This action might not be possible to undo. Are you sure you want to continue?

Xin-She Yang Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, UK. New Address: Mathematics and Scientiﬁc Computing, National Physical Laboratory, Teddington TW11 0LW, UK.

Abstract Engineering optimization is typically multiobjective and multidisciplinary with complex constraints, and the solution of such complex problems requires eﬃcient optimization algorithms. Recently, Xin-She Yang proposed a bat-inspired algorithm for solving nonlinear, global optimisation problems. In this paper, we extend this algorithm to solve multiobjective optimisation problems. The proposed multiobjective bat algorithm (MOBA) is ﬁrst validated against a subset of test functions, and then applied to solve multiobjective design problems such as welded beam design. Simulation results suggest that the proposed algorithm works eﬃciently.

arXiv:1203.6571v1 [math.OC] 29 Mar 2012

Keywords: Bat algorithm; cuckoo search; ﬁreﬂy algorithm; optimisation; multiobjective optimisation. Reference to this paper should be made as follows: Yang, X. S., (2011), Bat Algorithm for Multiobjective Optimization, Int. J. Bio-Inspired Computation, Vol. 3, No. 5, pp.267-274.

1

Introduction

Design optimisation in engineering often concerns multiple design objectives under complex, highly nonlinear constraints. Diﬀerent objectives often conﬂict each other, and sometimes, truly optimal solutions do not exist, and some tradeoﬀ and approximations are often needed. Further to this complexity, a design problem is subjected to various design constraints, limited by design codes or standards, material properties and choice of available resources and costs (Deb, 2001; Farina et al., 2004). Even for global optimisation problems with a single objective, if the design functions are highly nonlinear, global optimality is not easy to reach. Metaheuristic algorithms are very powerful in dealing with this kind of optimization, and there are many review articles and excellent textbooks (Coello, 1999; Deb, 2001; Isasi and Hernandez, 2004; Yang, 2008; Talbi, 2009; Yang, 2010c). In contrast with single objective optimization, multiobjective problems are much diﬃcult and complex (Coello, 1999; Floudas et al., 1999; Gong et al., 2009; Yang and Koziel, 2010). Firstly, no single unique solution is the best; instead, a set of non-dominated solutions should be found in order to get a good approximation to the true Pareto front. Secondly, even if an algorithm can ﬁnd solution points on the Pareto front, there is no guarantee that multiple Pareto points will distribute along the front uniformly, often they do not. Thirdly, algorithms work well for single objective optimization usually do not directly work for multiobjective problems, unless under special circumstances such as combining multiobjectives into a single objective using some weighted sum methods. Substantial modiﬁcations are often needed. In addition to these diﬃculties, a further challenge is how to generate solutions with enough diversity so that new solutions can sample the search space eﬃciently (Talbi, 2009; Erfani and Utyuzhnikov, 2011; Yang and Koziel, 2011).

