Sei sulla pagina 1di 14

OPTIMAL DESIGN OF THREE TRANSLATIONAL 3 DOF PARALLEL MECHANISMS FOR MACHINING PROCESSES WITH GENETIC ALGORITHMS S. Mitsi1, K.-D.

Bouzakis1, L. Misopolinos1
1. Laboratory for Machine Tools and Manufacturing Engineering, Mechanical Engineering Department, Aristoteles University of Thessaloniki, Greece

ABSTRACT The performance of a parallel kinematic manipulator can be improved by optimal evaluation of different parameters. Especially when the space in which the robot will be placed is limited then an optimal design must be carried out. The paper presents the optimal design of three parallel kinematic manipulators: a Delta mechanism with prismatic actuators, an orthoglide mechanism and a 3 DOF translational parallel manipulator (TPM). The optimization process is performed with the aid of a genetic algorithm considering the same geometric limits, such as the actuator length limits and design constraints. The genetic algorithm is implemented for maximizing the workspace, manipulability and stiffness. The obtained results can be used to identify an optimal region of the workspace of each mechanism regarding the manipulability and stiffness and furthermore by comparing these performances to select the appropriate mechanism for machining process. KEYWORDS: Delta parallel mechanism, Orthoglide, Genetic algorithm, Optimization 1. INTRODUCTION

Many industrial applications which involve machining processes require tasks to be performed by machine tools with specified dimensions. It is also a demand of the industry for machines tools with higher flexibility in machining capabilities such as the machining centers. In this direction machine tools with parallel kinematics have been developed. The advantages of parallel manipulators over serial such as high stiffness, high accuracy and low inertia make this type of robots candidates for machining processes /1/. Small workspace is a well known problem for parallel mechanisms. Without a desired workspace it is not possible the mechanism to perform any of the prescribed tasks. For this reason the workspace is the most important index to design a manipulator. Another priority is achieving high manipulability into the workspace. Also the optimization of stiffness is of high importance because the mechanisms are developed for machining processes. The optimal design of the parallel manipulator must be performed into specific actuator length limits. Since most machining operations only require a maximum of five axes, the configurations with 3 axes (degrees of freedom) are more appropriate. Different parallel kinematic mechanisms with three degrees of freedom (DOF) can be used for high-precision machine tools. From these mechanisms well known are the Delta mechanism, the orthoglide and the 3-DOF translational parallel manipulator (TPM). However it is observed /2-20/ that a comparison of different types of 3 DOF parallel mechanisms optimized for 3 criteria, workspace volume, manipulability and stiffness index, for same actuators lengths limits and design constraints has not received much attention. In the present paper a multi-criterion design method is used to optimize three parallel kinematic manipulators: a Delta parallel mechanism with prismatic actuators, an orthoglide mechanism and a 3 DOF translational parallel manipulator (TPM). The objective of the optimization is to find
Proceedings of the 3rd International Conference on Manufacturing Engineering (ICMEN), 1-3 October 2008, Chalkidiki, Greece Edited by Prof. K.-D. Bouzakis, Director of the Laboratory for Machine Tools and Manufacturing Engineering (), Aristoteles University of Thessaloniki and of the Fraunhofer Project Center Coatings in Manufacturing (PCCM), a joint initiative by Fraunhofer-Gesellschaft and Centre for Research and Technology Hellas, Published by: and PCCM

511

the dimensions of the mechanisms maximizing the workspace, manipulability and stiffness considering the same actuator length limits and design variable limits. Following previous experience /10, 16/ the multi objective optimization problem is solved with the aid of a genetic algorithm. The proposed algorithm uses a variation of the simple genetic algorithm /21, 22/ with multipoint crossover, constraints handling method and hill climbing method. The advantage of GAs over other search methods is the ability to search the whole space of solutions and at the same time not to be trapped in a local minimum. Furthermore a constraints handling method (CHM) is applied in order to limit the design variables /23, 24/. Also the application of the hill climbing method /22/, which is a local search method, reduces the computational time. In the rest of the paper, the mathematical formulation of the problem is presented considering workspace, manipulability and stiffness index number and is stated in Section 2. The proposed optimization method is analyzed in Section 3 and the results for all three mechanisms are given in Section 4. In Section 5 the conclusions of this paper are presented. 2. PROBLEM FORMULATION

