C. A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275--308, 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Vol. 2, N. 6, December 2002. 68 Table IV Reactive Bounds Busbar number Q min Q max 2 20.0000 100.0000 5 15.0000 80.0000 8 15.0000 60.0000 11 10.0000 50.0000 13 15.0000 60.0000 This problem was solved using NSGA and the inequalities constraints were handled as objectives [9] [10], together with the P M elitism that was modified as described in [9] The first front, Fig. 7.a, was obtained using population size and generation number equal to 40 and 150, respectively. The load flow problem was solved using the fast decoupled method [11] The maximum error accepted to ....

C. A. C. Coello, `Treating Constraints as Objectives for Single-Objective Evolutionary optimization', Engineering Optimization, v. 32, no.3, pp. 275-308, 2000.

....points. In this paper, constraints in constrained multiobjective optimization problems are handled as objectives and the resulting problem is solved by the Niched Pareto Genetic Algorithm NPGA [1] Handling constraints as objectives was recently presented for single objective optimization [2]. The original NPGA was modified by incorporating the Parks Miller elitist technique (P M) 3] which needed some changes when constraints were treated as objectives. The required changes were essential to avoid convergence toward an infeasible space. Two analytical test problems, TBU [4] and ....

C. A. C. Coello, `Treating Constraints as Objectives for Single-Objective Evolutionary optimization', Engineering Optimization, v. 32, no.3, pp. 275-308, 2000.

....propose to use multiple scenarios as if they were separate objectives, and to use a multi objective EA to dynamically evolve adapted client admission strategies. A research topic that has gained attention is the transformation of a constrained single objective problem to a multi objective problem [5]. Another interesting development is the application of multi objective EAs to separate subproblems of a single objective problem [25] Analogously, we propose to transform dynamic single objective problems to multiple objectives. 2 De nition of the Problem 2.1 On line batching for Near Video on ....

Carlos A. Coello Coello. Treating constraints as objectives for single-objective evolutionary optimization. Engineering Optimization, 32(3):275-308, 2000.

....included in this bibliography. The list is arranged in alphabetical order by the name of the journal. Archives of Control Sciences, 285] Arti cial Intelligence, 356] Civil Engineering Systems, 132] Complex Systems, 403] Computers in Industry, 139] DIMACS, 196] Engineering Optimization, [136, 138] Engineering with Computers, 89] Evolutionary Computation, 87, 88, 95, 99, 112, 116, 412] Expert Systems with Applications, 114] Fuzzy Systems Arti cial Intelligence Reports and Letters, 176] IEE Proceedings C: Generation, Transmission and Distribution, 151] IEE Proceedings, Vision, ....

....[315] Chiva, Emmanual, 155] Chorafas, Dimitris N. 12] Christiansen, Alan D. 13, 52, 57, 63, 65, 89, 119, 132, 136] Chu, Chee Hung H. 154] Cli , David T. 156, 159, 161, 192, 367, 368, 370, 371, 372, 373, 374, 376] Clote, Peter, 86] Cockcroft, Victor, 248] Coello Coello, Carlos A. [13, 26, 32, 35, 36, 37, 40, 52, 57, 59, 60, 61, 62, 63, 65, 114, 119, 128, 130, 131, 132, 136, 137, 138, 139] Coello Coello, Carlos C. 89] Collins, Robert James, 398] Colombetti, Marco, 355, 358] Cootes, T. F. 45] Cord on, Oscar, 193, 221, 232, 239, 243, 252, 275, 285, 288, 290, 292] Coveney, Peter V. 125] Crutch eld, James P. 47, 387, 388] Cziko, Gary, 41] Das, Rajarshi, 31, 47] ....

[Article contains additional citation context not shown here]

Carlos A. Coello Coello. Treating constraints as objectives for single-objective evolutionary optimization. Engineering Optimization, 32, 1999. (Accepted for publication; URL: http://www.lania.mx/~ccoello/papers.html) yCoelloCoello Key: ga99dCoelloCoello.

....Intelligence Review, 239] Av. Ing. Quim. 92] Bad. Oper. Decyzje (Poland) 166] Biophysical Journal, 238] Chromatographia, 132] Civil Engineering Systems, 96, 105] Computers in Industry, 109] Egypt. Comput. J. Egypt) 11] Electronics Letters, 197, 20] Engineering Optimization, [106, 108] Engineering with Computers, 94] European Journal of Operational Research, 237] EvoNews, 194] Expert Systems with Applications, 99] Finite Elements in Analysis and Design, 264] Fuzzy Systems Artificial Intelligence Reports and Letters, 144] IEEE Transaction on Power Systems, 240] ....

