The Study Of Analog Circuits Auto-design Based On Evolutionary Methods

Evolutionary circuit design is an important branch of the Evolvable Hardware: EHW, it has already become a research hotspot of EHW recent years. Evolutionary circuit design applies evolutionary techniques to design circuits without human intervention, which simulates the process of biological evolution. However, evolutionary circuit design is still in the initial stage, many issues need in-depth research. For example,how to improve the convergence rate of evolutionary process, the scale of the circuits we designed, and the applications, etc, all of them are challenging problems. The purpose of this dissertation is to study Analog Circuits Auto_Design Based on Evolutionary Methods. The research includes the co-evolution circuit evolution, the Circuit Tolerance Design based on Active Deviation and multi-objective optimization of analog circuit. The main research works of this dissertation can be summarized as follows:(1) In order to design circuits with more components and to improve the speed of convergence, this dissertation has studied the co-evolution theory. The co-evolution theory was combined with genetic programming, and COGP algorithm was proposed. On this basis, this dissertation has combined the COGP algorithm with current decomposition method, which is an evolutionary circuit design method for solving the scalability problem, and it can also improve the convergence rate of evolutionary. The experimental results demonstrate the validity of the proposed method.(2) In analog circuits, the value of discrete devices is susceptible to manufacturing tolerances, device aging and environmental factors, so the tolerance design of analog circuits is very necessary. In this dissertation a new circuit tolerance design method was proposed. Firstly, the evolutionary model(Active Deviation model) of the device value drift was given, which was used to mimic the phenomenon of device value drift. Then, this paper combined the Active Deviation and GP, and proposed the Active Deviation based Hybrid GP(AD-Based HGP). With AD-Based HGP the tolerance range of the target circuit can be set in accordance with the Active Deviation range, this is very flexible to change the tolerance range in the design process. The experimental results show that the method can effectively improve the tolerance range of the circuit we designed, and the tolerance range is adjustable, which also shows its great potential in analog circuit tolerance design .(3) In this paper, we implement a nondominated sorting-based multiobjective genetic programming, called Nondominated Sorting Genetic Programming(NSGP), which was used to multi-objective circuits design. In the design process,this method can take into account the circuit performance and circuit size. Experimental results demonstrate that, during an evolution process,NSGP can give out more than ten Pareto-optimal solutions,each Pareto-optimal solution represent a target circuit,it also showed the superior performance of NSGP in multi-objective circuits design.