2.1. PARALLEL MANIPULATORS DESCRIPTION In the present paper three parallel manipulators with 3 DOF are considered: a Delta robot (Figure 1), an orthoglide (Figure 2) and a 3 DOF translational parallel manipulator (Figure 3). Each mechanism consists from the base, the moving platform and three identical fixed-length limbs, connected to the base by prismatic actuated joints. Each of the three limbs contains one parallelogram. The three parallelograms are connected to the moving platform and the sliders by revolute joints respectively. The axes of the prismatic joints of the Delta, orthoglide and 3 DOF TPM mechanisms are parallel, orthogonal and inclined with angle with respect to x-y plane, respectively. The angle i is the angle between x-axis and OAi vector (i=1, 2, 3) for the Delta and 3 DOF TPM mechanisms and so 1 = 0, 2 = 120 and 3 = 240 (Figure 1 and Figure 3 right). The moving platform of the mechanisms has three translational DOF with respect to the base, so that it retains a constant orientation during the motion. For kinematic analysis of each mechanism two reference systems are used (see Figure 1, 2 and 3): a global fixed reference system O-xyz located at the base and a mobile reference system PxPyPzP located at the mobile platform parallel with the global one. The geometric parameters of each mechanism are the base length Ri = OAi, the moving platform length ri = PCi and the limbs length ci = BiCi, where i=1, 2, 3. For the 3 DOF TPM the angle

Figure 1: Delta robot with linear actuators.

512

3rd ICMEN 2008

Figure 2: Orthoglide mechanism.

Figure 3: 3-DOF translational parallel manipulator (TPM).

is also a geometric parameter. AiBi is the scalar variable of the actuated prismatic joint with length di (i=1, 2, 3). The objective of the inverse kinematics is to find the displacement of the prismatic joints with the given position of the reference point P. The closed-form solutions for the inverse kinematics of the Delta, orthoglide and 3 DOF TPM mechanisms developed in /6/, /11/ and /17/ respectively are used during the optimization procedure. 2.2. OBJECTIVE FUNCTION The aim of the present paper is to determine the design variables of the mechanism in order to maximize the workspace volume, the conditioning index and the stiffness of the mechanism when the limits of the prismatic actuated joint displacements are imposed by the mechanism design:

Robots in Manufacturing Environments

513

d i,lower < d i < d i,upper , i = 1, 2, 3

(1)

where di, lower and di, upper are the low and upper limits of the prismatic joint displacement i . The design variables of each considered mechanism are the base length Ri, the moving platform length ri and the limbs length ci (i=1, 2, 3). Moreover, for the 3DOF TPM the angle is also a design variable. These variables must be inside the limits imposed through mechanism design. Thus an optimization problem can be formulated with the following objective function:

F = a F1 + b F2 + c F3
where F1, F2 and F3 are given by the following relations:

(2)

k V + k2 , for the Delta mechanism F1 = 1 k1 logV , for the orthoglide and 3 DOF TPM

(3) (4) (5)

F2 = n1 GCI
F3 = m1 m2 log(GSI )

with a, b and c are the weighting factors and k1, k2, n1, m1 and m2 are the factors for the normalization of the workspace V, global conditioning index GCI and global stiffness index GSI respectively. The weighting factors a, b and c are used in order to scale the contribution of each term, taking the values between 0 and 1. Very critical is the selection of the normalization factors, so as the parts F1, F2 and F3 to have the same order of magnitude. The selection is made through the trial and error method with the use of the first random generation of each mechanism. The normalization factors are selected so as the values of terms F1, F2 and F3 to be between 0 and 100. It must be pointed out that the value of the fitness function is not a global optimum for the three mechanisms but for each one separately. From the maximization of the objective function occur the values of the design variables. The methods for determination of the workspace, conditioning index and stiffness index used during the optimization procedure are presented following. 2.3. WORKSPACE ANALYSIS To determine the workspace of a mechanism a variation of the box method /1/ is used. An initial box sufficiently large is chosen to contain completely the workspace. Then a smaller sub box is chosen and scans in directions x, y, z the whole initial box. For each given position of the center of the sub box, solving the inverse kinematics the displacement of each actuated prismatic joint is calculated. If the solution is inside the limits of prismatic actuated joints displacement then the given point and the corresponding sub box belong to the workspace volume. The volume of the workspace V is obtained as a sum of all the allowable sub boxes given by:

V = SBV
1

(6)

where n is the number of all the allowable sub boxes and SBV is the volume of each sub box. The box size affects the accuracy of the formed workspace and as well the computational time. A small box leads to better workspace accuracy but the computational time increases. The choice of the box size is made by use of a trial and error method, in order to achieve good workspace accuracy with fast computational time.

