solutions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

13
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 15. PCM 3𝑃 𝐹 𝑅: IK solution for mixed modes.

Fig. 16. Four FK solutions of the 3𝑅𝐹 𝑅: mode 111.

6. Workspace analysis

Workspace Analysis in the design stage consists in finding all reachable poses of a mechanism considering all possible assembly
modes of the system. Each of the reachable singularity-free volumes are called aspects in classic rigid-link parallel kinematics. Those
aspects can then be evaluated pose by pose to get kinematic performance indices to asses manipulability or to do an optimal design,
being this still a controversial issue in Kinematics [27]. For Parallel Continuum manipulator the approach chosen is the analogous
one. First, upon a choice of output pose and load, find the multiple stable solutions of the IK problem for the set of plausible
buckling modes. Then, use each of them to start a numerical evaluation of the surrounding poses making use of the values of the
state parameters obtained in the closest predetermined pose. Checking that the new obtained poses are continuous to the previous
ones we can obtain each aspect, which requires evaluating Kinematic Jacobians to detect singularity boundaries.
In planar three degrees of freedom mechanisms, being the output pose variables 𝑋 , 𝑌 , 𝜃 , the common approach is the
evaluation of constant orientation workspaces for a set of desired orientations 𝜃 . For each of these constant orientation workspaces,
the iterative procedure starts from a given pose with the desired orientation at some coordinates 𝑋 , 𝑌 . Then, a wave propagation is
initiated emanating from that position by step increments of those coordinates [28]. A direct numerical integration of the mechanism
ODEs equations as in [14] is very efficient for this purpose, provided that the reference state parameters values are not far from the
solution, which can be enforced by choosing a sufficient small step size. When no solution is reached, a boundary of the workspace
corresponding to an IK singularity has been detected, being this verified by keeping a track of the evolution of the IK Jacobian norm.
Upon evaluation of FK Jacobian, FK singularities are also detected, but are most efficiently found upon evaluation of instability which
occurs simultaneously [23]. Wave propagation is stopped in the direction of singularity encounters.

14
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 17. One Aspect of the Workspace of the 3𝑃 𝐹 𝑅 with initial modes 111.

As an example, different constant orientation workspaces are shown in Fig. 17 for the unique solution obtained of the 3𝑃 𝐹 𝑅
with the mode combination 111 (as shown in Fig. 14(a)). IK singularities curves (depicted in black colour) can be found at the
boundary of the workspace, while FK singularities, in orange, appear as boundaries of the workspace with instability areas.
An important issue is that, once the iterative process starts from a given pose with a choice of d-mode combination, there are
continuous transitions between solutions that may involve a change of d-mode in some rod. The change of d-mode of rods is not a
singularity or any other type of problem, it is a natural transition in the deformation mode of the flexible mechanism.

7. Experimental verification

In this section some issues related to practical application of the method are discussed and verified via experiments. The method
assumes some simplifications that require some verification, from the use of the Kirchhoff rod model itself, to the final results of
the position analysis and stability of the multiple solutions. Two test benches have been devised for this purpose.

15
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 18. Kirchhoff rod model testing.

Fig. 19. Experimental Validation of Kirchhoff Model.

First, the Kirchhoff rod model is verified experimentally on the test arrangement shown in Fig. 18. A Nitinol elastic rod is tested,
length of 𝐿 = 0.908 [m] and a circular cross section of radius 𝑟 = 1.5 [mm]. The material is modelled with a linear constitutive law
whose Elastic modulus is 𝐸 = 68𝑒9 [Pa] and Shear modulus 𝐺 = 25.5𝑒9 [Pa]. The rod is pinned at its end-tip to a Kistler dynamometer
where end-tip force components on a fixed cartesian frame are measured, while the other end is clamped to a linear actuator so
that different bending configurations as in the geometric condition shown in Fig. 5(a) can be easily generated.
In order to justify the use of the Kirchhoff rod model, Eqs. (1)–(3) have to be considered. As the rod to be used has a Euler’s
critical load 𝑃𝑐𝑟𝑖 = 𝜋 2 𝐸𝐼∕(0.7𝐿)2 = 6.6 [N], expected forces will be on the tenths of [N], and with values of 𝐸𝐴 = 4.8𝑒5 and
𝐺𝐴 = 1.8𝑒5 in Eqs. (2) and (3), the difference with respect to the Kirchhoff model values of 𝑣𝑥 = 1 and 𝑣𝑦 = 0 is of 𝑒−5 which
justifies neglecting shear and extension effects.
Experimental results are obtained for a fixed 𝑦 coordinate at the end-tip of −0.2125 [m] and a variable 𝑥 coordinate going from
0.86 to 0.51 [m] as for Fig. 5(b). These experimental measurements are compared to those obtained with a nonlinear FEM analysis
using ANSYS and the Kirchhoff rod model used in this paper, and are shown in Fig. 19. The differences are small enough to justify
the simplification.
Once the elastic rod model has been verified, a second test bench is assembled to evaluate the effectiveness of the method to
solve multiple solutions of a 3𝑃 𝐹 𝑅 mechanism. As shown in Figs. 20, a 3𝑃 𝐹 𝑅 mechanism is placed on a horizontal table where the
end-effector has a reduced friction contact on the surface and it is unloaded. The dimensions chosen for the rods are the ones that
verified the use of Kirchhoff model. These rods are clamped to linear actuators that are integrated in the system but the motion
evaluated will be quasistatic. These linear guides are on the edges of a fixed equilateral triangle of circumcircle radius 𝑟 = 0.9 [m].
Elastic rods are connected by revolute joints to the end-effector, that is an equilateral triangle with an edge length 𝑎 = 0.09 [m].

