Path Planning for Formation Control of Autonomous

Similar documents
Investigations of Topology and Shape of Multi-material Optimum Design of Structures

Priority queues and heaps Professors Clark F. Olson and Carol Zander

A Binarization Algorithm specialized on Document Images and Photos

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields

Machine Learning. K-means Algorithm

Overview. Basic Setup [9] Motivation and Tasks. Modularization 2008/2/20 IMPROVED COVERAGE CONTROL USING ONLY LOCAL INFORMATION

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

Meta-heuristics for Multidimensional Knapsack Problems

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Kent State University CS 4/ Design and Analysis of Algorithms. Dept. of Math & Computer Science LECT-16. Dynamic Programming

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION

An Optimal Algorithm for Prufer Codes *

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

PARETO BAYESIAN OPTIMIZATION ALGORITHM FOR THE MULTIOBJECTIVE 0/1 KNAPSACK PROBLEM

Cluster Analysis of Electrical Behavior

Wavefront Reconstructor

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation

Cooperative UAV Trajectory Planning with Multiple Dynamic Targets

Smoothing Spline ANOVA for variable screening

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

3D Virtual Eyeglass Frames Modeling from Multiple Camera Image Data Based on the GFFD Deformation Method

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Active Contours/Snakes

Cost-efficient deployment of distributed software services

XV International PhD Workshop OWD 2013, October Machine Learning for the Efficient Control of a Multi-Wheeled Mobile Robot

Application of Improved Fish Swarm Algorithm in Cloud Computing Resource Scheduling

Dynamic wetting property investigation of AFM tips in micro/nanoscale

The Shortest Path of Touring Lines given in the Plane

Fitting: Deformable contours April 26 th, 2018

Support Vector Machines

Fast Computation of Shortest Path for Visiting Segments in the Plane

Anytime Predictive Navigation of an Autonomous Robot

EVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS

An Efficient Genetic Algorithm Based Approach for the Minimum Graph Bisection Problem

CS 534: Computer Vision Model Fitting

Prof. Feng Liu. Spring /24/2017

Topology Design using LS-TaSC Version 2 and LS-DYNA

Pose, Posture, Formation and Contortion in Kinematic Systems

Accounting for the Use of Different Length Scale Factors in x, y and z Directions

Positive Semi-definite Programming Localization in Wireless Sensor Networks

Sensory Redundant Parallel Mobile Mechanism

Exact solution, the Direct Linear Transfo. ct solution, the Direct Linear Transform

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning

A Saturation Binary Neural Network for Crossbar Switching Problem

Hermite Splines in Lie Groups as Products of Geodesics

Comparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments

Reliable and Efficient Routing Using Adaptive Genetic Algorithm in Packet Switched Networks

GSLM Operations Research II Fall 13/14

Classifier Swarms for Human Detection in Infrared Imagery

Mode-seeking by Medoidshifts

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data

ON THE ONE METHOD OF A THIRD-DEGREE BEZIER TYPE SPLINE CURVE CONSTRUCTION

DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT

Classifier Selection Based on Data Complexity Measures *

Design for Reliability: Case Studies in Manufacturing Process Synthesis

Optimal Design of Nonlinear Fuzzy Model by Means of Independent Fuzzy Scatter Partition

Clustering Algorithm Combining CPSO with K-Means Chunqin Gu 1, a, Qian Tao 2, b

Support Vector Machines. CS534 - Machine Learning

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

UNIT 2 : INEQUALITIES AND CONVEX SETS

Rational Interpolants with Tension Parameters

Complexity Analysis of Problem-Dimension Using PSO

Hierarchical clustering for gene expression data analysis

Load-Balanced Anycast Routing

Resolving Ambiguity in Depth Extraction for Motion Capture using Genetic Algorithm

Lecture #15 Lecture Notes

Cracking of the Merkle Hellman Cryptosystem Using Genetic Algorithm

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

VFH*: Local Obstacle Avoidance with Look-Ahead Verification

APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT

Predator-Prey Pigeon-Inspired Optimization for UAV Three-Dimensional Path Planning

