Open access

Type Design of Decoupled Parallel Manipulators with Lower Mobility

Written By

Weimin Li

Published: 01 April 2008

DOI: 10.5772/5384

From the Edited Volume

Parallel Manipulators, New Developments

Edited by Jee-Hwan Ryu

Chapter metrics overview

3,303 Chapter Downloads

View Full Metrics

1. Introduction

A typical parallel mechanism consists of a moving platform, a fixed base, and several kinematical chains (also called the legs or limbs) which connect the moving platform to its base. Only some kinematical pairs are actuated, whose number usually equals to the number of degrees of freedom (dofs) that the platform possesses with respect to the base. Frequently, the number of legs equals to that of dofs. This makes it possible to actuate only one pair per leg, allowing all motors to be mounted close to the base. Such mechanisms show desirable characteristics, such as large payload and weight ratio, large stiffness, low inertia, and high dynamic performance. However, compared with serial manipulators, the disadvantages include lower dexterity, smaller workspace, singularity, and more noticeable, coupled geometry, by which it is very difficult to determine the initial value of actuators while the end effector stands at its original position.

In an engineering point of view, it is always important to develop a simple and efficient original position calibration method to determine initial values of all actuators. This calibration method usually becomes one of the key techniques that a type of mechanism can be simply and successfully used to the precision applications. Accordingly, few have been reported that the parallel manipulators being applied to high precision situations except micro-movement ones.

The study of movement decoupling for parallel manipulators shows an opportunity to simply the original position calibration and to improve the precision of parallel manipulators in a handy way. One of the most important things in the study of movement decoupling of parallel manipulators is how to design a new type with decoupled geometry.

Decoupled parallel manipulators with less mobility (LM-DPMs) are parallel mechanisms with less than six dofs and with decoupled geometry. This type of manipulators has attracted more and more attention of academic researchers in recent years. Till now, it is difficult to design a decoupled parallel manipulator which has translational and rotational movement simultaneously (Zhang et al., 2006a, 2006b, 2006c). Nevertheless, under some rules, it is relatively easy to design a decoupled parallel manipulator which can produce pure translational (Baron & Bernier, 2001; Carricato, & Parenti-Castelli, 2001a; Gao et al., 2005; Hervé, & Sparacino, 1992; Kim & Tsai, 2003; Kong & Gosselin, 2002; Li et al., 2005a, 2005b, 2006a; Tsai, 1996; Tsai et al., 1996; Zhao & Huang, 2000) or rotational (Carricato & Parenti-Castelli, 2001b, 2004; Gogu, 2005; Li et al., 2006b, 2007a, 2007b) movements.

This chapter attempts to provide a unified frame for the type design of decoupled parallel manipulators with pure translational or rotational movements.

The chapter starts with the introduction of the LM-DPMs, and then, introduce a general idea for type design. Finally, divide the specific subjects into two independent aspects, pure translational and rotational. Each of them is discussed separately. Special attention is paid to the kinds of joins or pairs, the limb topology, the type design, and etc.


2. The general idea for decoupled parallel manipulators with less mobility

The general idea for the type design of decoupled parallel manipulators with less mobility can be expressed as the following theory.

Theory: A movement is independent with others if one of the following conditions is satisfied:
  1. To the pure translational mechanisms, the translational actuator is orthogonal with the plane composed of other translational actuators.

  2. To the pure rotational mechanisms (spherical mechanisms), the translational actuator is parallel with the axis of rotational actuator.

Depend on part (1) of the theory, we can design some kinds of 3-dofs pure translational decoupled parallel manipulators. Also we can get some kinds of 2-dofs spherical mechanism based on part (2) of the theory.

For the convenience, first, let us define some letters to denote the joints (or pairs). They are the revolute joint (R), the spherical joint (S), the prismatic pair (P), and the planar pair or flat pair (F). They possess one revolute dof, three revolute dofs, one translational dof and three dofs (two translational and one revolute) respectively. Then the theory can be expressed by figure 1 and figure 2 separately.

Figure 1 illustrates the limb topology. The actuator should be installed with the prismatic pair. The flat pair can be composed in deferent way. Using this kind of limb, we can design some kinds of 3-dofs pure translational decoupled parallel manipulators.

Figure 1.

The idea for limb which can be used to compose decoupled translational mechanisms.

Figure 2(a) illustrates the general one geometry of a decoupled 2-dofs spherical mechanism. The moving platform is anchored to the base by two legs. A leg consists of two revolute joints, R1 and R2, whose axes, z1 and z2, intersect at point o and connect to each other perpendicularly to form a universal joint; so the value of is /2. The other leg consists of a revolute joint, R3, a flat pair, F, and a prismatic pair P, in which the moving direction of P is perpendicular to the working plane of F and the axis of R3. The revolute joints R2 and R3 are mounted on the moving platform in parallel. The prismatic pair P and the revolute joint R1 are assembled to the base, in which the moving direction of P is parallel to the axis of R1.

Suppose that the input parameters, q1 and q2, represent the positions of the revolute joint R1 and the prismatic pair P, which are driven by a rotary actuator and a linear actuator separately. The pose of the moving platform is defined by the Euler angles 1 and 2 of the platform. When the value of q1 changes and q2 holds the line, only 1 alters. On the other hand, when the value of q2 changes, only 2 changes. So, 1 and 2 are independently determined by q1 and q2 respectively, i.e., one output parameter only relates to one input parameter. In other words, the platform rotations around two axes are decoupled.

Figure 2(b) is an improved idea of figure 2(a). Using this idea, we can get a decoupled 2-dofs spherical mechanism with a hemi-sphere work space.

Figure 2.

The idea for decoupled 2-dof spherical mechanisms.


3. Design of 3-dofs translational manipulators with decoupled geometry

3.1. Type design

The Type design of 3-dofs translational manipulators is based on the analysis of limb topology shown in figure 1.

Figure 3.

The substitutes for the flat pair and the prismatic pair.

Firstly, we construct deferent structures to replace the flat pair and the prismatic pair. Some substitutes for the flat pair and the prismatic pair are shown in figure 3. Then, using the pairs to form variational kinds of limbs. Figure 4 shows three examples. Finally, we can constitute the 3-dofs translational manipulators by installing the specified limbs in orthogonal as shown in figure 5, 6 and 7.

Figure 4.

The examples of limbs.

Figure 5.

3-PPP manipulator.

Figure 6.

3-7R manipulator.

Figure 7.

Modified 3-7R manipulator.

3.2. Kinematics

The forward and inverse kinematic analyses for the 3-PPP manipulator shown in figure 5 are trivial since there exists a one-to-one correspondence between the moving platform position and the input pair displacements. So the velocity jacobia matrix is a 33 identity matrix.

The kinematics of 3-7R manipulator can be analysed as follows. Referring to figure 6(b), each limb constrains point P to lie on a plane which passes through points Mj2, Mj3, and Bj, and is perpendicular to the axis of x, y, and z, respectively. The position of jth plane is determined only by j whenever the length lj1 is given. Consequently, the position of P is determined by the intersection of three planes, i.e., the intersection of j for j=1,2,3. If the distance from Mj1 to Mj2 is m0j, then a simple kinematic relation can be written as


where p=[px py pz]T denotes the position vector of the end-effector. Taking the time derivative of equation (1) yields


where J is a diagonal matrix that holds


The kinematics of the modified 3-7R manipulator are the same.

3.3. Original position calibration

The calibration of 3-PPP manipulator is the same as a pure translational 3-dofs serial manipulator. So we just consider the manipulator of 3-7R and modified 3-7R, they can be expressed in the same way as shown in figure 8(a).

Figure 8.

Original position calibration.

For convenience, we suppose,

  1. The input j (j=1,2,3) is within [-jm, jm], where jm >0, and j = jm denotes the initial position of the jth limb;

  2. In the initial position (see figure 8), the angle between the link lj1 (j=1,2,3) and the axis u(u=x,y,z) is 0j (j=1,2,3).

Then the initial value jm of j can be determined as


So we can determine 0j first, thenjm, the steps of the calibration can be as follows. From the initial position 0j of the arm in figure 8, rotate the driving arm twice in a specified angle s, which satisfies


During the process, record the two moving distances l0 and l of the platform in the direction of axis u(u=x,y,z), they satisfy


expand cos(θ0j+θs) and cos(θ0j+2θs) in equation (6),and eliminatesinθ0j,we get

If θ0π/3 and letθ=π/3, then equation (7) yields

The geometric signification of the equation (8) is shown in figure 8(b), which is very sententious and convenient to industrial applications. θjmcan be get from equation (4).

3.4. Singularity

The 3-PPP manipulator has no singularity, so we just discuss the manipulator of 3-7R and modified 3-7R, they can be expressed in the same.

From equation (2) we can find out that the rotational actuator speed is nonlinear to the velocity of the end-effector. Moreover, ifθj=±90°, thendet|J|=0, for any expected velocity of the end-effector, the rotational speed of the actuator will be infinite. When θj is not equal but close to±90°, thendet|J|0, the required rotational speed of the actuator may be still too high to reach. So the value of the θj must be designed in an appropriate range whenever the speed limit of the end-effector is given.

Suppose the desired velocity of the end-effector isve, and the permissible rotational speed of the actuator isne, then the absolute maximum value of the θj for j=1,2,3 can be obtained from equation (2), that is


Then θj should satisfy


Whenever the mechanism design satisfies equation (11), no singularity will exist.


4. Design of 2-dofs spherical manipulators with decoupled geometry

4.1. Type Design

The Type design of 2-dofs spherical manipulators is based on the general idea shown in figure 2. Using the 3R and 4R pairs in figure 3 to replace the F and P pairs separately, a new structure (2R&8R manipulator) for figure 2(a) is constructed as shown in figure 9. Similarly, figure 10 shows the improved configuration of figure 2(b), a 2R&PRR manipulator, but distinguishingly, additional modification is that a through hole is added to the center of the revolute joint R1, so the prismatic pair P can be set in the center of the hole and rotates with R1. As a result, the workspace of 1 can reach 2.

Figure 9.

2R&8R manipulator.

Figure 10.

2R&PRR manipulator.

4.2. Kinematics

Firstly, the 2R&8R manipulator in figure 9 will be discussed. Let e be the distance between the axes of R2 and R3, m be the distance between the axes of R8 and R10 (or R7 and R9). Also, suppose that, when the moving platform is on the initial position, the axis of R1 is perpendicular to the plane consisting of the axes of R2 and R3. Then the displacement relationships between input and output for the 2R&8R manipulator are:


In the structure design, it is easy to set the length m of R7R9¯ and R8R10¯ equal to the distance e between the axes of R2 and R3 so as to get the one-to-one input-output mapping. Let m = e, it follows that:


This implies that the direct linear one-to-one input-output correlation, so the velocity jacobia matrix becomes an identity one.

Now we discuss the the 2R&PRR manipulator shown in figure 10. Suppose that the input parameters, q1 and q2, represent the angular displacement of the revolute joint R1 and the distance between the axes of R2 and R4 separately. They are driven by a rotary actuator and a linear actuator. The pose of the moving platform is defined by the Euler angles 1 and 2 of the platform. Let e be the distance between the axes of R2 and R3, m be the distance between the axes of R3 and R4. Also suppose that, axis z3 is through the point o and always perpendicular to the plane of z1-z2 and moreover, define the value of 2 is zero whenever the axis of R3 is on the plane of z1-z2. Then the coordinates of R4 and R3 for the axes z1 and z3 are


The displacement relationship between input and output is:


Taking the derivative of equation (15), it follows that




4.3. Singularity and workspace

The 2R&8R manipulator shown in figure 9 has two legs. The first leg (R1 to R2) produces the Euler angle 1 of the platform by the input of q1; while the second one (R10 to R3) produces 2 by q2. To illustrate the motional relationship, let us introduce a transition parameter z to equation (12), it follows that:


where, z is the displacement of F-pair (R4 to R6) in the direction of z1.

From equation (18), it is seen that the Euler angle 1 is produced from the input of q1 directly by the first leg; while 2 is produced from q2 by the second leg through two transformations, which include (1) rotary to linear motion q2z usingmsinq2=z, and (2) linear to rotary motion zθ2 usingz=esinθ2. In the second transformation, there exists a limitation related to friction circle. Let denote the radius of the friction circle of R2, which is determined by the product of the radius r of the revolute joint’s axis and the equivalent friction coefficient as follows.


Figure 11.

Force and torque of R2.

Let denote the angle between z1 and the linkR2R3¯, and decompose the force F into two parts, the radial component Fr and the tangent component Ft (see figure 11). Then the force F acts on R2 is equivalent to a force Q and a torque M, which can be calculated from the following equations.


As a basic law in mechanics, the effect of a force Q and a torque M acting on a rigid body is equivalent to a force Qh with an offset h, which is shown in figure 12 and can be calculated as follows


where, h is the distance between the action lines of force Qh and Q.

Figure 12.

Force couple equivalent.

There exist three instances for the different relationship between h and , which are (1) h < , the revolute joint R2 will never rotate regardless the value of Qh; (2) h > , revolute joint R2 can rotate; and (3) h = , the critical condition. In the critical condition of h = , using equation (21), it follows that:


Then the workspace of 2 satisfies:


On the other hand, the angle 1 produced by the first leg is limited only by the structure design of the F-pair and the base, so the workspace of 1 can reach a designated area through proper design. Assume that the workspace of 1 is from – /2 to /2, then the workspace of the spherical mechanism can be depicted by the reachable range of the point P as shown in figure 13. The workspace is smaller than a hemisphere, so it would be limitted in some applications.

When the mechanism is running, the direction of axis z1 keeps unchanged, while the direction of axis z2 alters according to 1. So the workspace represented by spherical surface in figure 13 can be interpreted as follows: point P draws latitude lines when only 1 changes and draws longitude lines while only 2 alters.

Figure 13.

The workspace denoted by the locus of point P.

Now we examine the 2R&PRR manipulator in figure 10. The only limitation of this mechanism is caused by the friction circle of R2. This limitation can be described by figure 14, from which we can see that the work space of θ2 satisfies


Where θ2min and θ2max are the minimum and the maximum boundaries, which can be simply calculated based on figure 14 as follows


Figure 14.

Workspace of θ2 limited by friction circle of R2.

It means that the workspace of the mechanism can not reach a hemisphere. Clearly, this is not desirable.

In fact, because the workspace of θ1 is [0, 2], the mechanism workspace can reach a hemisphere only if the workspace of θ2 is chosen [0, /2] or [/2, ]. So there exist two methods to get a hemisphere workspace.

Figure 15 shows the critical instances for both of them; each one uses the similar technique to offset the axis of R4 from the axis z1. Let n denotes the axis offset of R4 (or the length of AR4), and nc is the special value of n for the critical configurations as shown in figure 15, then n should be chosen equation (27). Using this technique, a hemisphere work space can be obtained.


Figure 15.

Two methods to modify the boundaries ofθ2: (a)θ2min=0, (b)θ2max=2π

Figure 16.

The improved mechanism for θ2[0,π/2].

The improved architectures are shown in figure 16 and figure 17, in which the workspace of θ2 includes the area of [0, /2] or [/2, ] separately.

A prototype model of the mechanism for the condition of θ2[0,π/2] is designed. Figure 18 shows the outline picture of this model. In this design, one leg is actuated by a servo motor through a tooth belt; while the other leg is actuated by the other servo motor through a ball screw, which converts the rotational movement into the translational one. Both motors are fixed on the base. Besides, another revolute joint R5 is added to connect the prismatic pair with the nut of the ball screw.

Figure 17.

The improved mechanism for θ2[π/2,π].

Figure 18.

The prototype model for θ2[0,π/2].


5. Conclusions

A general idea for type design of decoupled parallel manipulators with less mobility is introduced. A unified frame for the type design is provided and divided into two independent aspects. Some kinds of decoupled parallel manipulators with 3-dofs pure translational and 2-dofs pure rotational movements are obtained.



The authors gratefully acknowledge the financial support of the National Natural Science Foundation of China (No. 50475055).


  1. 1. BaronL.BernierG. 2001 The design of parallel manipulators of star topology under isotropic constraint, 2001 ASME Design Engineering Technical Conferences, Paper DAC-21025 , Pittsburgh, PA.
  2. 2. CarricatoM.Parenti-CastelliV. 2001a A family of 3-DOF translational parallel manipulators, in: Proceedings of the 2001 ASME Design Engineering Technical Conferences, Pittsburgh, PA, DAC-21035.
  3. 3. CarricatoM.Parenti-CastelliV. 2001b A two-decoupled-dof spherical parallel mechanism for replication of human joints, Servicerob 2001- European Workshop on Service and Humanoid Robots, 512 , Santorini, Greece.
  4. 4. CarricatoM.Parenti-CastelliV. 2004 A Novel Fully Decoupled Two-Degrees-of-Freedom Parallel Wrist, The International Journal of Robotics Research, 23(6), 661-667.
  5. 5. GaoF.ZhangY.LiW. M. 2005 Type synthesis of 3 -dof reducible translational mechanisms, Robotica, 23(2), 239-245.
  6. 6. GoguG. 2005 Fully-Isotropic Over-constrained parallel wrists with two degrees of freedom, in Proceedings of 2005 IEEE International Conference on Robotics and Automation (ICRA 2005), Barcelona, Spain, 18-22, 0-78038-915-840254030 .
  7. 7. HervéJ. M.SparacinoF. 1992 STAR: a new concept in robotics, International Conference of 3K-ARK, 176183 .
  8. 8. KimH. S.TsaiL. W. 2003 Design optimization of a cartesian parallel manipulator, Journal of Mechanical Design, 125(1), 43-51.
  9. 9. KongX. W.GosselinC. 2002 Kinematics and singularity analysis of a novel type of 3-CRR 3-dof translational parallel manipulator, International Journal of Robotic Research, 21(9), 791-798.
  10. 10. LiW. M.GaoF.ZhangJ. J. 2005a R-CUBE, a Decoupled Parallel Manipulator Only with Revolute Joints, Mechanism and Machine Theory, 40(4), 467-473.
  11. 11. LiW. M.GaoF.ZhangJ. J. 2005b A three-DOF translational manipulator with decoupled geometry, Robotica, 23 6 805808 .
  12. 12. LiW. M.ZhangJ. J.GaoF. 2006a P-CUBE, A decoupled parallel robot only with prismatic pairs, The 2nd IEEE/ASME International Conference on Mechatronic and Embedded Systems and Applications, Beijing, China.
  13. 13. LiW. M.SunJ. G.ZhangJ. J.HeK.DuR. 2006b A novel parallel 2 -dof spherical mechanism with one-to-one input-output mapping, WSEAS Transactions on Systems 5(6).
  14. 14. LiW. M.HeK.QuY. X.ZhangJ. J.DuR. 2007a On the type design of a fully decoupled 2-DOF spherical mechanism with a hemisphere workspace, WSEAS Transactions on Systems 6(10).
  15. 15. LiW. M.HeK.QuY. X.ZhangJ. J.DuR. 2007b HEMISPHERE, a fully decoupled parallel2 -DOFspherical mechanism. Proceedings of the 7th WSEAS int. Conference on Robotics, Control and Manufacturing Technology, 301-306, Hangzhou, China.
  16. 16. TsaiL. W.WalshG. C.StamperR. 1996 Kinematics of a novel three DOF translational platform, Proceedings of the 1996 IEEE International Conference on Robotics and Automation, 34463451 , Minneapolis, MM.
  17. 17. TsaiL. W. 1996 kinematics of a three-DOF platform manipulator with three extensible limbs, Recent Advances in Robot Kinematics, J. Lenarcic, and V. Parenti-Castelli, ed., 401410 , Kluwer Academic Publishers, London.
  18. 18. ZhangJ. J.LiW. M.WangX. H.GaoF. 2006a Study on kinematics decoupling for parallel manipulator with perpendicular structures. Proceedings of the 2006 IEEE/RSJ International Conference on Intelligent Robots and Systems, 748753 , Beijing, China.
  19. 19. ZhangJ. J.WangX. H.LiW. M.GaoF. 2006b On the study of type design for PMWPSs and their kinematics characteristics. WSEAS Transactions on Systems 5 6 13281334 .
  20. 20. ZhangJ. J.WangX. H.LiW. M.GaoF. 2006c Mechanism Design for 3-, 4-, 5- and 6-DOF PMWPSs. Proceedings of the 6th WSEAS International Conference on Robotics, Control and Manufacturing Technology, 93-98, Hangzhou, China.
  21. 21. ZhaoT. S.HuangZ. 2000 A novel three-DOF translational platform mechanism and its kinematics, 2000 ASME Design Engineering Technical Conferences, Paper MECH-14101 , Baltimore, MD.

Written By

Weimin Li

Published: 01 April 2008