Longmei Li

Building and Using an Ontology of Preference-Based Multiobjective Evolutionary Algorithms

Integrating user preferences in Evolutionary Multiobjective Optimization EMO is currently a prevalent research topic. There is a large variety of preference handling methods originated from Multicriteria decision making, MCDM and EMO methods, which have been combined in various ways. This paper proposes a Web Ontology Language OWL ontology to model and systematize the knowledge of preference-based multiobjective evolutionary algorithms PMOEAs. Detailed procedure is given on how to build and use the ontology with the help of Protege. Different use-cases, including training new learners, querying and reasoning are exemplified and show remarkable benefit for both EMO and MCDM communities.

Preference-based multiobjective optimization using truncated expected hypervolume improvement

The ultimate goal of multi-objective optimization is to provide potential solutions to a decision maker. Usually, what they are concerned with is a Pareto front in an interesting/preferred region, instead of the whole Pareto front. In this paper, a method for effectively approximating a preferred Pareto front, based on multiobjective efficient global optimization (EGO), is introduced. EGO uses Gaussian processes (or Kriging) to build a model of the objective function. Our variant of EGO uses truncated expected hypervolume improvement (TEHVI) as an infill criterion, which takes predictive mean, variance and preference region in the objective space into consideration. Compared to expected hypervolume improvement (EHVI), the probability density function in TEHVI follows a truncated normal distribution. This paper proposes a TEHVI method that makes it possible to set a region of interest on the Pareto front and focus search effectively on this preferred region. An expression for the exact and efficient computation of the TEHVI for truncation over a two dimensional range is derived, and benchmark results on standard bi-objective problems for small budget of evaluations are computed, confirming the effectiveness of the new approach.

Preference incorporation to solve multi-objective mission planning of agile earth observation satellites

This paper investigates earth observation scheduling of agile satellite constellation based on evolutionary multiobjective optimization (EMO). The mission planning of agile earth observation satellite (AEOS) is to select and specify the observation activities to acquire images on the earth surface. This should be done in accordance with operational constraints and in order to maximize certain objectives. In this paper three objectives are considered, i. e. profit, quality and timeliness. Preference-based EMO methods are introduced to generate solutions preferable for the decision maker, whose preference is expressed by a reference point. An improved algorithm based on R-NSGA-II, which is named CD-NSGA-II, is proposed and compared with other algorithms in various scheduling scenarios. Results show that the chosen preference modeling paradigm allows to focus search on the interesting part of the Pareto front. Moreover, the tested algorithms behave differently with the change of reference point and degree of conflicts, among them CD-NSGA-II has the best overall performance. Suggestions on applying these algorithms in practice are also given in this paper.

A new approach to target region based multiobjective evolutionary algorithms

In this paper, a target region based multiobjective evolutionary algorithm framework is proposed to incorporate preference into the optimization process. It aims at finding a more fine-grained resolution of a target region without exploring the whole set of Pareto optimal solutions. It can guide the search towards the regions on the Pareto Front which are of real interest to the decision maker. The algorithm framework has been combined with SMS-EMOA, R2-EMOA, NSGA-II to form three target region based multiobjective evolutionary algorithms: T-SMS-EMOA, T-R2-EMOA and T-NSGA-II. In these algorithms, three ranking criteria are applied to achieve a well-converged and well-distributed set of Pareto optimal solutions in the target region. The three criteria are: 1. Non-dominated sorting; 2. indicators (hypervolume or R2 indicator) or crowding distance in the new coordinate space (i.e. target region) after coordinate transformation; 3. the Chebyshev distance to the target region. Rectangular and spherical target regions have been tested on some benchmark problems, including continuous problems and discrete problems. Experimental results show that new algorithms can handle the preference information very well and find an adequate set of Pareto-optimal solutions in the preferred regions quickly. Moreover, the proposed algorithms have been enhanced to support multiple target regions and preference information based on a target point or multiple target points. Some results of enhanced algorithms are presented.

Preference-based evolutionary algorithms for many-objective mission planning of agile earth observation satellites

The mission planning of agile earth observation satellite (AEOS) involves multiple objectives to be optimized simultaneously. Total profit, observed target number, averaged image quality, satellite usage balance and averaged timeliness are the five objectives considered in this paper. The problem is a mixed-integer problem with constraints and belongs to the class of NP-hard problems. Two preference-based evolutionary algorithms, i.e., T-NSGA-III and T-MOEA/D, are proposed with problem-specific coding and decoding strategies to solve the problem. Target region, which is defined by preferred range of each objective, is used for preference articulation. Experiments show that the proposed algorithms can obtain Pareto optimal solutions within the target region efficiently. They outperform Pareto-based T-NSGA-II with regard to Hypervolume indicator and CPU runtime.