Study On Multi-objective Optimization Theory And Application Based On Biogeography Algorithm

Biogeography-based optimization algorithm (BBO) is a new random search algorithmbased on biogeography model. It is proved that the algorithm has good ability in exploitationof the population information and shows excellent performance for the single objectiveoptimization problem and real application. In this paper, BBO is extended to multi-objectiveoptimization field for solving multi-objective optimization problems (MOPs). Firstly,multi-objective optimization algorithm based on BBO is proposed and is studied its feasibilityand effectiveness. And then a new memetic algorithm combining BBO with differentialevolution (DE) is put forward to improve the performance of the optimization algorithm. Inpractical applications, some multi-objective optimization problems often are subject to certainconstraints. In view of this, some constraint handling methods are involved to propose aconstrained multi-objective optimization algorithm based on BBO. Finally, the proposedconstrained multi-objective optimization algorithm is applied for robot path planning.Simulation results show the performance of algorithm in practical application.In this paper, the detailed key problems are summarized as follows:At first, the original BBO is combined with multi-objective evolutionary algorithmframework to solve MOPs, which will lead to some problems including loss of the populationdiversity and falling into the local Pareto front. To overcome some shortcomings, newdisturbance migration operator is proposed based on original migration operator. Thedisturbance migration operator introduces a random disturbance factor which makes thechanged characters of excellent individuals shared by the other individuals. And then thedisturbance migration operator and mutation operator are used to generate the offspringpopulation. An archive is applied to conserve the non-dominated solutions, and thencrowding-distance is used to update the archive. By testing on benchmark functions andcomparing with classic algorithms, simulation results show that the proposed algorithm isfeasible and effective for MOPs.To improve the performance of optimization algorithm, a hybrid multi-objectivealgorithm is proposed, which combines the exploitation ability of BBO and the explorationability of DE. In the multi-objective algorithm, a hybrid disturbance migration operator isproposed by using a across parameter to combine disturbance operator based on the boundaryindividuals and mutation operator of DE. The hybrid operator takes into account the diversityand spread of the population, simultaneously. Then the hybrid operator and mutation operatorare applied to produce the offspring population. The archive is used to conserve the non-dominated solutions gained, and the crowding-distance is used to update the archive.Qualitative and quantitative experiment results demonstrate that the hybrid algorithm cansearch a wider, more uniform and better diversity Pareto front for various test functions.Constrained multi-objective optimization algorithm based on BBO is put forwards forsolving constrained multi-objective problems (CMOPs). Considering that the excellentinfeasible solutions can contribute to good performance, they make crossover with the nearestnon-dominated feasible solutions so that they can approximate the feasibility to improve thediversity of population. At the same time, during the process of species migration, someindividuals are often affected by the other individuals and random factors so that a newmigration operator is proposed based on the difference of the other individuals, then themigration operator is applied to the evolution of feasible solutions to produce many morenon-dominated feasible solutions. So the algorithm make the population approximate Paretofront from both inside and outside of feasible region, simutantiously. Benchmark test andcomparison experiments show that the proposed algorithm has better performance inconvergence and distribution.At last, in order to confirm the effectiveness of the designed algorithm in practicalapplication, the constrained multi-objective optimization algorithm based on BBO is appliedto solve robot path planning (RPP). The path length and smoothness are as optimizationsubjective, and the violation of obstacles is as a constraint, so RPP is a constrainedmulti-objective optimization problems. Then paths and their notes are taken as individualsand genes of optimization algorithm, constrained multi-objective biogeography–basedoptimization algorithm is applied to plan the path of robot. Compared with the classicalconstraint algorithms, experimental results show the superiority of the proposed algorithm forRPP.