514

3rd ICMEN 2008

2.4. CONDITIONING INDEX The manipulability of a parallel mechanism is estimated through the local conditioning index (LCI), which is the reciprocal of the condition number of the Jacobian matrix /25, 26/:

LCI=

1 J -1 J

(7)

where denotes the second norm of the Jacobian matrix J. The LCI is used to evaluate the dexterity, isotropy, as well as the static stiffness of a manipulator [27]. LCI takes values from 0 to 1. When LCI=0 the mechanism is in a singular configuration and when LCI=1 in an isotropic one. That means that when the LCI is near 0 the mechanism will be out of control. The purpose of the optimization is the LCI to reach value 1, if this is possible. The value of the LCI depends on the position of the mechanism within its workspace. For that reason, during the optimization procedure the RMS value of the LCI is used. This value is called global conditioning index (GCI) and is given by:

GCI = i =1

LCIi2
n
(8)

2.5. STIFFNESS INDEX The stiffness performance of the mechanism can be evaluated with the aid of different indices /7, 15, 17, 26/ using a stiffness model of the designed mechanism. In the present paper the stiffness model of a parallel mechanism takes into account only the stiffness of the actuators /7/. The maximum deformation on the end-effector is defined as the local stiffness index (LSI) and is calculated with the method presented in /7/. This methodology can be easily incorporated into the optimization algorithm. The purpose of the optimization is that the LSI value to be as small as possible. The value of the LSI varies with the variation of the mechanism configuration within its workspace. During the optimization procedure the global stiffness index (GSI) is used to evaluate the stiffness of the parallel mechanism. This index is given by:

GSI =

i =1

LSIi2
n
(9)

3. OPTIMIZATION METHOD The optimal design of each considered mechanism is solved with the aid of a genetic algorithm. The GAs have the advantage of searching the whole space of the solutions as well as not being entrapped in a local minimum. Also they can solve any kind of objective functions and any kind of constraints. GAs use payoff (objective function) information, not derivatives and so they can solve linear and non linear problems. An initial randomized population that consists of a group of chromosomes and represents the design variables, produces new population through successive iterations, using various genetic operators. The common operators are selection, crossover, mutation and elitism. A function called fitness function determines when a new chromosome will replace a previous one or not, according to its worth. Through several repetitions the evolution of the individuals leads to the domination of the stronger ones.

Robots in Manufacturing Environments

515

In the present paper the GA uses binary coding for the design variables, multi point crossover, adaptive mutation rates, a hill climbing method (HCM) and a constraints handling method (CHM). The HCM is a local search method used to accelerate the optimization procedure. It starts with the elite vector of variables of each generation. Then it changes the value of a variable, one at a time, and recalculates the fitness function. If a better value results then this new vector of variables replaces the elite of the current generation. The CHM defines bounds of the design variables, the relations between the design variables and the limits regarding each term of the fitness function in order to maximize the workspace, the manipulability and stiffness of the mechanism simultaneously. The flow chart of the proposed algorithm is illustrated in Figure 4. The input data for the algorithm are the variables bounds, the actuated joints limits, the design and other constraints and the genetic algorithm parameters. In these parameters are included the population size, crossover rate, initial mutation rate and number of generations. The limits for design variables and actuators length as well as the relations between the design variables are imposed by design requirements. The fitness function defined by the equation (1) is used in all steps of the algorithm. In the proposed algorithm the loop begins with the 1st step where a random population is firstly generated to set the variable values of the generation. Next the genetic algorithm uses CHM to eliminate the variables that are outside the constraints limits. Also the constraints regarding the value of the terms F1, F2 and F3 of the fitness function are checked. The purpose of these constraints is to decrease the contribution of the individuals that give values of terms outside predefined limits from the next generation.