....267] Cesteros, A. M. F. P. 183] Chacon, P. 238] Chang, O. 255, 256, 260, 262] Chaves, R. O. 39, 55] Chavez, Margarita G. 268] Cheim, L. 47] Chowdhury, M. M. M. 18] Christiansen, Alan D. 73, 80, 85, 94, 96, 97, 105, 106] Cluitmans, L. J. M. 16] Coello Coello, Carlos A. [71, 73, 74, 75, 76, 77, 78, 80, 81, 82, 83, 84, 85, 87, 94, 96, 97, 99, 102, 103, 104, 105, 106, 107, 108, 109] Colin, A. 92] Colmenares, A. 263] Comellas, F. 245] Conejo, A. J. 184] Conejo, A. 195] Cord on, Oscar, 150, 164, 177, 185, 198, 204, 205, 206, 211, 229, 235] Cornejo Rodriguez, A. 79, 93] Cortez, P. 121] Costa, Ernesto, 130] Costa, J. P. 37] Cotta, Carlos, 186, 209] Cotta ....

[Article contains additional citation context not shown here]

Carlos A. Coello Coello. Treating constraints as objectives for single-objective evolutionary optimization. Engineering Optimization, 32, 1999. (Accepted for publication; available via www URL: http://www.lania.mx/~ccoello/papers.html) yCoelloCoello ga99dCoelloCoello. 44 Genetic algorithms in the Latin America, Portugal and Spain

....of implemented MOEAs use only two fitness functions, most probably for ease and understanding. Several use three to nine, and the currently known maximum is 23 fitness functions within a single MOEA. This approach used an MOEA to solve a heavily constrained single objective optimization problem (Coello, 2000). Here, one objective was the fitness function and the other 22 were constraints cast as objectives. The highest number of conceptually different implemented fitness functions is found in a linkage design problem (Sandgren, 1994) where nine objectives are used. Howmany fitness functions are ....

Coello, C. A. C. (2000). Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275--308.

No context found.

C. A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275--308, 2000.

No context found.

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275--308, 2000.

....concepts to exploit the search space. Instead, it uses VEGA just to reach the feasible region. The use of special operators that preserve feasibility makes this approach highly specific to one application domain rather than providing a general methodology to handle constraints. Coello [7] used a population based approach similar to VEGA [23] to handle constraints in single objective optimization problems. At each generation, the population was split into 8 96 subpopulations of equal fixed size, where is the number of constraints of the problem. The additional subpopulation ....

....tries to reach the feasible region corresponding to one individual constraint. By combining these different subpopulations, the approach will reach the feasible region of the problem considering all of its constraints simultaneously. This approach was tested with some engineering problems [7] in which it produced competitive results. It has also been successfully used to solve combinational circuit design problems [8] The main drawback of this approach is that the number of subpopulations required increases linearly with the number of constraints of the problem. This has some obvious ....

[Article contains additional citation context not shown here]

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275--308, 2000.

....optimization concepts to exploit the search space. Instead, it uses VEGA just to reach the feasible region. The use of special operators that preserve feasibility make this approach highly speci c to one application domain rather than providing a general methodology to handle constraints. Coello [12] used a population based approach similar to VEGA [40] to handle constraints in single objective optimization problems. At each generation, the population was split into m 1 subpopulations of equal xed size, where m is the number of constraints of the problem. The additional subpopulation ....

....(including its own, of course) In this way, individuals who satisfy constraints are combined with individuals with a good tness value. At the end, it is expected to have a population of feasible individuals with high tness values. This approach was tested with some engineering problems [12] in which it produced competitive results. It has also been successfully used to solve combinational circuit design problems [13] The main drawback of this approach is that the number of subpopulations required increases linearly with the number of constraints of the problem. This has some ....

[Article contains additional citation context not shown here]

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275-308, 2000.

....is disjoint) we might land in an inappropriate part of the feasible region from which we will not be able to escape. However, this approach (as in the case of Parmee and Purchase s [12] technique) may be a good alternative to find a feasible point in a heavily constrained search space. Coello [4] proposed the use of a population based multiobjective optimization technique such as VEGA [13] to handle each of the constraints of a single objective optimization problem as an objective. At each generation, the population is split into 1 m sub populations ( m is the number of constraints) ....

