IEEE WCCI/CEC2020 Competition

IEEE WCCI/CEC2020 Competition on Constrained Multiobjective Optimization

Organized by:

Yong Wang1, Zhi-Zhong Liu2, Zhongwei Ma1, and Bing-Chuan Wang1

1 School of Automation, Central South University, Changsha 410083, China (; mzw;

2 Department of Computer Science and Engineering, Southern University of Science and
Technology, Shenzhen 518055, China (

Scope and Topics

Constrained multiobjective optimization problems (CMOPs) are frequently encountered in diverse science and engineering. CMOPs contain both conflicting objective functions and diverse constraints. Therefore, to properly address a CMOP, not only the tradeoff among objective functions but also the balance between objective functions and constraints should be carefully considered. Undoubtedly, this is not an easy task for current evolutionary algorithms (EAs).

Fortunately, many researchers in the community of evolutionary computation have been devoting to coping with CMOPs; thus, many constrained multiobjective EAs (CMOEAs) have already been proposed during the past two decades. In practice, these CMOEAs are roughly grouped into three classes, i.e., dominance-based CMOEAs, decomposition-based CMOEAs, and indicator-based CMOEAs, in which constraint-handling techniques are combined with dominance-based, decomposition-based, and indicator-based MOEAs, respectively. To assess the performance of these CMOEAs, various artificial CMOPs have been designed including CTP, C-DTLZ, DAS-CMOPsnd NCTPs. These artificial CMOPs are more suitable than real-world ones as benchmark functions. It is because the latter may require special hardware or software in the simulation processes. Anyway, these artificial CMOPs can help researchers to analyze and understand the performance of different CMOEAs and encourage users to select the desired ones. At present, most of them have been successfully solved by current peer CMOEAs.

To further boost the development of CMOEAs, two novel CMOP test suites (i.e., DOC and MW) were proposed in 2019 and both of them draw the inspirations from the CMOPs in the real-life applications.

  • DOC considers the fact that, in the real-world CMOPs, both decision and objective constraints are involved, which are easy to understand in decision space and objective space, respectively. To simulate the real-world scenes better, DOC provides a quite convenient approach to construct a CMOP with both decision and objective constraints. Note that these decision constraints can make the feasible region in the decision space have different properties (e.g., nonlinear, extremely small, and multimodal), while these objective constraints can reduce the feasible region in the objective space and make the Pareto front (PF) have diverse characteristics (e.g., continuous, discrete, mixed, and degenerate). As a result, DOC poses a great challenge to the current CMOEAs to obtain a set of well-distributed and well-converged feasible solutions.

  • Due to the presence of constraints, some or all the original Pareto optimal solutions of the unconstrained MOPs may become infeasible, and some solutions on the boundary of the feasible region of a CMOP may become the Pareto optimal solutions. With that in mind, MW designs four different types of CMOPs: 1) Type I: the constrained PF is the same with the unconstrained PF; 2) Type II: the constrained PF is a part of the unconstrained PF; 3) Type III: the constrained PF consists of a part of the unconstrained PF and a part of the boundary of the feasible region; and 4) Type IV: the unconstrained PF is all located outside the feasible region. All these four different types of CMOPs are constructed by making use of a new construction method, in which a global control process and a local adjustment process are included. The CMOPs in MW have complex geometry of constrained Pareto front and controllable size of feasible region in the objective space, which can also set up a lot of obstacles for a CMOEA to obtain promising results.

For convenience, we do not design any new CMOPs, instead, we select several CMOPs in the previous two test suites as the benchmark test functions for this competition. Specifically, CMOPs 1-6 come from MW5, MW6, MW10, MW13, MW9, and MW11, and CMOPs 7-10 come from DOC-1, DOC-4, DOC-5, and DOC-8. It is believed that the blending of these two test suites has the capability to assess an algorithm's  capability for solving CMOPs in the real-life applications.


  • Z.-Z. Liu and Y. Wang. Handling constrained multiobjective optimization problems with constraints in both the decision and objective spaces. IEEE Transactions on Evolutionary Computation, vol. 23, no. 5, pp. 870-884, 2019.

  • Z. Ma and Y. Wang. Evolutionary constrained multiobjective optimization: Test suite construction
    and performance comparisons. IEEE Transactions on Evolutionary Computation, vol. 23, no. 6, pp. 972-986, 2019.

CEC 2020



Recently Visited

Similar Subject

Same institution

Similar Interests