Home         Authors   Papers   Year of conference   Themes   Organizations        To MES conference

Placement of VLSI Elements Based on Swarm Intelligence Models
Authors
	Lebedev B.K.
	Lebedev O.B.
	Zhiglaty A.A.
Date of publication
	2020
DOI
	10.31114/2078-7707-2020-4-118-125
Abstract
	The paper presents the architecture of a multi-agent system for deploying VLSI elements based on the integration of swarm intelligence models. The concept of placing elements by the pair P (A1, A2) of ordered lists of the same set of elements is used. The transition from pair P to placement is carried out in two stages. First, a constraint graph is formed on the basis of the pair P, and then a plan or arrangement of elements is constructed from the constraint graph. New structures are proposed for representing the solution to the problem of placing elements of super-large integrated circuits in the form of chromosomes. A modified paradigm of a particle swarm is presented, which differs from the canonical one, with the possibility of using positions with integer parameter values in the affine space. The mechanisms of particle movement in the affine space are considered to reduce the weight of affine bonds using the developed operator directed mutation. The directed mutation operators are described, the essence of which is to change the integer values of the genes in the chromosome. A modified structure of the bee algorithm is proposed. The key operation of the algorithm is the study of promising positions lying in the vicinity of the base positions. The bee paradigm is a two-tier search strategy. At the first level, many base positions are generated in the search space. At the second level, in the search space, the neighborhoods of base positions are investigated. A method of forming positions in the vicinity of base positions is proposed. The best positions found in each neighborhood of the base positions at iteration t are part of the set of base positions used at the iteration (t + 1). The architecture of a hybrid algorithm based on the integration of bee swarm and bee colony methods was developed. Hybridization is as follows. Initially, a swarm of particles is formed. After moving a swarm of particles to new positions, these positions are considered as the basic positions found by the swarm of scout bees. Further, in accordance with the mechanisms of the bee colony, a new set of basic positions is formed, which, in turn, at the next iteration is considered as a swarm of particles. Test tests have shown that when integrating the behavior patterns of a swarm of bees and a swarm of particles, the results of the new hybrid algorithm are 11 to 18% better than each algorithm individually.
Keywords
	VLSI, placement, swarm intelligence, bee algorithm, swarm of particles, hybridization. multi-agent system, affine search space, directional mutation operator, neighborhood of base positions, bionic search.
Library reference
	Lebedev B.K., Lebedev O.B., Zhiglaty A.A. Placement of VLSI Elements Based on Swarm Intelligence Models // Problems of Perspective Micro- and Nanoelectronic Systems Development - 2020. Issue 4. P. 118-125. doi:10.31114/2078-7707-2020-4-118-125
URL of paper
	http://www.mes-conference.ru/data/year2020/pdf/D030.pdf