Figure 4: Proposed algorithm flow chart.

516

3rd ICMEN 2008

After the CHM, the optimum individual is selected and is processed with HCM a local search method to achieve better results. The hill climbing method searches a small space around the optimum individual. This is done by changing the variables, one at a time, by small steps. If a better optimum value is found then this one is the elite individual of the generation. In the next ith step the crossover and mutation take place to form the new generation. The elite is passed in all the generations without being modified from crossover and mutation. In this way it is certain that the best of each generation will not be destroyed and its a way to guide the optimization process. Then the procedure continues until the maximum generations number is achieved. Finally the results are examined and if they are satisfactory the optimization is finished. In other case the optimization procedure begins again with different parameters like crossover rate, number of generation or different random first generation. 4. APPLICATION RESULTS

The proposed algorithm is applied for the optimum design of a parallel Delta mechanism, an orthoglide and a 3 DOF TPM to maximize the workspace volume, the conditioning index and the stiffness. In all numerical examples the input data as the design variables bounds, the actuated joints displacement limits and the parameters of the genetic algorithm are same (Table 1). The GA parameters are selected as optimums through many applied tests. The design variables are the base length Ri, the moving platform length ri, the leg length ci (i=1, 2, 3) of the Delta, orthoglide and the 3-DOF TPM mechanism. As shown in Table 1 the base length Ri must always be 100 mm longer with respect to the moving platform length ri in order to assure the stability of the mechanism. For the 3-DOF TPM the angle of the base OAi with the x-y plane of the mechanism is also a design variable. This angle is limited between 0 and 80 degrees. The values between 80 and 90 degrees are excluded because the axes of the prismatic actuator joints tend to become parallel and the mechanism to be transformed in a parallel Delta mechanism which is separately studied. Table 1: Input data. Design constraints (mm) 200 Ri 600 70 ri 470 200 ci 600 0 80 (degrees) (Ri - ri) > 100 Actuated joints limits (mm) 0 di 300 i = 1, 2, 3 Initial genetic algorithm parameters Crossover rate = 0.8 Probability of mutation = 0.05 Population size = 60 atoms Number of generations = 130

For the 3-DOF TPM an additional constraint is set, all the points of the workspace must always be in the negative side of the z-axis in order the moving platform to work below the base as shown in Figure 3 and to avoid the collision with the base. Furthermore the lower threshold for the term F2 of the objective function F is set to 10 for all three mechanisms. The term F3 has an upper threshold of 1300 in order to keep the GSI value into realistic values. If the terms F2 and F3 are outside these thresholds then the fitness function is multiplied by a number which is 0.5, to decrease the contribution of the atom to the next generation. There is also a constraint that checks the difference between terms F1 and F2 in order to assure the maximization of the workspace and the manipulability of the mechanism simultaneously. So if this difference is above 50 units then again the fitness function is multiplied by the coefficient 0.5 to decrease the contribution of the atom to the next generation. It is reminded here once again that the values of the terms F1, F2 and F3 are between 0 and 100.

Robots in Manufacturing Environments

517

During the optimization procedure it is possible to obtain design variables that lead to a zero value of the fitness function, which is the minimum value of the generations fitness function. To avoid this problem the zero value of the fitness function is replaced with a small value. With this technique the solution has almost no chance to reproduce to the next generation and also the generations fitness function average is not affected. In all cases the weighting factors used in equation (1) are a=b=c=1. The corresponding normalization factors k1, k2, n1, m1 and m2 are given in Table 2. Table 2: Normalization factors. Mechanism Delta Orthoglide 3 DOF TPM k1 1.1*10 13.0 13.0
-6

Normalization factors k2 n1 m1 10.0 0.0 0.0 100.0 100.0 100.0 80.0 80.0 80.0

m2 18.0 25.0 25.0

After running the algorithm for 130 generations, the procedure is finished and the optimum variables design values are obtained. During the optimization procedure the workspace is calculated several times in one generation. In order to reduce the computational time without deteriorating the workspace accuracy, after several tests, the value 64 mm3 of the box size is chosen as optimum value for the workspace volume evaluation. The optimum design variables of the three mechanisms are inserted in Table 3. Furthermore for the 3 DOF TPM the optimized value of the angle is 79.25 degrees. Table 3: Optimum design variables and results. Design variables Mechanism Base length Moving platform Link length (mm) length (mm) (mm) Delta Orthoglide 3 DOF TPM 453.000 599.875 565.625 92.000 70.125 336.625 586.500 399.000 596.875 GCI GSI Workspace (mm3)

0.4852 1.4410 13,654,720.0 0.4764 9.1341 28,603,008.0 0.1715 35.0679 22,815,552.0

To compare the performance of the designed mechanisms the workspace, the conditioning and stiffness indices are evaluated. The workspace shape of the Delta, orthoglide and 3 DOF TPM mechanisms is presented in Figure 5, Figure 6 and Figure 7 respectively. In comparison with the workspace shape of the Delta and 3 DOF TPM mechanisms, the orthoglide workspace is almost a cube. The LCI distribution in the z-axis level corresponding to the maximum x-y workspace section of the Delta, orthoglide and 3 DOF TPM mechanisms is illustrated in Figure 8, Figure 9 and Figure 10 respectively. For other z-levels the distribution shape of the LCI remains the same. The highest LCI value of the Delta mechanism is near the workspace center (see Figure 8), while for the orthoglide mechanism (Figure 9) the highest value is near the workspace edge and is better than the corresponding value of the Delta mechanism. It is noted that inside workspace of the Delta and orthoglide mechanisms the singularities are avoided. In contrast, the lowest LCI value of the 3 DOF TPM (Figure 10) takes zero value and singularities are created inside its workspace. A comparison of these results reveals that the Delta and orthoglide mechanisms have better manipulability with respect to 3 DOF TPM. Figure 11, Figure 12 and Figure 13 show the LSI distribution in a z-axis workspace level of the Delta, orthoglide and 3 DOF TPM mechanisms respectively. It can be observed that the LSI

518

3rd ICMEN 2008

Figure 5: Workspace of Delta parallel mechanism.

Figure 6: Workspace of orthoglide mechanism.

Figure 7: Workspace of the 3DOF TPM.

Robots in Manufacturing Environments

519

Figure 8: LCI distribution for z-axis level 709.5 mm of the Delta mechanism.

distribution in a z = -709.5 mm level of the Delta mechanism (Figure 11) is 120 degrees symmetrical. This symmetry coincides with the three axes of the Delta mechanism. In addition the LSI values are acceptable and the difference between the highest and lowest value of the LSI is small. For the orthoglide mechanism (Figure 12), the LSI in the z = -74 mm level takes low values inside the workspace except only one small workspace part, where the LSI takes highest value. It is observed that for this mechanism the difference between the highest and lowest value of the LSI is greater than the corresponding one of the Delta mechanism. Finally, for the 3 DOF TPM (Figure13), in the z = -92 mm workspace level, the LSI takes low values in the upper right part of the workspace level and in the remaining part the increase of the LSI values becomes very steep. A comparison of the results regarding the distribution and the values of the LSI in the level workspace show that the Delta and orthoglide mechanisms have good stiffness properties with respect to 3 DOF TPM.

Figure 9: LCI distribution for z-axis level -74 mm of the orthoglide mechanism.

520

3rd ICMEN 2008

Figure 10:

LCI distribution z-axis level -92 mm of the 3 DOF TPM.

In order to have a global image of the performance regarding the workspace, manipulability and stiffness, the value of the workspace, GCI and GSI are taken into account (Table 3 right). Also are considered the maximum and minimum values of the LCI and LSI (Table 4) into the whole workspace of the three mechanisms. A comparison of the results (Table 3 right) obtained for the three designed mechanisms reveals that, although the Delta mechanism has the best values of the GCI and GSI, its workspace volume takes the lowest value. On the other hand, the maximum workspace volume is obtained with orthoglide mechanism which has still GCI value neighbor to the corresponding value of the Delta mechanism. It is noted that the criteria of the manipulability and stiffness are antagonistic for the 3 DOF TPM. As shown in Table 4, for the Delta mechanism the maximum and minimum values of the both LSI and LCI are acceptable and the difference between these values is small. In contrary, the other two mechanisms present a high difference between maximum and minimum values in both LCI and LSI. Specifically for the orthoglide the maximum LCI reaches the value of 1 which is the overall distribution in a z = -709.5 mm level of the Delta mechanism (Figure 11) is 120 degrees symmetrical. This symmetry coincides with the three axes of the Delta mechanism. In addition the optimum of the CI. But at the same time there are regions inside its workspace that must be avoided because the LCI approaches the value of zero. Analog remarks can be done for the maximum and minimum LSI values of the orthoglide. Final for the 3 DOF TPM the mini-

Figure 11: LSI distribution for z-axis level 709.5 mm of the Delta mechanism.

