Hardware Implementation Methodologies Of Computing Models Inspired By Biological Cells

Cell-inspired computing models are computing models enlightened by the structures,functioning and bio-chemical reactions of living cells.Parallelism and non-determinism are two characteristics of cell-inspired computing models.The research significance and value consist in:(1)Hardware implementation is the important method to validate the theoretical par-allelism and non-determinism of these models;(2)Hardware implementation bridge the gap between theories and applications of these models;(3)Hardware implemented cell-inspired computing models emulate their computation processes with digital circuits essentially.These designed circuits can lay the foundation for the manufacturing of application specific integrated circuits(ASICs)of these models;(4)Hardware architectures of cell-inspired computing mod-els are parallel heterogeneous.Hardware implementation of these architectures has consider-able referential value for the research and development of novel processors/controllers.Nevertheless,hardware implementation of cell-inspired computing models is a research work with high challenges.Its core businesses are the characterization of syntactical structures and imitation of computation processes of these models.The implementation of parallelism and non-determinism of these models are the major challenges.The research objects of this thesis are two kinds of cell-inspired computing models—reaction systems and cell-like P systems.Difficulties of reaction system hardware implementation lie in:(1)The representa-tion of qualitative bio-chemical reactions and chemical entities of reaction systems;(2)The imitation of interactive process of reaction systems,including the simultaneous input of chem-ical entities and parallel execution of reactions.Difficulties of cell-like membrane system hardware implementation consist in:(1)The representation of objects/variants and evolution rules/programs;(2)The representation of membrane structures;(3)The imitation of config-uration transition,including the determination of applicable rules/programs and their parallel and non-deterministic executions.The main innovations of this thesis are:1.The FPGA implementation method of reaction systems is put forward.Innovations of this method lie in:(1)Qualitative reactions are transformed to Boolean expressions;(2)Ap-plicable and non-applicable reactions are executed simultaneously to avoid the determination of usability of reactions in each interactive process step;(3)Chemical entities in the context sequence are treated as intermediate variables to avoid communication process.Reaction sys-tems implemented by this method have the computing speed of 2×10~8step/second.2.The generalized numerical P system(GNPS)is proposed.Advantages of GNPS con-sist in:(1)Presburger arithmetic is introduced to present usability predicates so their forms and scope are expanded;(2)The position restrictions of variables of(enzymatic)numerical P systems((E)NPS)are eliminated and allow variables coming from any membranes,so that all the programs updating the same variable can be combined into one program.As results,mem-brane structures are simplified and GNPS have better flexibility and adaptability.(3)GNPS has specialized input-output variables to facilitate data input of data acquisition facilities and results output.GNPS extends(E)NPS and can be implemented more easy.3.The fixed/floating point variable numerical P sytems(NPS)FPGA implementation method are presented.novelties of these methods are:(1)Membranes are represented by Ver-ilog basic functional blocks,the modules;(2)For fixed-point variables,1)program usability predicates are transformed to logic expressions;2)multiple distributed values to the same vari-able are summed and the summation is assigned this variable to refrain from conflicts due to multiple programs updating the same value;(3)For floating-point variables,1)sequentially triggered floating-point arithmetic units are designed to perform associated operations;2)edge detection technique is used to generate trigger signals which are equivalent to program usabil-ity predicates for floating-point arithmetic units.Numerical P systems implemented by these two methods achieve the speedup of 10~4magnitude order comparing with their software sim-ulation speed.4.The coefficient non-deterministic selection(CNS)method for implementing the non determinism of symbol-object P systems(SPS)is devised.New ideas of CNS method include:(1)Non-repetitive Cartesian products are built from possible multiplicities of variables in the left hand side of rules;(2)The system of restriction inequalities are constructed according to maximally parallel derivation mode and object consumption.Non-deterministically choose a Cartesian product and substitute it into those combined systems of restriction inequalities con-structed before.The Solution of combined systems of restriction inequalities is the applicable rule multiset which implements the non-determinism of SPS.Contrast with existing methods,the strong point of CNS is that it can find all the applicable rule multisets.5.The symbol-object P systems FPGA implementation method is proposed based on NPS FPGA implementation method.Innovations of this method are:(1)rules of SPS are converted to programs of NPS;(2)Design programs to execute the algorithm implementing non-determinism of PSP;(3)Assemble these programs to build a GNPS.Implement this GNPS with fixed-point variable NPS FPGA implementation method so that PSP can be implemented indirectly.Comparing with software simulation,SPS implemented by this method achieves the speedup of 10~4magnitude order.On the basis of the above achievements,the future researches include:(1)using FPGA membrane controller to control target robot directly;(2)Study and implement simultaneous location and mapping(SLAM)based on membrane computing to provide environmental map for robot path planning;(3)Investigate the FPGA implementation method of tissue/population P systems and spiking neural P systems;(4)Explore the implementation method of biological cell computing models by combining software and hardware.Research works of this thesis provide efficient FPGA implementation methods for re-action systems and the two cell-like P systems.By these methods,the parallelism and non-determinism of these models are obtained.These methods are applied in biology/ecology mod-elling,robot motion control,path planning and image processing efficiently,obtaining con-siderable speedup comparing with software simulations.