Research And Design Of Controllers Based On Optimization Algorithm Of Membrane Computing

Membrane computing (P Systems) is aiming at abstract computing models from thefunctioning and structures of living cells, as well as from the way the cells are organized intissues or higher order structures. The obtained models which are called membrane systems orP systems are distributed and parallel computing models. The obtained models, calledmembrane systems or P systems, are distributed and parallel computing models. Membranecomputing is not simple simulation at functions of biomembranes but a computation moduleconsisting of fundamental characteristics which are abstracted from the functioning andstructures of varietal biomembranes. These models are composed of three key ingredients:membrane hierarchical structures, multisets of symbol-objects and evolution rules.The membrane computing optimization algorithm is a sort of optimization programwhich simulates the internal evolutionary mechanism of biological cells. The programpossesses complex construction of multi-hierarchy and region-compartment with features offlexible membrane structure and powerful compatibility of which it can integrate merit ofother algorithms to solve practical problems. In this paper, based on presented research resultson membrane computing and algorithms, the particle swarm optimization based on P systems(PSOPS) algorithm and its application in optimization of PID controller parameters areinvestigated in detail. Besides, one improved algorithm called membrane computingoptimization algorithm based on mutated PSO (MPSO-MC) is proposed to apply in thecontrolling of non-linear systems.The main researching contents and results are shown as follows:(1) The PSOPS algorithm is employed to optimize parameters of PID controllers basedon four typical SISO objects with different performance indexes due to its powerful searchingability. The experimental results demonstrate that PSOPS performs better than traditionalmethods with the best control performance.(2) The particles of PSOPS in each membrane may show great homoplasy in thesearching process and then lead the algorithm into premature. To solve this problem, one newalgorithm called membrane computing optimization algorithm based on mutated PSO (inshort, MPSO-MC) is proposed to make fully use of the parallelism feature and uncertainty ofmembrane computing. The algorithm is an appropriate combination of a horizontal hybridmutation operator, membrane computing approaches (OLMC, communication rules) andevolution rules of PSO algorithms. To evaluate the proposed algorithm in terms of the searchcapability, the performance of MPSO-MC is compared with PSO and PSOPS using ten testfunctions with characteristics of large searching space, numerous local minimum points andstrong fraudulence. The types of these functions include unimodal (containing only one optimum) and multimodal (containing many local optima, but only one global optimum),continuous and discontinuous, or nimization and maximization.(3) Furthermore, we develop the MIMO PID neural network (MPIDNN) controller basedon PIDNN and employ the proposed algorithm MPSO-MC to take the place of BP algorithmto control the strong coupled nonlinear MIMO system. Comparison of simulation results ofthe MPIDNN controller based on BP, PSO, CPSO and MPSO-MC algorithms indicate thatMPSO-MC-based MPIDNN controller is more effective than the other three in controllingMIMO systems.