Logical Function And Evolutionary Circuit Design Based On Cellular Neural Network

In recent years, building dynamic logic gates based on cellular neural network is a new research direction. As the nonlinear system has strong nonlinear characteristics and rich dynamic model, it has an advantage in the construction of flexible, reconfigurable logic gate. Functions of these logical gates can alter according to changes of external control signals. Thus, this design brings new possibilities to more flexible universal computing device and even dynamical computing architecture.In this paper, firstly, the design of logical functions based on cellular neural networks is proposed. The template parameters of standard uncoupled cellular neural network with two inputs and three inputs linearly separable Boolean function are analyzed, and the corresponding circuit is designed. At the same time, for the cell design of linearly separable Boolean function, the dynamic transformation of logical gates can be realized by changing the parameters when the circuit structure is invariable. For linearly non-separable Boolean function, through two types of improved cellular neural networks which are canonical universal and multi-nested universal cells, the corresponding template parameters are designed. And a full adder design is realized by using the canonical universal cellular neural network. Simulation results show that the logical circuit based on the canonical universal cellular neural network has certain advantages. Then illustrated by the case of Parity(4) function, the resource utilization of multi-nested cells in circuit design is superior to that of canonical universal cells.Secondly, evolutionary circuit design based on CNN is studied.The genetic algorithm is used to evolve the circuit. In view of the circuit structure characteristics, the encoding way of the circuit individual is the matrix form; By adding three inputs LSBF as logic genes, the diversity of genes and circuit evolution can be enriched, and the effectiveness of the circuit can be improved.Improved genetic algorithm, through the improvement of the selection operator, crossover operator and mutation operator, can improve the rate of evolution and the overall performance of the circuit. On the basis of this, a full adder and two bit multiplier evolution circuit are designed.To sum up, we studied how to use standard CNN and its modified cells to realize Boolean function, and gave the method and circuit implementations. The proposed method is of clear physical meanings, simple circuit and easy to understand, which is the foundation of CNN to realize the VLSI circuit, and has some engineering significance to the design of the reconfigurable chips.