Research On Bio-inspired Optimization Algorithms Based On Membrane Computing And Applications

Complex optimization problems exist widely in the fields of science and engineering. Biologically inspired computing is a class of available methods to solve these problems. Membrane computing (also known as P systems) is presented originally as a distributed and parallel computational model, which is abstracted from living cell structure, functioning and interrelationships in tissues or organs. This dissertation focuses on novel computational frameworks, rules and search strategies distinguished from other optimization algorithms, which are based on distributed and parallel computational model, ideas and rules of membrane computing, inspired by eukaryotic cells and combined the fruits of evolutionary computing and mathematical programming optimization algorithms, and using them in the proposed algorithms to solve difficult nonlinear optimization problems. Typical test functions and industrial case studies are employed to validate and analyse these algorithms. The contributions and innovations of this dissertation are as follows:(1) Aiming at solving unconstrained optimization problems, a bio-inspired algorithm based on membrane computing (BIAMC) is proposed. The netted membrane structure used in BIAMC is based on a commonly computational model of membrane computing and inspired by the shape, structure and function of the Golgi apparatus of eukaryotic cells. In parallel identical membranes, there are rules of rewriting, pairing, selection and communication. These rules in BIAMC explore optimum solutions in the global search space randomly. The rules of transition, abstraction, selection and communication in the membrane of quasi-Golgi synthesize new objects in the neighborhood of current best objects. Then the exploration and exploitation of searching for a global optimum solution in BIAMC are balanced. The nondeterministic versions of membrane computing and the function of Golgi appareturs are implemented as the computational strategy and the search strategy of BIAMC, which are different from the computational strategy and uniformly distributed approaching strategy used in the evolutionary computation. Fifteen typical unconstrained functions are used for validating the proposed algorithm and analyzing the sensitivity of algorithmic parameters. The results and comparison show that the proposed approach is available and efficient.(2) An improved bio-inspired algorithm based on membrane computing is proposed to solve both unconstrained and constrained optimization problems. Compared with BIAMC, the membrane structure is similar to that used in BIAMC. The rewriting rule in the parallel identical membranes is improved, the communication rule is different from that of BIAMC, and a new rule of target indication is embeded in the membrane of quasi-Golgi, which enhances the computational efficiency of approaching to optimum solutions. A quadratic penalty method is employed to solve the constrained problems. Fourteen various unconstrained and constrained functions are used for performance testing. Then the proposed algorithm is applied to solve a nonlinear optimization problem of gasoline blending and scheduling in chemical process. The results and comparison with others show that the proposed approach can find optimal or close-to-optimal solutions efficiently.(3) For the purpose of solving complex constrained optimization problems, the three hybrid optimization algorithms are proposed. In these algorithms, a membrane of S is embedded into their netted membrane structure, which has a rule of sequencial quadratic programming (SQP) algorithm. The first hybrid optimization algorithm uses a serial aggregating mode. The second one uses an inbuilt aggregating mode. The third hybrid optimization algorithm with a dynamic membrane structure is designed for solving optimal robot path planning problems, which uses a serial aggregating mode. The superiority of three proposed algorithms benefits from the distributed computational structure to other hybrid optimization methods. Benchmark constrained problems and simulation examples in chemical process are utilized to test the proposed hybrid algorithms. The results and comparison show that the three hybrid optimization algorithms based on membrane computing are available and valid.(4) A multi-objective bio-inspired algorithm based on membrane computing (MOBIAMC) is proposed. The membrane structure used in MOBIAMC is a network of membranes, which is similar to those used in BIAMC. A novel computational strategy of multi-objective optimization algorithm is applied to obtain the high-performance with a cheaper computational cost and more efficient computational capability. Benchmark multi-objective test problems are applied to validate the proposed algorithm. The test results show that the non-dominated solution set solved by MOBIAMC reaches or is close to the Pareto front, and the comparison with other algorithms reveals its superior computational performance.