1

2001. Microbats use a type of sonar.2 to 11cm.5 to 2g) to the giant bats with wingspan of about 2 m and weight up to about 1 kg. Microbats typically have forearm length of about 2. Konak et al. the rate of pulse emission can be sped up to about 200 pulses per second when they ﬂy near their prey. Zhang and Li. 1995. In this paper. 2010a. let us brieﬂy review the basics of the bat algorithm for single objective optimization. and locate their roosting crevices in the dark. avoid obstacles. Their pulses vary in properties and can be correlated with their hunting strategies. we will outline the basic ideas and steps of the proposed algorithm. In addition. Though each pulse only lasts a few thousandths of a second (up to about 8 to 10 ms). such as bi-objective beam design. 2006. 1995. while others more often use constant-frequency signals for echolocation.. 2009.. 2006. 2. Deb. 2007. studies shows the integration time of the bat ear is typically about 300 to 400 µs. Yang. Yang. depending on the species. Each ultrasonic burst may last typically 5 to 20 ms. Their size ranges from the tiny bumblebee bat (of about 1. We will ﬁrst validate it against a subset of multiobjective test functions. They are the only mammals with wings and they also have advanced capability of echolocation. Yang. called bat algorithm (BA). 2000). microbats are a famous example as microbats use echolocation extensively while megabats do not (Richardson. In fact. Recently. Reyes-Sierra and Coello. it has a constant frequency which is usually in the region of 25kHz to 150 kHz. As the speed of sound 2 . they often mimic the successful characteristics in nature.Furthermore. 2005. 2010b). and microbats emit about 10 to 20 such sound bursts every second. a new metaheuristic search algorithm. Osyczka and Kundu. has been developed by Xin-She Yang (2010a). Most bats uses echolocation to a certain degree. For example. among all the species. Such short sound bursts imply the fantastic ability of the signal processing power of bats. real-world optimization problems always involve certain degree of uncertainty or noise. please refer to Yang (2011). 2002.1 Echolocation of Microbats Bats are fascinating animals. Rangaiah. multiobjective optimization has many powerful algorithms with many successful applications (Abbass and Sarker. 2007. materials properties for a design product may vary signiﬁcantly. For example. frequencymodulated signals to sweep through about an octave. Colin. Luna et al. When hunting for prey. 2 Bat Behaviour and Bat Algorithm In order to extend the bat-inspired algorithm for single optimization to solve multiobjective problems. 2008. 2011). 1995. echolocation. Many new algorithms are emerging with many important applications (Kennedy and Eberhart. an optimal design should be robust enough to allow such inhomogeneity and also provides good choice for decision-makers or designers.. we will apply it to solve design optimization problems in engineering. Most bats use short. 1996. 2010b. metaheuristic algorithms start to emerge as a major player for multiobjective global optimization. however. Yang et al. a new cuckoo search algorithm was developed by Xin-She Yang and Suash Deb (2009) and more detailed studies by the same authors (Yang and Deb. Then. 2004). though some species can emit higher frequencies up to 150 kHz. 2004. we will extend BA to solve multiobjective problems and formulate a multiobjective bat algorithm (MOBA). These bats emit a very loud sound pulse and listen for the echo that bounces back from the surrounding objects. Banks et al. we will discuss the unique features of the proposed algorithm as well as topics for further studies. 2009. Tabli. to detect prey. Then. It is estimated that there are about 996 diﬀerent species which account for up to 20% of all mammal species (Altringham. especially biological systems (Kennedy and Eberhart. 2010a) suggested that it is very eﬃcient for solving nonlinear engineering design problems. Yang. Finally. For a recent review of popular metaheuristics. Farina et al. 2008. 2010c. Despite these challenges. 2008). Cui and Cai. Marler and Arora. Their signal bandwidth varies depends on the species. Yang and Deb. Preliminary studies show that it is very promising and could outperform existing algorithms.. The typical range of frequencies for most bat species are in the region between 25kHz and 100kHz. and often increased by using more harmonics.. called.

