The Research Of HW/SW Partitioning Algorithm Based On Artificial Bee Colony Algorithm

The artificial bee colony is a novel group of intelligent optimization algorithm proposed in 2005, which is a heuristic optimization algorithm based on intelligent search strategy for some populations. Due to the advantages of few control parameters, computed conveniently and carried out easily. But at present, artificial bee colony algorithm has some shortcomings, such as local search ability, low accuracy, convergence is slow and other issues.In this paper, HW/SW partitioning for embedded systems is introduced, Take deep discussions about the principles, models. And then propose an improved artificial bee colony algorithm. Compare the original colony algorithm with particle swarm optimization, ant colony algorithm, artificial fish swarm algorithm and genetic algorithm, the results showed that the artificial bee colony algorithm has obvious advantages in HW/SW partitioning. To achieve the desired objectives, the improved artificial bee colony algorithm can greatly improve the efficiency in HW/SW partitioning. Detailed work in the paper is as follows:Firstly, this paper introduces the theory of the HW/SW partitioning modeling, some of the problems in it, and then a comprehensive analysis of the artificial bee colony algorithm, including the principles of artificial bee colony algorithm, the algorithm features, as well as the encoding algorithm fitness function.Secondly, it analyzes the performance index in the model, system architecture, and transplants the artificial bee colony algorithm into HW/SW partitioning.Thirdly, this article proposed an improved method based on adaptive neighborhood search. Then it is verified by experiment. In the DAG graph partitioning experiments, the times of cycle is 1200, the optimal solution of the improved ABC algorithm did not deteriorate, in reverse, it is even better. Meanwhile, the times of the optimal solution's appearance is increased, approximately 3.75 times, 3 times, 1 times, 1.43 times compared with the solutions of original algorithm. The searching precision of optimal solution is higher than original ABC algorithm, and the average time expenditure is reduced.Finally, the paper makes a brief summary and conclusion on the future research directions.