Adaptive Weighted Sum Method for Bi-objective Optimization

The Codesign Challenge

The ray density estimation of a CT system by a supervised learning algorithm

A NEW HYBRID APPROACH FOR PREDICTION OF MOVING VEHICLE LOCATION USING PARTICLE SWARM OPTIMIZATION AND NEURAL NETWORK

The Research of Support Vector Machine in Agricultural Data Classification

Lecture 5: Multilayer Perceptrons

Feature Selection for Target Detection in SAR Images

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Air Transport Demand. Ta-Hui Yang Associate Professor Department of Logistics Management National Kaohsiung First Univ. of Sci. & Tech.

Mathematics 256 a course in differential equations for engineering students

Constructing Minimum Connected Dominating Set: Algorithmic approach

Path Planning and State Estimation for Unmanned Aerial Vehicles in Hostile Environments

Simplification of 3D Meshes

The Comparison of Calibration Method of Binocular Stereo Vision System Ke Zhang a *, Zhao Gao b

Reducing Frame Rate for Object Tracking

Integration of Planning and Control in Robotic Formations

Module 6: FEM for Plates and Shells Lecture 6: Finite Element Analysis of Shell

Graph-based Clustering

Optimization of machining fixture layout for tolerance requirements under the influence of locating errors

Review of approximation techniques

Inverse kinematic Modeling of 3RRR Parallel Robot

Analysis on the Workspace of Six-degrees-of-freedom Industrial Robot Based on AutoCAD

Wishing you all a Total Quality New Year!

Transcription:

Pat Plannng for Formaton Control of Autonomous Vecles 1 E.K. Xdas, 2 C. Palotta, 3 N.A. Aspragatos and 2 K.Y. Pettersen 1 Department of Product and Systems Desgn engneerng, Unversty of te Aegean, 84100 Ermoupols, Syros, Greece Emal: xdas@aegean.gr 2 Department of Engneerng Cybernetcs, Norwegan Unversty of Scence and Tecnology, NTNU, NO-7491 Trondem, Norway Emal: {claudo.palotta, krstn.y.pettersen}@tk.ntnu.no 3 Department of Mecancal Engneerng & Aeronautcs, Unversty of Patras, 82200 Patras Greece Emal: asprag@mec.upatras.gr Abstract. In ts paper a two-stage approac s ntroduced for optmum pat plannng of a team of autonomous vecles n an envronment cluttered wt obstacles. Te vecles are requested to move n formaton from an ntal pont to a fnal pont. Te Bump-Surface concept s used for te representaton of te envronment wle te formaton of te vecles s presented by a deformable Delaunay trangulaton. Te proposed approac s presented n detal and test cases wt multple vecles are smulated to demonstrate te effcency of te metod. Keywords. Pat Plannng, Formaton Control, Autonomous Vecles 1 Introducton Teams of autonomous vecles are wdely used n many applcatons, were te vecles are requested to meet formatons or oter constrants to accompls complex tasks, suc as transportaton of large obects [1], localzaton and mappng [2], searc and rescue mssons [3]. Ts nterest s motvated by te necessty of avng more vecles performng tasks wc are more dffcult to perform wt only one vecle, for nstance survellance mssons [4]. Furtermore, te moton n formaton s partcularly mportant wen spatally dstrbuted tasks ave to be accomplsed, lke for nstance, source seekng mssons [5]. In ts paper we present an approac for te pat plannng problem for a mult-vecle system. In partcular, we consder a mult-vecle system wc s consstng of autonomous vecles. Te obectve s to fnd an optmal pat for eac vecle wc