which is in the range of 2mm to 14mm for the typical frequency range from 25kHz to 150 kHz.7mm to 17mm in reality. Here x∗ is the current global best location (solution) which is located after comparing all the solutions among all the n bats at each iteration t. 500kHz] corresponds to a range of wavelengths from 0. Such echolocation behaviour of microbats can be formulated in such a way that it can be associated with the objective function to be optimized. In reality. we can choose the ranges freely to suit diﬀerent applications. λmax ]. and most bats also have very sensitive smell sense. Although the loudness can vary in many ways. 1]. Another obvious simpliﬁcation is that no ray tracing is used in estimating the time delay and three dimensional topography. and they also ‘know’ the diﬀerence between food/prey and background barriers in some magical way. In general the frequency f in a range [fmin . In our implementation. The new solutions xt and velocities v t at time step t are i i given by fi = fmin + (fmax − fmin)β. They can automatically adjust the wavelength (or frequency) of their emitted pulses and adjust the rate of pulse emission r ∈ [0. for simplicity. 1996). 2. the type of prey. 1] is a random vector drawn from a uniform distribution. the time diﬀerence between their two ears.3 Bat Motion For the bats in simulations. as it is more computationally extensive in multidimensional cases.2 Bat Algorithm If we idealize some of the echolocation characteristics of microbats. the following approximate or idealized rules were used. Bats ﬂy randomly with velocity v i at position xi with a frequency fmin . For example a frequency range of [20kHz. In the basic bat algorithm developed by Xin-She Yang (2010a). Though this might be a good feature for the application in computational geometry. i i i xt+1 = xt + v t . varying wavelength λ and loudness A0 to search for prey. some bats have good eyesight. they will use all the senses as a combination to maximize the eﬃcient detection of prey and smooth navigation. (1) v t+1 = v t + (xt − x∗ )fi . 2. depending on the proximity of their target. we can develop various batinspired algorithms or bat algorithms. and even the moving speed of the prey such as small insects (Altringham.in air is typically v = 340 m/s. Studies show that microbats use the time delay from the emission and detection of the echo. However. we assume that the loudness varies from a large (positive) A0 to a minimum constant value Amin . we will not use this feature. i i i (2) (3) where β ∈ [0. They can detect the distance and orientation of the target. we can use fi (or λi ) to adjust the velocity change while ﬁxing the other factor λi (or fi ). Obviously. 3. fmax ] corresponds to a range of wavelengths [λmin . In addition to these simpliﬁed assumptions. we also use the following approximations. the wavelength λ of the ultrasonic sound bursts with a constant frequency f is given by λ = v/f . here we are only interested in the echolocation and the associated behaviour. we have to deﬁne the rules how their positions xi and velocities v i in a d-dimensional search space are updated. we will use fmin = 0 and fmax = O(1). All bats use echolocation to sense distance. 1. and this makes it possible to formulate new optimization algorithms. As the product λi fi is the velocity increment. depending on the domain 3 . and the loudness variations of the echoes to build up three dimensional scenario of the surrounding. 2. Such wavelengths are in the same order of their prey sizes. Obviously. however. depending on the type of the problem of interest.

1]. and at least one component is smaller. while the rate of pulse emission increases. a new solution for each bat is generated locally using random walk xnew = xold + ǫ At . we can also use A0 = 1 and Amin = 0. the loudness Ai and the rate ri of pulse emission have to be updated accordingly as the iterations proceed. we can deﬁne another dominance relationship by u v ⇐⇒ u ≺ v ∨ u = v. BA can be considered as a balanced combination of the standard particle swarm optimization and the intensive local search controlled by the loudness and pulse rate... Similarly. and we have used α = γ = 0. For any 0 < α < 1 and γ > 0. as t → ∞. no component of u is larger than the corresponding component of v.4 Loudness and Pulse Emission Furthermore. n} : ui < vi . . we can use α = γ. .. each bat is randomly assigned a frequency which is drawn uniformly from [fmin ... (7) It is worth pointing out that for maximization problems. vn )T if and only if ui ≤ vi for ∀i ∈ {1. i (6) In the simplest case. In other words. The Pareto front P F of a multiobjective can be deﬁned as the set of non-dominated solutions so that P F = {s ∈ S ∃ s′ ∈ S : s′ ≺ s}. . we can use A0 = 100 and Amin = 1. the dominance can be deﬁned by replacing ≺ with ≻. As the loudness usually decreases once a bat has found its prey. n} and ∃i ∈ {1. 3 Multiobjective Bat Algorithm Multiobjective optimization problems are more complicated than single objective optimization. / (8) 4 . (4) where ǫ is a random number vector drawn from [−1. assuming Amin = 0 means that a bat has just found the prey and temporarily stop emitting any sound.. un )T ∈ F . fmax ]. For simplicity. 1983). In fact. Therefore.. Initially. Now we have t 0 At+1 = αAt .. a point x∗ ∈ F is called a non-dominated solution if no solution can be found that dominates it (Coello. ri = ri [1 − exp(−γt)]. .. The update of the velocities and positions of bats have some similarity to the procedure in the standard particle swarm optimization.. 2. For example.. To a degree. is said to dominate another vector v = (v1 . In addition. ri → ri . Preliminary studies by Yang (2010a) suggested that bat algorithm is very promising for solving nonlinear global optimization problems. and we have to ﬁnd and/or approximate the optimality fronts. 3. (5) i i where α and γ are constants. as fi essentially controls the pace and range of the movement of the swarming particles. while At =<At > is the average loudness of i all the bats at this time step.9 in our simulations. Now we extend it to solve multiobjective optimization problems.size of the problem of interest. once a solution is selected among the current best solutions. we have t 0 At → 0. the loudness can be chosen as any value of convenience. algorithms have to be modiﬁed to accommodate multiobjectives properly. α is similar to the cooling factor of a cooling schedule in the simulated annealing (Kirkpatrick et al.1 Pareto Optimality A solution vector u = (u1 .. For the local search part. 1999).

