Research On Multi-objective Optimization Algorithm Based On Membrane Computing Models

Membrane computing, as a branch of natural computing, aims to abstract computing models from the structure and its function, organization, organs and their collaborative relations of the living cells. Membrane computing model is also called membrane systems or P systems, which has the advantages of good parallelism, distribution and non-determinacy and other characteristics. Now, membrane computing has been widely used in many optimization fields. Therefore, membrane computing is an important researching field which has important theoretical and practical values.Multi-objective optimization problems exist widely in the field of engineering and science. Usually, such optimization problems are characterized by multiple objectives which conflict each other. Many nature-inspired methods, such as genetic algorithms, particle swarm optimization algorithms and membrane algorithms based on membrane computing model, are proposed to solve the problems. In these methods, because membrane computing models can provide a rich framework in solving optimization problems, P system-based optimization algorithms have been the hot topic so far. This paper, based on the membrane computing model, proposes two multi-objective optimization algorithms based on the membrane computing model, respectively:(1) Proposes a multi-objective membrane algorithm guided by skin membrane. Membrane computing as a branch of natural computing, has achieved good results in the single objective optimization. Nevertheless, its problem solving potentials in the multi-objective optimization remains to be explored. For the current multi-objective membrane optimization algorithms, many algorithms put skin membrane as an archive, where the optimal solutions during the search are stored. But, they all neglect to harness the information of skin membrane, where the optimal solutions are stored, to guide the evolution of inner membranes. Thus, in this paper we propose an effective skin membrane guiding strategy by considering both convergence and distribution respectively, which can accelerate the convergence of solutions. Experimental results on ZDT and DTLZ test suites show that SMG-MOMA performs better than popular multi-objective optimization algorithms and multi-objective optimization membrane algorithms.(2) Proposes a many-objective membrane algorithm guided by skin membrane. Due to membrane computing models can provide a rich framework in solving optimization problems, which also demonstrate its unique advantages in the multi-objective optimization problems. But when the objectives exceed three, those existing attempts are not effective. SMG-MOMA algorithm has proved the skin membrane guided strategy played a significant role in population evolution. But, when there are many objectives, it may lose its convergence advantages. Therefore, we propose a more effective convergence allocation strategy, i.e., setting two archive populations in the skin membrane:one population used to guarantee the convergence and the other to guarantee the distribution of solutions respectively in order to guide the evolution of internal membrane population. Experimental results on the scalable DTLZ and WFG test suites show that SMG-MaOMA can effectively handle the many-objective optimization problems.