Robots in Manufacturing Environments

521

Figure 12: LSI distribution for z-axis level -74 mm of the orthoglide mechanism.

Figure 13: LSI distribution for z-axis level -92 mm of the 3 DOF TPM. Table 4: Maximum and minimum values of LCI and LSI. LCI Mechanism Delta Orthoglide 3 DOF TPM Max 0.5516 1.0000 0.3301 Min 0.2982 0.0325 0.0000 Max 1.5867 32.4306 65.7383 LSI Min 1.2003 0.0005 2.0694

mum value of the LCI and the maximum value of the LSI show that there are regions inside its workspace that are entirely unusable.

522

3rd ICMEN 2008

The aforementioned remarks can be used for the selection of the appropriate mechanism for the machining process. For example, in terms of the workspace performance the orthoglide mechanism is the best (see Table 3 right). If, at the same time, a minimum LCI value is prescribed, then the designer can select the mechanism that has the maximum exploited workspace region. For example, if the minimum prescribed LCI value is 0.2982, which is the minimum LCI value of the Delta mechanism, then the exploited workspace region of the orthoglide and 3 DOF TPM mechanisms become 24,331,456.0 mm3 and 372,672.0 mm3 respectively. Therefore, the appropriate candidate for machining process considering the workspace and manipulability performance is the orthoglide mechanism. Furthermore, if the manipulability and stiffness are simultaneously prescribed, a similar procedure can be used for the selection of the appropriate mechanism. If the minimum prescribed LCI value is 0.2982 and the maximum prescribed LSI value is 1.5867, then the exploited workspace region volume of orthoglide and 3 DOF TPM mechanisms become 1,077,760.0 mm3 and zero respectively. It is concluded, that in this case the Delta mechanism is the best candidate for machining process having the biggest volume. 5. CONCLUSIONS

In the present paper the optimum design of a parallel Delta mechanism, an orthoglide and a 3 DOF TPM is developed considering the same actuator displacement limits and design constraints. The fitness function takes into account the workspace volume, the manipulability measure and global stiffness index of the mechanism. The developed optimization procedure, based on genetic algorithms, guarantees the optimal results, although some of the considered criteria are antagonistic. The obtained results can be used to identify an optimal region of the workspace of each mechanism regarding the manipulability and stiffness and furthermore by comparing these performances to select the appropriate candidate for machining process. Moreover, the proposed methodology can be extended for the optimal design of 3 DOF parallel mechanisms considering also the stiffness properties of the legs. The developed algorithms for the optimization process are written in Visual Studio .NET C++ 2005. The solid graphics are designed in SolidWorks environment. The workspace LCI and LSI graphs are made with MatLab and the 3-D workspace in RapidForm. 6. REFERENCES

1. Merlet JP. Parallel robots. Kluwer Academic Publishers; 2000. 2. Clavel R. Delta: a fast robot with parallel geometry. 18th Int. Symp. on Industrial Robot (ISIR), Springer-Verlag, Berlin 1988;91100. 3. Pierrot F, Reynaud C, Fournier A. Delta: a simple and efficient parallel robot. Robotica 1990;8:105109. 4. Codourey A. Dynamic modelling and mass matrix evaluation of the Delta robot for axes decoupling control. Proceedings of IEEE IROS 1996;12111218. 5. Demaurex MO. The Delta robot within the industry. In: Boer CR, Molinari-Tosatti L, Smith KS, editors. Parallel Kinematic Machines Springer; 1999, p. 395399. 6. Liu XJ, Jinsong W, Kun-Ku O, Jongwon K. A new approach to the design of a Delta robot with a desired workspace. Journal of Intelligent and Robotic Systems 2004;39:209225. 7. Liu XJ. Optimal kinematic design of a three translational DoFs parallel manipulator. Robotica 2006;24:239-250. 8. Liu XJ, Wang J, Zheng H. Workspace atlases for the computer-aided design of the Delta robot. Proceedings of the I. MECH. E. Part C: Journal of Mechanical Engineering Science 2003;217:861869.