(10) As the weights are generated randomly from a uniform distribution.. / (9) where f = (f1 . the basic steps of the multiobjective bat algorithm (MOBA) can be summarized as the pseudo code shown in Fig. it is possible to vary the weights with suﬃcient diversity so that the Pareto front can be approximated correctly. 2. Konak et al.. 0 For example. 1. and increase ri & reduce Ai end if Rank the bats and ﬁnd the current best x∗ end while Record x∗ as a non-dominated solution end Postprocess results and visualization Figure 1: Multiobjective bat algorithm (MOBA). n) and v i for j = 1 to N (points on Pareto fronts) K Generate K weights wk ≥ 0 so that k=1 wk = 1 K Form a single objective f = k=1 wk fk while (t <Max number of iterations) Generate new solutions and update by (1) to (3) if (rand > ri ) Random walk around a selected best solution end if Generate a new solution by ﬂying randomly if (rand < Ai & f (xi ) < f (x∗ )) Accept the new solutions. 3. we use a weighted sum to combine all objectives fk into a single objective K K f= k=1 wk fk . and we have used α = γ = 0. To obtain a good approximation to Pareto front.. xd )T Initialize the bat population xi (i = 1.. The choice of parameters requires some experimenting. x = (x1 ..9 for all simulations.. 2006). Their loudness and emission rates will be updated only if the new solutions are improved.. or in term of the Pareto optimal set in the search space P F ∗ = {x ∈ F ∃ x′ ∈ F : f (x′ ) ≺ f (x)}. each bat should have diﬀerent values of loudness and pulse emission rate. fK (x).. .. a diverse range of solutions should be generated using eﬃcient techniques (Gujarathi and Babu. 5 . and this can be achieved by randomization. or any value ri ∈ [0. . 1] if using (5). In our simulations.... For simplicity here.Objective functions f1 (x). the initial loudness A0 can typically be [1.. 2009. which means that these bats are moving towards the optimal solution. we have carried out parametric studies. Initially.2 MOBA Algorithm Based on these approximations and idealization. . while the initial emission rate ri can be i 0 around zero. . fK )T . 2]. k=1 wk = 1.