....guided by the objective function, the evaluation of such objective function for a given vector x is used directly as the fitness function (multiplied by ( 1) if it is a minimization problem) with no penalties of any sort. For all the other sub populations, the algorithm used was the following [4]: if ( 0 x g then fitness = x g else if 0 v then fitness = v else fitness = f where ( x g refers to the constraint corresponding to sub population 1 j (this is assuming that the first sub population is assigned to the objective function f ) and v refers to the number ....

[Article contains additional citation context not shown here]

Coello C. (2000) Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275---308.

....The individual with the lowest maximum violation wins. Notice the great similarity between this approach and the technique proposed by Deb [44] that was described in section 5.2. The main di erence is that in this case, no extra mechanism is used to preserve diversity in the population. Coello [24] proposed the use of a population based multiobjective optimization technique such as VEGA [151] to handle each of the constraints of a single objective optimization problem as an objective. At each generation, the population is split into m 1 sub populations (m is the number of constraints) so ....

....guided by the objective function, the evaluation of such objective function for a given vector x is used directly as the tness function (multiplied by ( 1) if it is a minimization problem) with no penalties of any sort. For all the other sub populations, the algorithm used is the following [24]: if g j ( x) 0:0 then tness = g j ( x) else if v 6= 0 then tness = v else tness = f( x) where g j ( x) refers to the constraint corresponding to sub population j 1 (this is assuming that the rst subpopulation is assigned to the objective function f( x) and v refers to the number of ....

[Article contains additional citation context not shown here]

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275-308, 2000.

....these goals are the optimum of each objective function, considered separately) Furthermore, these techniques will yield a nondominated solution only if the goals are chosen in the feasible domain, and such condition may certainly limit their applicability. 4.6. 2 Some Applications Truss design [56, 7]. Design of a robot arm [10] Synthesis of low power operational ampli ers [72] 4.7 Recent approaches Recently, several new EMOO approaches have been developed. We consider important to discuss brie y at least two of them: PAES and SPEA. The Pareto Archived Evolution Strategy (PAES) was ....

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275-308, 2000.

....algorithms. First, we aim to optimize circuits (using a certain metric) in a di erent way, and intuitively, we can think of producing novel designs (since there is no human intervention) Such novel designs have been shown in the past (Miller et al. 2000, Miller, Kalganova, Lipnitskaya Job 1999, Coello et al. 2000). Second, it would be extremely useful to extract design patterns from such evolutionary generated solutions. This could lead to a practical design process in which a small (optimal) circuit is used as a building block to produce complex circuits. Such a divide and conquer approach has also been ....

.... we have approached this problem using a GA with a matrix encoding scheme, and an n cardinality alphabet (after a series of experiments, we found this n cardinality representation scheme to be more robust than the traditional binary representation (Coello 1996, Coello, Christiansen Aguirre 1997, Coello et al. 2000)) Our original GA based approach presents great resemblance with the one proposed by Miller (1997) and further developed by Miller and his colleagues (2000, 1999, 1998) The two main di erences between the two approaches are the encoding scheme and the tness function as we will explain later in ....

[Article contains additional citation context not shown here]

Coello, C. A. C. (2000), `Treating Constraints as Objectives for Single-Objective Evolutionary Optimization', Engineering Optimization 32(3), 275-308.

....is disjoint) we might land in an inappropriate part of the feasible region from which we will not be able to escape. However, this approach (as in the case of Parmee and Purchase s [37] technique) may be a good alternative to find a feasible point in a heavily constrained search space. Coello [10] proposed the use of a population based multiobjective optimization technique such as VEGA [43] to handle each of the constraints of a singleobjective optimization problem as an objective. At each generation, the population is split into m 1 sub populations (m is the number of constraints) so ....

....guided by the objective function, the evaluation of such objective function for a given vector x is used directly as the fitness function (multiplied by ( 1) if it is a minimization problem) with no penalties of any sort. For all the other sub populations, the algorithm used was the following [10]: if g j ( x) 0:0 then fitness = g j ( x) else if v 6= 0 then fitness = Gammav else fitness = f where g j ( x) refers to the constraint corresponding to sub population j 1 (this is assuming that the first sub population is assigned to the objective function f ) and v refers to the number ....

[Article contains additional citation context not shown here]

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275--308, 2000.

....a multiobjective fitness function) to design combinational circuits. There is some (relatively scarce) previous work on using multiobjective techniques to handle constraints. This work, however, has concentrated on numerical optimization only. Our approach, originally introduced in a recent paper [6], was probably the first attempt to use this kind of technique in the design of circuits (we presented one example of the design of a circuit in [6] Our proposal is to handle each of the matches between a solution generated by a GA and the values specified by the truth table as equality ....

....to handle constraints. This work, however, has concentrated on numerical optimization only. Our approach, originally introduced in a recent paper [6] was probably the first attempt to use this kind of technique in the design of circuits (we presented one example of the design of a circuit in [6]) Our proposal is to handle each of the matches between a solution generated by a GA and the values specified by the truth table as equality constraints. This, however, introduces some dimensionality problems for conventional multiobjective optimization techniques (this is because checking for ....

[Article contains additional citation context not shown here]

C. A. C. Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32(3):275--308, 2000.

....lowest maximum violation wins. Notice the great similarity between this approach and the technique proposed by Deb [31] that was described in a previous section. Jim enez and Verdegay [67] used a real coded GA with uniform crossover [133] nonuniform mutation [83] and tournament selection. Coello [17] proposed the use of a population based multiobjective optimization technique such as VEGA [119] to handle each of the constraints of a single objective optimization problem as an objective. The technique may be better illustrated by Figure 2. At each generation, the population is split into m 1 ....

....by the objective function, the evaluation of such objective function for a given vector X is used directly as the fitness function (probably multiplied by ( 1) if it is a minimization problem) with no penalties of any sort. For all the other sub populations, the algorithm used was the following [17]: if OE j (X) 0:0 then fitness = OE j (X) else if v 6= 0 then fitness = Gammav else fitness = f f(x) g (x) g (x) g (x) 1 f(x) g (x) g (x) 1 2 2 Sub populations Old Sub populations New m 1 3 1 2 1 2 3 m 1 genetic operators Apply g (x) m m Figure 2: Graphical representation of the ....

[Article contains additional citation context not shown here]

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32, 1999. (Accepted for publication).

....single objective optimization of f as a multiobjective optimization problem in which we will have m 1 objectives, where m is the number of constraints. Then, we can apply any multiobjec Please send all correspondence to: PO Box 60326 394, Houston, Texas 77205. tive optimization technique [4, 5] to the new vector v = f; f 1 ; fm ) where f 1 ; fm are the original constraints of the problem. An ideal solution X would thus have f i (X) 0 for 1 i m and f(X) f (Y) for all feasible Y (assuming minimization) The use of an evolutionary multiobjective optimization ....

....concentrate in optimizing the objective function, then we would be sampling points in the feasible space at random and it would be later very difficult to approach the region where the optimum resides. We have recently proposed a population based approach similar to VEGA to handle constraints [4]. This technique does not use dominance to impose an order on the constraints based on their violation (like in the case of COMOGA [11] which is a more expensive process (in terms of CPU time) that also requires additional parameters. The proposed approach does not rank individuals, but it uses ....

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32, 1999. (Accepted for publication).

.... multiobjective optimization techniques to real world problems will increase over the years, and a probable trend in research could be to reformulate many problems that are currently considered as if they only had one objective (e.g. constraint handling in single objective optimization [66]) This will constitute a more realistic approach to the solution of problems that frequently arise in areas such as engineering, because they are normally reduced to a single objective and the remaining objectives are treated as constraints instead of handling all (conflicting) objectives ....

Carlos A. Coello Coello. Treating Constraints as Objectives for Single-Objective Evolutionary Optimization. Engineering Optimization, 32, 1999. (Accepted for publication).

....to avoid stagnation. The new approach has been tested with several functions of different degrees of difficulty, and has been able to deal properly with both inequality and equality constraints. However, due to space limitations, only 2 examples will be shown. The interested reader may consult [6] for further information. 3. Examples 3.1. Example 1 : Design of a hydrostatic thrust bearing In this problem we want to minimize the power loss during the operation of a hydrostatic thrust bearing (see Figure 2) which has to withstand a specified load while providing an axial support. Four design ....

Coello, Carlos A. (1999). Treating constraints as objectives for single-objective optimization. Computers in Industry, (accepted for publication).

No context found.

C.A.C. Coello, Treating Constraints as Objectives for Single-Objective Evolutionary Optimization, Engineering Optimization, vol. 32 (3), pp. 275-308 (2000).

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC