Background and Objectives: Today, the use of methods derived from Reinforcement learning-based approaches, due to their powerful in learning and extracting optimal/desirable solutions to various problems, shows a significant wideness and success. This paper presents the application of reinforcement learning in automatic analog integrated circuit design.
Methods: In this work, the multi-objective approach by learning automata is evaluated for accommodating required functionalities and performance specifications considering optimal minimizing the MOSFETs area and power consumption for two famous CMOS op-amps.
Results: The performance of the circuits is evaluated through HSPICE and the approach is implemented in MATLAB, so a combination of MATLAB and HSPICE is performed. The two-stage and single-ended folded-cascode op-amps are designed in 0.25μm and 0.18μm CMOS technologies, respectively.
According to the simulation results, a power of 560.42 and an area of 72.825 are obtained for a two-stage CMOS op-amp, and also a power of 214.15 and an area of 13.76 are obtained for a single-ended folded-cascode op-amp. In addition, in terms of total optimality index, MOLA for both cases has the best performance between the applied methods, and other research works with values of -25.683 and -34.162 dB, respectively.
Conclusion: The results shown the ability of the proposed method to optimize aforementioned objectives, compared with three multi-objective well-known algorithms.