30.. non-convex and discontinuous Pareto fronts. 0.7 to 0. LZ4 function (Li and Zhang. 100. which is suﬃcient for most problems. xi ∈ [0. 1 and γ = 0.8. α = 0. ZDT2 function with a non-convex front f1 (x) = x1 . Zitzler and Thiele. 0. 20. By varying n = 5. but a subset of a few widely used functions provides a wide range of diverse properties in terms Pareto front and Pareto optimal set.9. 10. ZDT3 function with a discontinuous front f1 (x) = x1 . 1999. 0. α = 0. This may be due to the fact that manipulations of large matrices or longer vectors usually take longer. Zitzler et al. and pulse reduction rate γ. 300. but also suitable to compare the closeness of Pareto front of diﬀerent functions.4 4.9. 2009. we found that to increase of n typically leads to a longer computing time than to re-run the program a few times.. depending on the problem of interest. d−1 where d is the number of dimensions. 0. (11) xi .9. 1.. we have tested the following four functions: ZDT1 function with a convex front (Zitzler and Thiele 1999. Zhang and Li. and computing time is within a few seconds to less than a minute. The Pareto-optimality is reached when g = 1. Zitzler et al.5. We can either use a given tolerance or a ﬁxed number of iterations.. . 2000). From the implementation point of view. a ﬁxed number of iterations is not only easy to implement. f1 varies from 0 to 0. 50 to 50.7 to 0. we will use the ﬁxed parameters: n = 50. 15.95. 30. g g f2 (x) = g(1 − f1 2 ) . Zhang et al. Another possibility is that simple restart can increase the diversity of solutions than more intensive search for longer iterations. 0.9 and γ = 0. we found that the best parameters for most applications are: n = 25 to 50. 2000) f1 (x) = x1 .1 Numerical Results Parametric Studies The proposed multiobjective bat algorithm (MOBA) is implemented in Matlab. 0. 0. In order to generate more optimal points on the Pareto front. We have tested it using a diﬀerent range of parameters such as population size (n). i = 1. The stopping criterion can be deﬁned in many ways. g where g in functions ZDT2 and ZDT3 is the same as in function ZDT1. To be more speciﬁc in this paper. 1985.2 Multiobjective Test Functions There are many diﬀerent test functions for multobjective optimization (Schaﬀer. 500. we have selected a subset of these functions with convex.7.9. we can do it in two ways: increase the population size n or run the program a few more times. we can also increase it to a larger number.6. To validate the proposed MOBA. 250. α = γ = 0. in all our simulations. 150.8. 200. Therefore. We also include functions with more complex Pareto sets. 2009. g =1+ 9 d i=2 f2 (x) = g(1 − f1 /g). 1]. In the ZDT3 function. Through simulations.6..5. 0. which is easily done within a few minutes. loudness reduction α. f2 (x) = g 1 − f1 f1 − sin(10πf1 ) .773 to 1. Zhang et al. 4.852 and f2 from −0.95. So we have set the ﬁxed number iterations as 5000.7. 0. 2007) f1 = x1 + 2 |J1 | 6 h(uj ) j∈J1 . 2003. 400. 0. 0. So to generate 200 points using a population size n = 50 requires to run the program 4 times. If necessary.

we will solve this design benchmark using MOBA.. Rangaiah. (15) where N is the number of points. We can see that exponential convergence can be achieved in all cases. In all the rest of the ﬁgures. and h(v) = |v| . d j = 2.. uj = xj − sin(6πx1 + jπ ). 3 and 4 show the exponential-like decrease of Ef as the iterations proceed. x1 ∈ [0.Table 1: Summary of results. 1]. 2). the depth d and thickness h of the main beam. the Pareto front generated by MOBA is compared √ with the true front f2 = 1 − f1 of ZDT1 (see Fig. 1999. The objectives are to minimize both the overall fabrication cost and the end deﬂection δ. The problem has four design variables: the width w and length L of the welded area.4E-4 5. 2002). 2009. As a result. Gong et al. In the rest of this paper. d x1 ∈ [0. j∈J2 (12) where J1 = {j|j is odd and 2 ≤ j ≤ d} and J2 = {j|j is even and 2 ≤ j ≤ d}.9E-4 Errort=5000 4. there are many diﬀerent benchmarks with detailed studies in the literature (Pham and Ghanbarzadeh. 5 Engineering Optimization Design optimization.7E-4 2.. 3. . Among the widely used benchmarks. The results for all the functions are summarized in Table 1. j = 2. the welded beam design is a well-known design problem. 2008). 7 . d. Figs. Multiobjective design of a welded beam is a classical benchmark which has been solved by many researchers (Deb. 1]. especially design of structures. (14) After generating 200 Pareto points by MOBA. Ray and Liew..2E-16 f2 = 1 − x2 + 1 2 |J2 | h(uj ). 2002.2E-19 1. 2007.5E-17 3.7E-15 1. the vertical axis is for f2 while the horizontal axis is for f1 . We can see clearly that our MOBA algorithm indeed converges almost exponentially. xj ∈ [−2. .. The least-square distance from the estimated front to the true front of ZDT1 for the ﬁrst 1000 iterations (Fig. 1 + e2|v| (13) 2 This function has a Pareto front f2 = 1 − f1 for 0 ≤ f1 ≤ 1 with a Pareto set xj = sin(6πx1 + jπ ). Ray and Liew. The convergence property can be viewed by following the iterations. 4). 3) and the logarithmic scale for 5000 iterations (Fig. 2]. d. Let us deﬁne the distance or error between the estimate Pareto front P F e to its correspond true front P F t as N Ef = ||P F e − P F t ||2 = j=1 (P Fje − P Ft )2 . Functions ZDT1 ZDT2 ZDT3 LZ4 Errort=2000 3.2E-5 2... has many applications in engineering and industry.