Robots in Manufacturing Environments

523

9. Laribi MA, Romdhane L, Zeghloul S. Analysis and dimensional synthesis of the Delta robot for a prescribed workspace. Mechanism and Machine Theory 2007;42:859-870. 10. Mitsi S, Bouzakis KD, Misopolinos L, Milutinovic D. Workspace and manipulability optimization with genetic algorithms of a 3 DOF spatial parallel mechanism used in machining processes. 2nd International Conference on Machining Engineering ICMEN, Sani, Chalkidiki, Greece 2005;351-361. 11. Wenger Ph and Chablat D. Kinematic analysis of a new parallel machine-tool: the orthoglide. In: ARK Piran 2000;305-314. 12. Wenger Ph, Chablat D. Design of a three-axis isotropic parallel manipulator for machining applications: the orthoglide. Proceedings of the Workshop on Fundamental Issues and Future Research Directions for Parallel Mechanisms and Manipulators, Quebec City 2002;1623. 13. Wenger Ph, Chablat D. Architecture optimization of a 3-DoF parallel mechanism for machining applications: the orthoglide. IEEE Transactions on Robotics and Automation 2003; 19:403410. 14. Pashkevich A, Chablat D, Wenger Ph. Kinematics and workspace analysis of a three-axis parallel manipulator: the orthoglide, Robotica 2006;24:3949. 15. Majou F, Gosselin C, Wenger P, Chablat D. Parametric stiffness analysis of the orthoglide. Mechanism and Machine Theory 2007;42:296311. 16. Mitsi S, Bouzakis KD, Misopolinos L. Optimal design of a modified orthoglide parallel kinematic mechanism used in a CNC milling machine. International Conference on Manufacturing Systems ICMS, Iasi, Romania 2007;377-384. 17. Xu Q, Li Y. An investigation on mobility and stiffness of a 3-DOF translational parallel manipulator via screw theory. Robot Comput Integr Manuf 2007, doi:10.1016/j.rcim.2007.02.022. 18. Tremblay A, Baron L. Geometrical synthesis of parallel manipulators of Star-like topology with a genetic algorithm. Proceedings of the 1999 IEEE International Conference on Robotics & Automation, Detroit 1999;3:2446-2451. 19. Zhang D, Xu Z, Mechefsk CM, Fengfeng X. Optimum design of parallel kinematic toolheads with genetic algorithms. Robotica 2004;22:7784. 20. Stamper RE, Tsai LW, Walsh GC. Optimization of a three DOF translational platform for well-conditioned workspace. Proceedings of IEEE International Conference on Robotics and Automation, New Mexico 1997;32503255. 21. Holland JH. Adaptation in natural and artificial systems. Ann Arbor, Michigan: University of Michigan Press; 1975. 22. Goldberg D. Genetic algorithms in search, optimization, and machine learning. AddisonWesley; 1989. 23. Carlson S. A general method for handling constraints in genetic algorithms. In: Proceedings of the Second Annual Joint Conference on Information Science 1995;663-667. 24. Michalewicz Z. Genetic algorithms, numerical optimization, and constraints. In: Proceedings of the Sixth International Conference on Genetic Algorithms, Pittsburgh, 1995;151-158. 25. Gosselin C, Angeles J. A global performance index for the kinematic optimization of robotic manipulators. Trans. ASME, J. Mech. Design 1991;113:220-226. 26. Xin-Jun L, Zhen-Lin J, Feng G. Optimum design of 3 DOF spherical parallel manipulators with respect to the conditioning and stiffness indices. Mechanism and Machine Theory 2000; 35:1257-1267. 27. Angeles J, Lopez-Cajun C. Kinematic isotropy and the conditioning index of serial robotic manipulators. International Journal of Robotics Research 1992;11:560-571.

524

3rd ICMEN 2008

Potrebbero piacerti anche