connects an ntal pont wt a fnal pont wle smultaneously te vecles sould be movng n a gven formaton. Te pat plannng problem for formatons control of a team of autonomous vecles as been nvestgated [6]-[8]. In [6], te autors presented a metod based on rapdly explorng random trees (RRT) for pat plannng of formatons wt under-actuated vecles. Ts metod randomly samples te envronment and cooses a free collson confguraton for eac vecle. Te autors revsed te classcal RRT to generate feasble pats for non-olonomc vecles. Furtermore, tey desgned a prorty strategy, wc makes te vecles to move n a gven formaton. Te work [7] descrbes a metod based on Vorono Fast Marcng (VFM) for formatons of fully actuated moble robots. Ts metod can be classfed as a potental feld metod but avods te drawbacks related to local mnma. In [8] an abstract manfold A was defned wc s te product of two manfolds G and S. Te manfold G s a Le Group, wc captures nformaton about te orentaton and poston of every vecle, wle S s a manfold, wc captures nformaton about te sape of te group of vecles. Te states n te two manfolds G and S are controlled ndependently. In ts paper we extend te metod proposed n [9] and [10] to mult-vecle systems consstng of autonomous vecles. Te vecles sould be movng n a gven formaton. By usng te proposed approac, t s possble to obtan smultaneously an optmum pat, for eac vecle. Eac pat s constructed consderng bot te envronment constrants and te formaton constrants. Te man contrbuton of ts paper s te ntroducton of a metod for te pat plannng of a flexble formaton of n autonomous vecles n an envronment cluttered wt statc obstacles. For te frst tme te formaton relatonsp s represented by a deformable Delaunay trangulaton, wc as te ablty to fnd a soluton even wen te vecles are requested to move troug narrow passages. Furtermore, te smootness of te pat s obtaned by controllng te angles between te controlpolygon segments, wc defne te system s pat. A multplcty of optmzaton crtera and constrants could be ncorporated easly to te formulated optmzaton problem accordng to te msson requrements of te team of vecles. 2 Basc assumptons and te two stage approac It s assumed tat a formaton of autonomous vecles sould move n a 2D envronment wc s cluttered wt known probted areas (obstacles-danger zones). A formal statement of te problem, te assumptons and te structure of te proposed metod are gven n ts secton. 2.1 Te pat plannng problem for a formaton of autonomous vecles Consder a team of 3 n ³ autonomous vecles wc sould move from an ntal to a goal locaton by keepng te desred formaton n a 2D envronment cluttered wt known statc obstacles. Te basc assumptons are:

Eac vecle s represented by a pont wc s movng only forward. Eac vecle s requested to move from an ntal pont S to a goal pont G nsde te desred formaton, wc sould not splt, wle te lengt of te pat sould be mnmum. Te pat of eac vecle sould be smoot. Te formaton s modeled as a deformable polygon. Te user defnes bot te mnmum and maxmum allowed formatons of te polygon. Te maxmum formaton s te desred one. 2.2 Te geometry and te representaton of te formaton We assume tat a team of autonomous vecles s enclosed n a deformable convex polygon, wle no splts are allowed. Te vertces of te convex polygon are te external vecles. Fg.1 sows a vsual representaton of a team of 15 pont-vecles. Te convex polygon s used n order to take te advantage of te fact tat te centrod R always les nsde te polygon. Furtermore, by usng convex polygon, we avod to ncrease te problem s complexty wc s appen wen we use non-convex polygons. Fg. 1. A team of 15 vecles (black crcles). Te correspondng: Delaunay trangles (das lnes) and te convex ull (black bold lne). Te Delaunay trangulaton [12] s used to facltate te geometrc relatons between te vecles wtn a geometry, were constrants could be defned easly. Generally, Delaunay trangulaton s caracterzed by ts smplcty and ts economy n data storage. Furtermore, te Delaunay trangulaton s ndependent of te order n wc te ponts are processed. For a team of n vecles we ave, 2n-2- m trangles and 3n-3-medges, were m s te number of vecles on te convex ull. Te lengt of mn max eac edge d, = 3,...,3n-3- m s assocated wt te constrant d d d, were mn max d s te mnmum safe dstance and d s te maxmum, wc s te desred dstance between a par of vecles.

3 Frst stage: te sub-optmal pat for te mnmum sze formaton In ts stage, te mnmum sze formaton s consdered as a fxed sape and te sortest pat s searced on te Bump-Surface. For te constructon of te Bump- Surface representng a gven 2D envronment, a normalzed workspace W s constructed by lnearly mappng te ntal envronment to [0,1]. Te constructon of te 2 correspondng Bump-Surface S s obtaned by a stragtforward extenson of te Z- value algortm [9]. It s assumed tat, te team as a reference pont wc les at te centrod of te polygon representng te mnmum sze formaton. Te reference pont traces a pat R() s n te normalzed W wc starts from te gven start pont and termnates at te desred goal pont. In order to defne R() s we use a B-Splne curve [13] to represent te pat of te fxed mnmum sze formaton: were, Q s te number of control ponts Q-1 2 () s = ån( s),0 s 1 = 0 R p (1) 2 p, N ( ) s are te B-Splne bass functons and 2 s te curve degree. Te goal of te proposed global pat plannng strategy s te determnaton of te poston of ( Q - 2) control ponts p wc defne te requested pat R () s. 3.1 Safe optmum moton of te mnmum formaton A safe pat R () s s one tat () does not collde wt te obstacles and () t s smoot. Followng te results from [10], te arc lengt of R () s approxmates te lengt L of ts mage SR ( ( s)) on te Bump-Surface S as long as R () s les onto te flat areas of S. Furtermore, n order to take nto account te sape of te formaton, a set of feature ponts A s selected on ts boundary accordng to ts sape and te requested accuracy [10]. Takng te above analyss nto consderaton, te pat plannng problem s formulated as an optmzaton problem wc s descrbed by, N ì m ï p ü a ï mnmze Ecomp = íl, åå H,1/ 1,...,1/ Q-2ý, subect to 180, = 1,..., Q -2 (2) ïî = 1 a= 0 ïþ were N denotes te number of ponts taken on R () s to dscretze t, s te -t p angle between te control-polygon segments and +1 R () s and H s te flatness

of ( s) A on S. Te R() s follows te sape of te defnng control polygon wc s derved by connectng te control ponts p [13]. Ten, a Genetc Algortm s adopted n order to searc for a soluton to te formulated optmzaton problem (Eq.(2)). A floatng pont representaton sceme s selected snce te coordnates of te control ponts and te angles of te control polygon are real numbers. A ftness assgnment strategy based on Pareto-optmal solutons called GPSIFF [14] s mplemented. Te followng tree genetc operators were selected. Reproducton: te proportonal selecton strategy s adopted, were cromosomes are selected to reproduce ter structures n te next generaton wt a rate proportonal to ter ftness. Crossover: te one-pont crossover was adopted. Mutaton: a boundary mutaton s used. 4 Stage 2: Determnng te smoot pat of eac vecle n te deformable formaton. Wt te reference pat R() s derved by te frst stage, te pat of every vecle s determned consderng te locaton of te vecles n te desred formaton. Snce te formaton as to pass troug areas, were te mnmum sze formaton s able to move wt safety, ten n te second pase a deformable formaton s consdered. A deformaton cost functon s formulated wt an optmum cost at te desred formaton. In ts stage eac vecle as ts own ndependent smoot pat but along ts pat t as to respect te desred formaton. In order to ensure tat te vecles n- m do not cross te border of te convex ull, te followng condton s taken nto account. At every pont R( s ), a= 1,..., N of te pat R() s te locaton of te vecle a p R, = 1,..., n-ms computed by te convex combnaton of te m vecles wc defne te convex ull. Terefore, were, te wegt factors m å R = w R, = 1,..., n-m (3) w satsfy, f f = 1 f f m wf ³ 0 and å wf = 1, = 1,..., n-m (4) Te goal of te proposed pat plannng strategy s te determnaton of te Q - 2 control ponts f = 1 q, wc defne te requested pat R () s for te -t vecle gven by te same equaton as Eq.(1). In order to take nto account tat te formaton of te vecles sould adapt to te geometrc caracterstcs of te envronment wle smultaneously tryng to keep te desred sape, te followng deformaton functon s proposed: max mn max ì k -, [, ] ï d d dî d d CD = e, = 3,...,3n -3- m, and k =í (5) ï îaverybg value (defned by te user), oterwse

Eq. (5) gves a penalzng functon, wc takes te optmum value wen d mn max and te worst wen d Ï [ d, d ]. Te mnmum sze polygonal sape of te formaton s obtaned wen d = d. mn Accordng to te above requrements te derved obectve functon s a vector wc s represented by, ì wf ³ 0 1 n ï mn E {,...,, 1,..., }, 3,...,3 3, m comp = L L CD CD = n- -m subecttoí ï å (6) wf = 1, = 1,..., n-m î f = 1 were L s te pat s lengt for te -t vecle wc s computed n a smlar way as n Stage 1. In te optmzaton problem defned by Eq. (6), te optmzaton varables are te control ponts wc defne te pat R () s of eac vecle and te wegt factors w. f A Mcro-GA s used to searc for a sub-optmum pat of eac vecle. Te man caracterstcs of te developed Mcro-GA are te followng: A floatng pont representaton sceme s selected for te cromosome syntax. Eac cromosome repre- sents a possble R () s as a sequence of te unknown control ponts q. A ftness assgnment strategy based on Pareto-optmal solutons s mplemented. It sould be notced tat te qualty of te ndvdual soluton generated n te ntal pase plays a crtcal role n determnng te qualty of te fnal optmal soluton. Tus, te soluton wc s derved from Stage 1 (te control ponts wc defne te pat R () s ), s used as an ntal soluton n te Mcro-GA (seedng) for eac vecle s pat R () s. Ts elps te Mcro-GA to converge n sort tme to te sub-optmal pat for eac vecle. Te same genetc operators as n Stage 1 are used except te mutaton operator wc s gnored. In most cases, a maxmum number of teratons (generatons) s defned n advance for te termnaton. However, t s dffcult to determne beforeand te number of generatons needed to fnd near-optmum solutons. Tus, an assessment of te qualty level of te Genetc Algortm s made on-lne. Te proposed algortm termnates eter wen te maxmum number of generatons s aceved or wen te same best cromosome appears for a maxmum number of generatons. = d max 5 Smulatons Te performance of te proposed metod s nvestgated troug a number of smulaton experments for a varety of formatons movng n 2D envronments. All smulatons are mplemented n Matlab. In all test cases, te grd sze s set to N g = 100. For te frst stage, te control parameters of te GA are te followng: populaton sze=250, maxmum number of generatons=500, crossover rate =0.75, boundary mutaton rate=0.004. For te Mcro-GA we set: populaton sze=50, maxmum number of generatons=30, crossover rate =0.75. It s wort notng tat te selecton of

te approprate control settngs s te result of extensve expermental efforts wt varous control scemes adopted followng te ndcatons of te lterature. Test Case: We assume te envronment of fg.2. Here, a team of 15 vecles s requested to move n formaton from te ntal ponts S to goal ponts G, = 1,...,15. Te convex ull s defned by sx vecles. A vsual representaton of bot te ntal and fnal formatons, te computed pat R() s and te correspondng Delaunay trangles are sown n Fg.2. Eac vecle s pat s defned by 8 control ponts. Te computed soluton takes about 10.24 mnutes. Furtermore, Fg. 2 sows te formaton of te convex ull wle te team of vecles s passng troug narrow passages, were te formaton s not ust srnked but t canged ts sape autonomously to adapt to te envronment. Fg. 2. Te resultng soluton pat R () s,te ntal and fnal convex ull wt te correspondng Delaunay trangles and te trace of te convex ull n two dfferent tme nstances. Despte te fact tat te problem under consderaton s off-lne, computatonal tme results wt respect to te number of te vecles s of mmense nterest. Te varaton of CPU s tme s ndcatve of te problem complexty. In tese experments, te envronment s te one sown n Fg.2 and te number of control ponts s constant, wle te number of vecles s canged from 3 to 15. Fg. 3 sows tat CPU tme ncreases almost lnearly wt te ncrease n te number of vecles. No of vecles Fg. 3. A CPU tme study

6 Concluson A new approac for te pat plannng of a deformable formaton of autonomous vecles s proposed. Te Delaunay trangulaton s proved to be very convenent for te modelng of a deformable formaton snce te extreme formatons (mnmum and maxmum allowed formatons) are defned by te user and tus we can easly determne te lmts of te dstances between te vecles. Furtermore, te smootness of te pat s obtaned by controllng te angles between te control-polygon segments, wc defne te system s pat. In future work te proposed approac sould be extended to sem-known 2D and 3D envronments, and n addton to te moton plannng te gudance and control of te vecles sould be consdered. References [1] Y. Wang and C. Slva, Sequental Q-learnng wt Kalman flterng for multrobot cooperatve transportaton, IEEE/ASME Trans. Mecatroncs, vol. 15, no. 2, pp. 261 268, Apr. 2010. [2] H. Cen, D. Sun, J. Yang, J. Cen, SLAM based global localzaton for mult-robot formatons n ndoor envronment, IEEE/ASME Trans. Mecatroncs, vol. 15, no. 4, pp. 561 574, Aug. 2010. [3] Z. Tang and U. Ozguner, Moton plannng for mult-target survellance wt moble sensor agents, IEEE Trans. Robot., vol. 21, no. 5, pp. 898 908, Oct. 2005. [4] M. Saska, J. Cudoba, L. Precl, J. Tomas, G. Loanno, A. Tresnak, V. Vonasek, and V. Kumar, Autonomous deployment of swarms of mcro-aeral vecles n cooperatve survellance, 2014 Int. Conf. Unmanned Arcr. Syst. ICUAS 2014 - Conf. Proc., pp. 584 595, 2014. [5] J. Han and Y. Cen, Multple UAV formatons for cooperatve source seekng and contour mappng of a radatve sgnal feld, J. Intell. Robot. Syst. Teory Appl., vol. 74, pp. 323 332, 2014. [6] S. Lu, D. Sun, and C. Zu, A dynamc prorty based pat plannng for cooperaton of multple moble robots n formaton formng, Robot. Comput. Integr. Manuf., vol. 30, no. 6, pp. 589 596, 2014. [7] S. Garrdo, L. Moreno, and P. U. Lma, Robot formaton moton plannng usng Fast Marcng, Rob. Auton. Syst., vol. 59, pp. 675 683, 2011. [8] C. Belta and V. Kumar, Moton generaton for formatons of robots: A geometrc approac, Proc. 2001 ICRA. IEEE Int. Conf. Robot. Autom. (Cat. No.01CH37164), vol. 2, 2001. [9] P.N. Azarads, N.A. Aspragatos, Obstacle representaton by Bump-surfaces for optmal motonplannng. Robotcs and Autonomous Systems 51 (2-3), pp. 129-150, 2005. [10] E.K. Xdas, P.N. Azarads, Msson desgn for a group of autonomous guded vecles. Robotcs and Autonomous Systems 59 (1), pp. 34-43, 2011. [11] M.S. LaValle, Plannng Algortms. Unversty of Illnos, 2004. [12] Preparata F. P., Samos M.I. Computatonal Geometry. Sprnger-Verlag, New York, 1985. [13] L. Pegl, W. Tller, Te NURBS Book. Sprnger-Verlag Berln Hedelberg, 1997. [14] J.H. Cen and S.Y. Ho, A novel approac to producton plannng of flexble manufacturng systems usng an effcent mult-obectve genetc algortm. Internatonal Journal of Macne Tools & Manufacture, no.45, pp.949-957, 2005.