6 0.2 0 0 0.2 0. 8 .8 0.1 True PF MOBA 0.4 0.4 0.6 0. 6 5 4 3 2 1 x 10 4 0 0 200 400 600 800 1000 Figure 3: Convergence of the MOBA.8 1 Figure 2: Estimated front and true front for ZDT1.

d ≤ 10 and 0.125 − w ≤ 0. g6 (x) = 0. 2 L2 + (w + d)2 .0 ≤ 0.10471w2 + 0. (19) 6 28 The simple limits or bounds are 0. subject to g1 (x) = w − h ≤ 0.000hd3 .10 0 10 −5 10 −10 10 −15 0 1000 2000 3000 4000 5000 Figure 4: Exponential convergence of the MOBA. g2 (x) = δ(x) − 0. d 30/48 dh3 (1 − ). 2002. The simulations for these benchmarks and functions suggest that MOBA is a very eﬃcient algorithm for multiobjective optimization. In addition. By using the MOBA. 2007). 6000 √ .25 ≤ 0.61423 × 106 9 . 5. minimize f2 = δ. 000 ≤ 0.1 ≤ L. h ≤ 2. (w+d)2 ]. 65.125 ≤ w. g4 (x) = σ(x) − 30. This is consistent with the results obtained by others (Ray and Liew. g7 (x) = 6000 − P (x) ≤ 0. (17) g5 (x) = 0. we have solved this design problem. J = β= √ 2 2 wL[ L + 6 QD J . The approximate Pareto front generated by the 50 non-dominated solutions after 1000 iterations are shown in Fig.04811dh(14. the results are more smooth with fewer iterations. Pham and Ghanbarzadeh.0 + L). P = 0.000 hd2 . 600 ≤ 0.856 30. The problem can be written as minimise f1 (x) = 1. 2wL (18) τ (x) = α2 + αβL D + β2. (16) g3 (x) = τ (x) − 13. 504.04811hd(14 + L) − 5.10471w2L + 0.0. It can deal with highly nonlinear problems with complex constraints and diverse Pareto optimal sets. where σ(x) = D= δ= α= 1 2 Q = 6000(14 + L 2 ).

In addition. T. we have successfully formulated a new algorithm for multiobjective optimization.x 10 8 −3 6 4 2 0 0 10 20 30 40 Figure 5: Pareto front for the bi-objective beam design. Artiﬁcial Intelligence Tools. and Sarker R. (2009) ‘Integral particle swarm optimisation with dispersed accelerator information’. combinatorial. hybridization with other algorithms may also prove to be fruitful. A. of 1999 Congress on Evolutionary Computation. and some combination with these techniques may improve MOBA even further. 11(4). In this paper..: The Varienty of Life. DOI 10. and Cai X. CEC99. J.: Bats: Biology and Behaviour. Additional tests and comparison of the proposed are highly needed. Fundam.781901 [5] olin.. including discrete and mixed type of optimization problems. we will focus on the parametric studies for a wider range of test problems. Further research can also emphasize the performance comparison of this algorithm with other popular methods for multiobjective optimization. and Anyakoha C. The proposed MOBA has been tested against a subset of well-chosen test functions. The Pareto diﬀential evolution algorithm.. D. References [1] bbass H. We will try to test the diversity of the Pareto front it can generate so as to identify the ways to improve this algorithm to suit a diverse range of problems. multicriteria and constrained optimization. Oxford University Press. (2008). (1996). Vol. J. and indicative applications. based on the recently developed bat algorithm. Part II: hydridisation. Vincent J. A... (2000). and then been applied to solve design optimization benchmarks in structural engineering. H. J. 95. There are a few eﬃcient techniques to generate diverse Pareto fronts (Erfani and Utyuzhnikov 2011). Int. Oxford Univesity Press. in: Proc. multiobjective bat algorithm. [3] anks A.1109/CEC. C. In the future work. A review of particle swarm optimization. 10 . namely. 531-552 (2002). (2002). (1999). 6 Conclusions Multiobjective optimization problems are typically very diﬃcult to solve. [4] oello C. [2] ltringham. 109-124 (2008).. Inform. Natural Computing.1999. 427–447. An updated survey of evolutionary multiobjective optimization techniques: state of the art and future trends. Results suggest that MOBA is an eﬃcient multiojective optimizer. [6] ui Z.

