Wolf pack algorithm for solving VLSI design tasks

Abstract

The paper deals with the development of new and modified heuristic mechanisms of searching optimal solutions, which is considered as one of the main problems of artificial intelligence. One of the promising areas of artificial intelligence development is application of methods and models of biological systems behavior for solving NP-complete and NP-difficult optimization tasks. The paper presents the statement of the placement task in designing very large-scale integration circuits (VLSI). The authors propose the algorithm for solving this task on the basis of biological system behavior in nature, e.g. wolf pack. Wolves are considered as typical social animals having clear separation of social work. The paper describes the actions and rules of wolf pack behaviour in nature. Based on the wolves’ behaviour rules and actions, the authors present the modified wolf pack algorithm. The benefits of the developed modified algorithm include the ability to improve each following iteration of the placement task. The wolf pack algorithm is implemented as a computer software on Java. To estimate the effectiveness of the proposed approach, the authors use the well-known IBM benchmarks to compare with the developed algorithm. The comparison is implemented with the results of the following algorithms: Capo 8.6, Feng Shui 2.0, Dragon 2.23. The results show that the wolf pack algorithm is more effective than the analogues.