A Hybrid GA-SQP Algorithm for Analog Circuits Sizing

Abstract

This study presents a hybrid algorithm obtained by combining a genetic algorithm (GA) with successive quadratic sequential programming (SQP), namely GA-SQP. GA is the main optimizer, whereas SQP is used to refine the results of GA, further improving the solution quality. The problem formulation is done in the framework named RUNE (fRamework for aUtomated aNalog dEsign), which targets solving nonlinear mono-objective and multi-objective optimization problems for analog circuits design. Two circuits are presented: a transimpedance amplifier (TIA) and an optical driver (Driver), which are both part of an Optical Network-on-Chip (ONoC). Furthermore, convergence characteristics and robustness of the proposed method have been explored through comparison with results obtained with SQP algorithm. The outcome is very encouraging and suggests that the hybrid proposed method is very efficient in solving analog design problems

Combined economic and emission dispatch considering conventional and wind power generating units

“…Then, an approximation is made of the Hessian matrix of the Lagrangian function by using the quasi‐Newton method at each iteration. Therefore, subproblems of the quadratic programming (QP) are generated to form the original search direction to a line search procedure . Theoretically, the resolution of the constrained smooth nonlinear optimization problem is very accurate through SQP, especially when the Karush‐Kuhn‐Tucker (KKT) conditions are applied …”

“…The GA algorithm is good for the global search. However, it needs a long simulation time and may not be very accurate in the local search . Moreover, from previous research, the SQP is a very accurate technique, but it is very sensitive to its initial points.…”

“…Therefore, subproblems of the quadratic programming (QP) are generated to form the original search direction to a line search procedure. [44][45][46] Theoretically, the resolution of the constrained smooth nonlinear optimization problem is very accurate through SQP, especially when the Karush-Kuhn-Tucker (KKT) conditions are applied. 38,[47][48][49][50][51]…”

“…However, it needs a long simulation time and may not be very accurate in the local search. 45 Moreover, from previous research, 38,[47][48][49][50][51][52] the SQP is a very accurate technique, but it is very sensitive to its initial points. A hybrid GA-SQP algorithm can reduce the computational time and ensure the accuracy, and it is applied in the present paper.…”

“…A hybrid GA-SQP algorithm can reduce the computational time and ensure the accuracy, and it is applied in the present paper. 38,45,52 Firstly, using GA as a first stage global optimizer, in order to obtain some decent starting points, by exploiting GA's global search ability. Secondly, using the obtained solution as found by GA as a starting point to the second stage local searching method SQP in order to refine the first stage result.…”