. New York.nhm. and Babu B. Science. Deb K. [12] ong W. Evol. 670-680 (1983). (2004).. Multi-Objective Optimisation using the Bees Algorithm. Zhu L. Also... Scheiger C. Engineering Optimization. Comput. [11] loudas C. Klepeis J. (1999).. Struct. 137-157 (2009). Dunbeath. Survey of multi-objective optimization methods for engineering.. L. T. D. Springer-Verlag. Harding S.. Wiley. Multiobjective optimization problems with complicated Paroto sets. Comput. 220. [24] ay L. Y. 8. 34(2).. P. NJ.. and Smith A. (2002)... 1st World Congr. Comput. (2011). Optimization. (EvoCOP’07). Dynamic multiobjective optimization problems: test cases. (1995). and Kundu S. (2008). of IEEE International Conference on Neural Networks. Int. C. T. and applications.. (2009). M... Optim. (2009). Improved Strategies of Multi-objective Diﬀerential Evolution (MODE) for Multi-objective Optimization. Optimization by simulated annealing. (in press) [10] arina M. P. Evolutionary algorithms for multi-criterion optimization in engieering design. 7th Euro. F. and Utyuzhnikov S. Meyer C. IEEE Trans.. 135-161.. S. December 16-18.. MOEA/D and NSGA-II. 369-395. [8] eb K.. (2007). 2009. Cai Z. Gumus Z. Optim. E.: Bats. Gellat C. A. [26] ichardson. Scotland. Multidisc. [17] onak A. A swarm metaphor for multiobjective design optimization. and Coello C. Natural History Museum. Multi-objective Optimization: Techniques and Applications in Chemical Engineering... (2001). Optim.uk 11 . 425-442. http://www.: The secrete life of bats. [19] una F. A. V. Evolutionary algorithms for real-world instances of the automatic frequency planning problem in GSM networks. P. Handbook of Test Problems in Local and Global Optimization. H. and Vecchi M. Proc. 284-302. 1942-1948. pp. [21] syczka A. Struct. 909-914. Evol. Esposito W. An eﬀective multiobjective diﬀerential evolution algorithm for engineering design. pp. and Amota P. Conf. in: Proc.. Combin. (1999). (2006). Evol. S. A genetic algorithm-based multicriteria optimization method. (2007)... World Scientiﬁc Publishing. [14] sasi P. J. M. Reliability Engineering and System Safety. (1995). 992-1007. A. Coit D. and Pedraza S. in: Evolutionary Aglorithms in Engineering and Computer Science. (2004). 38. and Ghanbarzadeh A. (2004). Comp. 141-153. Proc. 20(3). [13] ujarathi A.. C. Particle swarm optimization. Piscataway... Computational Intelligence.. T. Opt. [9] rfani T.. Directed search domain: a method for even generation of Pareto frontier in multiobjective optimization. in: 3rd International Virtual Conference on Intelligent Production Machines and Systems (IPROMS 2007): Whittles. and Eberhart R.. Proc. London. Eng. 445-449 [15] ennedy J. (2008). Multidisc. [18] i H.. W. and Arora J. and Hernandez J... IEEE Trans. Pardalos P. Adjiman C. Multi-objective particle swarm optimizers: A survey of the state-of-the-art. approximations. Struct. 26. John Wiley & Sons.. of 4th Indian International Conference on Artiﬁcial Intelligence (IICAI-09). Nebro A.[7] eb K. 2007 [23] angaiah G. Multidisc..ac. M. Introduction to the Applications of Evolutionary Computation in Computer Security. Elsevier Sciencce. [16] irkpatrick S. 2(3). A. (2006).. 13. R. [25] eyes-Sierra M.. C. 287-308. Richardson. Multi-Objective optimization using evolutionary algorithms.. [20] arler R.. [22] ham D. and Zhang Q. (2007).. and Liew K. pp. Multiobjective optimization using genetic algorithms: a tutorial... H. Alba E. Intelligence Res. 91. J. Springer.

(2007). (1985). (2010a) Engineering optimisation by cuckoo search. B... [38] ang X. Engineering Optimisation: An Introduction with Metaheuristic Applications.. (2011). MOEA/D: a multiobjective evolutionary algorithm based on decomposition. S. vol 4. Vol. J. [32] ang. and Thiele L.. Liu W. S. 317-323. [28] albi E.. Springer. S. Bio-Inspired Compuation. 2nd Edition. 1291-1294 (2010). in: Artiﬁcial Intelligence and Knowledge Engineering Applications: A Bioinspired Approach. S. and Clemson University.. John Wiley and Sons. Technical Report CES487. 93-100.. (2010c). Canberra. [37] ang X. [39] ang X. Deb K.D. F. (2009) Cuckoo search via L´vy ﬂights. [29] ang X. USA. 624 pp. in: Proc. Liang Y. Int.. 77-84 (2011). (2010b). R. April 2009. 330-343. Studies in Computational Intelligence. Genetic Aglorithms. Optimisation. Comparison of multiobjective evolutionary algorithms: Empirical results. [41] hang L. J. 136. pp. [30] ang X. (2000).. 712-731 (2007). Computational optimization. 3. H. [45] itzler E.. (2009). Nature-Inspired Metaheuristic Algorithms.). S.. pp.. S. 1. S... in: NDT 2011.. Comput. N.. Evol. and Li H. Modelling & Num. 1st Int.. [31] ang X. S. [43] hang Q. In: Proceedings of the 2003 congress on evolutionary computation (CEC2003).. December 2003 [42] hang Q. and Fong S. and Thiele L. in: Proc. Conf... Accelerated particle swarm optimization and support vector machine for business optimization and applications. and Koziel S.. [44] itzler E. Math. Engineering optimization via nature-inspired virtual bee algorithms... 24002405. SCI Vol. World Scientiﬁc Publishing. Lecture Notes in Computer Science. John Wiley and Sons. S. IEEE Press.. modelling and simulation – a paradigm shift. (2005). Multiobjective optimization test instances for the CEC 2009 special session and competition. Gonzalez et al. 1. Suganthan P. 210-214. Comp. in: Nature-Inspiref Cooperative Strategies for Optimization (NICSO 2010).. and Koziel S. Zhou A.. IEEE Evol. 284. [40] ang X. pp. pp. and Deb S. University of Essex. Computational Optimization and Applications in Engineering and Industry. Zhao S. Eagle strategy using L´vy walk and ﬁreﬂy algorithms for stochastic e optimization. pp. pp. Procedia in Computer Science. Luniver Press. and Deb S. C. Nanyang Technological University. Evol. of World Congress on Nature e & Biologically Inspired Computing (NaBic 2009). G. (2010b). S. (2009). S. pp. 284. [35] ang X. Communications in Computer and Information Science. Australia. Metaheuristics: From Design to Implementation.. Int. (1999). 8. 53-66.[27] chaﬀer J.. Multiobjective evolutonary algorithms: A comparative case study and the strength pareto approach. 257-271. Comput. Zhou C. Z. 3(2). [33] ang X. X. Liu X. F. IEEE Trans. [34] ang X. M. J. 11. 3562. [36] ang X. Springer. Review of metaheuristics and generalized evolutionary walk algorithm. in: Nature Inspired Cooperative Strategies for Optimization (NICSO 2010) (Eds. Ma Z. Solving multi objective optimization problems using particle swarm optimization. Tiwari S. 65-74.. UK. and Deb S. A new metaheuristic bat-inspired algorithm. (2003). IEEE Publications. 101-111. (2008) Introduction to Computational Mathematics. 173195 12 . Q. (2011). S.. Multiple objective optimization with vector evaluated genetic algorithms.-G. Deb S.. (2010a).

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd