模拟集成电路(IC）的设计自动化面临诸多困难。人工设计是通过推导设计方程进行设计推理，其中需要近似推理以获得设计知识。要把具有推理成分的设计过程自动化是目前模拟IC自动化课题的最大困难。项目组近期在符号化方法生成设计方程方向取得了积极进展，形成了一些解决瓶颈问题的初步方案；由此激发了我们在该方向上继续进行有价值探索的研究动机，开发能执行自动化近似推理的模拟IC新颖设计自动化工具，并在电路拓扑设计、器件尺寸定制、性能预测建模等相关问题上得到应用，以弥补传统以数值仿真为中心方法的不足。项目将着重处理：1）基于电路结构自动化约简方法解决电路的自动化拓扑综合问题；2）基于电路的自动化拓扑降阶方案和设计方程自动导出方法解决Sizing自动化；3）为上述两项问题构造良好的互动式界面，提高以人机互动近似推理方式获取设计知识的效率。系统探索计算机辅助近似推理机制对模拟IC设计自动化的本质作用。

Analog IC design automation faces many difficult problems. Manual design of analog IC highly relies on design equation based approximate reasoning. It is very hard to automate a design process involving such approximate reasoning. Our recent research effort has made good progress in automatic generation of design equations for analog circuits by applying symbolic computation techniques. Inspired by the new progress, in this proposal we are going to pursue research on automatic synthesis of analog ICs, including topology synthesis and device sizing. The research goal is to develop new mechanism for analog IC design automation based on automated reasoning to compensate the inadequacy in traditional methods. We intend to attack the following problems: (1) Developing easy-to-use schemes for circuit topology design automation by automatic topological circuit reduction and its inversion；(2) developing design equation based sizing methods based on topological circuit reduction, and (3) developing friendly user interface to enable interactive computer-aid approximate reasoning, etc. Based on a systematic research on these problems, we intend to explore the knowledge discovery mechanism with the help from computer-aided approximate reasoning, validate the feasibility of intelligent analog IC design synthesis, and establish a new paradigm for analog IC design automation.