16
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Fig. 20. Experimental Validation of 3𝑃 𝐹 𝑅 mechanism.

Table 1
Theoretical and experimental results for the FK problem.
Outputs FK solution 1 FK solution 2
Theory Measured Theory Measured
𝑥 0 −2 mm 0 8 mm
𝑦 0 −2 mm 0 −1 mm
𝜃 90◦ 99◦ −75.7◦ −84◦

First, a set of inputs is chosen and the multiple stable solutions of end-effector poses are measured to verify the accordance to
solutions of the FK problem in Section 5. Input values are all equal to 𝜌𝑖 = 0.025 [m], theoretical output poses and measured ones
are presented in Table 1, and plot in Fig. 20.
Second, the input solutions found in the IK problem in Section 4 for given output poses are compared to the ones needed in the
prototype to arrive to them. Several output poses with 𝑥 = 𝑦 = 0 and the following orientations 𝜃 = 60◦ , 90◦ , 120◦ are tested, with
results given in Table 2.
These experimental results on prototypes with large dimensions and no calibration show a good agreement between the model
and the actual mechanism behaviour. Nevertheless, if accuracy is important for the objective task a more complex model should be
used as explained in the next section.

17
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

Table 2
Theoretical and experimental results for the IK problem.
Inputs 𝜃 = 60◦ 𝜃 = 90◦ 𝜃 = 120◦
Theory Measured Theory Measured Theory Measured
𝜌1 [m] −0.007 −0.017 0.025 0.019 0.058 0.051
𝜌2 [m] −0.007 −0.013 0.025 0.018 0.058 0.051
𝜌3 [m] −0.007 −0.018 0.025 0.013 0.058 0.040

8. Conclusions

In this paper, we propose a procedure to solve the full position problem of planar parallel continuum mechanisms of three degrees
of freedom with the following conditions: flexible elements are straight rods in free-stress reference configurations, no distributed
forces or moments are considered to act along the length of the rod, a linear material constitutive law is used, and bending is the
only source of deformation. A Kirchhoff model is used for the deformation, and a procedure is devised to take advantage of the
special properties of the elliptic integrals functions that result in the integration of the nonlinear system of differential equations.
The number of solutions obtained in the position problems is multiple, as it is commonly the case in their rigid-link counterparts,
but the variability of the d-mode of the slender rods adds additional solutions. Deformation modes are important in contributing to
such multiplicity, and must be addressed through the use of a suitable model, in this case, the elliptic integral solution method.
The procedure relies heavily in numerical routines subjected to tolerance thresholds that can burden the computational cost
or lose some solution. To alleviate the first drawback, the procedure can be used with tolerance values that produce preliminary
solutions with a limited accuracy, that is further improved using a direct integration approach [14] in a second stage.
The mechanisms chosen have rods connected to the end-effector via revolute joints, but a rigid connection could be also used. In
such a case, the method should be adjusted to the new boundary condition with a change of 𝜙2 in Eq. (11), a change of the boundary
conditions at 𝐵𝑖 for the three mechanisms considered, and including the end-tip bending moments in the balance of moments in
Eq. (15)
Also, if the deformation model is more complex, because other modes of deformation as shear or extension are relevant, or
because a more complex material law is required, a Cosserat model should be used. Bear in mind that common superelastic materials
have material laws that diverge from a linear elastic behaviour characterized by Young’s Modulus E, in fact they may present an
hysteretic behaviour, and that will affect the accuracy of results. The main problem to include such complex models is that such a
model does not allow an easy integration resulting in elliptic integrals functions. We would like to emphasize that the focus of the
paper is in the kinematic understanding of the performance of flexible parallel mechanisms that leads to the preliminary stages of
design of such a device. The contribution is a method to solve the complete position analysis considering multiple solutions, stability
issues, and eventually assembly modes and working modes of the mechanism that provide the best singularity free workspace. In
order to be able to do that in a reasonable time, the price to pay is a simplification of the model so that elliptic integral approach
can ensure a certified solution. Once the kinematician has chosen the mechanism for the task application, the next stage in the
design will require a more detailed model to solve the position problem with the accuracy needed. That model will be a Cosserat
model with the most appropriate material law, considering other effects than bending, and all distributed loads as own weight. Such
a model will not be solvable with the elliptic integral approach and will require a direct integration of differential equations, but
the results obtained with the proposed approach, by applying the Kirchhoff model, can be used as first approximations, that is, as
initial guess, in a direct integration of those more complex models.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared
to influence the work reported in this paper.

Data availability

No data was used for the research described in the article.

Acknowledgements

The authors wish to acknowledge the financial support received from the Spanish Government through the Ministerio de Ciencia
e Innovación (Project PID2020-116176GB-I00) financed by MCIN/AEI/ 10.13039/501100011033 and the support for the research
group through Project IT1480-22 provided by the Departamento de Educación from the Regional Basque Government, Spain. Also,
we want to express our gratitude to Javier Corral for his help with the ANSYS nonlinear analysis, and Luigi Tagliavini for his work
in the prototype assembly.

18
O. Altuzarra et al. Mechanism and Machine Theory 185 (2023) 105311

References

[1] C.E. Bryson, D.C. Rucker, Toward parallel continuum manipulators, in: 2014 IEEE International Conference on Robotics and Automation, ICRA, 2014, pp.
778–785, http://dx.doi.org/10.1109/ICRA.2014.6906943.
[2] P.L. Anderson, A.W. Mahoney, R.J. Webster, Continuum reconfigurable parallel robots for surgery: Shape sensing and state estimation with uncertainty,
IEEE Robot. Autom. Lett. 2 (3) (2017) 1617–1624, http://dx.doi.org/10.1109/LRA.2017.2678606.
[3] J. Burgner-Kahrs, D.C. Rucker, H. Choset, Continuum robots for medical applications: A survey, IEEE Trans. Robot. 31 (6) (2015) 1261–1280, http:
//dx.doi.org/10.1109/TRO.2015.2489500.
[4] L.L. Howell, Compliant Mechanisms, John Wiley and Sons, 2002.
[5] H. McClintock, F.Z. Temel, N. Doshi, J.-s. Koh, R.J. Wood, The milliDelta: A high-bandwidth, high-precision, millimeter-scale Delta robot, Science Robotics
3 (14) (2018) http://dx.doi.org/10.1126/scirobotics.aar3018.
[6] S.V. Levyakov, V.V. Kuznetsov, Stability analysis of planar equilibrium configurations of elastic rods subjected to end loads, Acta Mech. 211 (1) (2010)
73–87, http://dx.doi.org/10.1007/s00707-009-0213-0.
[7] S.M. Lyon, L.L. Howell, A simplified pseudo-rigid-body model for fixed-fixed flexible segments, in: Volume 5: 27th Biennial Mechanisms and Robotics
Conference, 2002, http://dx.doi.org/10.1115/detc2002/mech-34203.
[8] G. Chen, Z. Zhang, H. Wang, A general approach to the large deflection problems of spatial flexible rods using principal axes decomposition of compliance
matrices, J. Mech. Robot. 10 (3) (2018) http://dx.doi.org/10.1115/1.4039223.
[9] G. Chen, Y. Kang, Z. Liang, Z. Zhang, H. Wang, Kinetostatics modeling and analysis of parallel continuum manipulators, Mech. Mach. Theory 163 (2021)
104380, http://dx.doi.org/10.1016/j.mechmachtheory.2021.104380.
[10] X. Duan, W. Yan, G. Chen, H. Wang, Analysis and validation of a planar parallel continuum manipulator with variable Cartesian stiffness, Mech. Mach.
Theory 177 (2022) 105030, http://dx.doi.org/10.1016/j.mechmachtheory.2022.105030.
[11] J.W. Wittwer, M.S. Baker, L.L. Howell, Simulation, measurement, and asymmetric buckling of thermal microactuators, Sensors Actuators A 128 (2) (2006)
395–401, http://dx.doi.org/10.1016/j.sna.2006.02.014.
[12] S.S. Antman, Nonlinear Problems of Elasticity, Springer, 2005.
[13] D.C. Rucker, R.J.W. III, Statics and dynamics of continuum robots with general tendon routing and external loading, IEEE Trans. Robot. 27 (6) (2011)
1033–1044, http://dx.doi.org/10.1109/TRO.2011.2160469.
[14] J. Till, C.E. Bryson, S. Chung, A. Orekhov, D.C. Rucker, Efficient computation of multiple coupled Cosserat rod models for real-time simulation and
control of parallel continuum manipulators, in: Robotics and Automation (ICRA), 2015 IEEE International Conference on, IEEE, 2015, pp. 5067–5074,
http://dx.doi.org/10.1109/ICRA.2015.7139904.
[15] A.L. Orekhov, V.A. Aloi, D.C. Rucker, Modeling parallel continuum robots with general intermediate constraints, in: 2017 IEEE International Conference
on Robotics and Automation, ICRA, 2017, pp. 6142–6149, http://dx.doi.org/10.1109/ICRA.2017.7989728.
[16] J. Till, D.C. Rucker, Elastic stability of Cosserat rods and parallel continuum robots, IEEE Trans. Robot. 33 (3) (2017) 718–733, http://dx.doi.org/10.
1109/TRO.2017.2664879.
[17] J. Till, D.C. Rucker, Elastic rod dynamics: Validation of a real-time implicit approach, in: 2017 IEEE/RSJ International Conference on Intelligent Robots
and Systems, IROS, 2017, pp. 3013–3019, http://dx.doi.org/10.1109/IROS.2017.8206139.
[18] A.A. Alqumsan, S. Khoo, M. Norton, Robust control of continuum robots using Cosserat rod theory, Mech. Mach. Theory 131 (2019) 48–61, http:
//dx.doi.org/10.1016/j.mechmachtheory.2018.09.011.
[19] G. Gao, H. Wang, J. Liu, Y. Zheng, Statics analysis of an extensible continuum manipulator with large deflection, Mech. Mach. Theory 141 (2019) 245–266,
http://dx.doi.org/10.1016/j.mechmachtheory.2019.07.015.
[20] O. Altuzarra, D. Caballero, F.J. Campa, C. Pinto, Position analysis in planar parallel continuum mechanisms, Mech. Mach. Theory 132 (2019) 13–29,
http://dx.doi.org/10.1016/j.mechmachtheory.2018.10.014.
[21] O. Altuzarra, J.P. Merlet, Certified kinematics solution of 2-DOF planar parallel continuum mechanisms, in: T. Uhl (Ed.), Advances in Mechanism and
Machine Science, Springer International Publishing, 2019, pp. 197–208, http://dx.doi.org/10.1007/978-3-030-20131-9_20.
[22] O. Altuzarra, D. Caballero, Q. Zhang, F.J. Campa, Kinematic characteristics of parallel continuum mechanisms, in: J. Lenarcic, V. Parenti-Castelli (Eds.),
Advances in Robot Kinematics 2018, Springer International Publishing, 2019, pp. 293–301, http://dx.doi.org/10.1007/978-3-319-93188-3_34.
[23] O. Altuzarra, F.J. Campa, On singularity and instability in a planar parallel continuum mechanism, in: J. Lenarcic, B. Siciliano (Eds.), Advances in Robot
Kinematics 2020, Springer International Publishing, 2021, pp. 327–334, http://dx.doi.org/10.1007/978-3-030-50975-0_40.
[24] A. Zhang, G. Chen, A comprehensive elliptic integral solution to the large deflection problems of thin beams in compliant mechanisms, J. Mech. Robot.
5 (2) (2013) 021006, http://dx.doi.org/10.1115/1.4023558.
[25] T. Bretl, Z. McCarthy, Quasi-static manipulation of a Kirchhoff elastic rod based on a geometric analysis of equilibrium configurations, Int. J. Robot. Res.
33 (1) (2014) 48–68, http://dx.doi.org/10.1177/0278364912473169.
[26] S. Briot, A. Goldsztejn, Singularity conditions for continuum parallel robots, IEEE Trans. Robot. 38 (1) (2022) 507–525, http://dx.doi.org/10.1109/TRO.
2021.3076830.
[27] J.-P. Merlet, Parallel robots: Open problems, in: J.M. Hollerbach, D.E. Koditschek (Eds.), Robotics Research, Springer London, London, 2000, pp. 27–32,
http://dx.doi.org/10.1007/978-1-4471-0765-1_4.
[28] E. Macho, O. Altuzarra, C. Pinto, A. Hernandez, Workspaces associated to assembly modes of the 5R planar parallel manipulator, Robotica 26 (3) (2008)
395–403, http://dx.doi.org/10.1017/S0263574707004109.

19