Research Of Dynamic Optimization Methods And Its Application On Dynamic SW-HW Co-partitioning

In the real-world applications and the research fields, there exist many complex optimization problems. Some problem's optimum or constrain function may change over time during the problem solving process and they are named dynamic optimization problems (DOPs). In recent years, evolutionary computation obtains significant progress in the fields of dynamic optimization, and many excellent algorithms have been proposed. The research of dynamic algorithms in this paper is based on the Evoluationary Algorithms; focus mainly on the experimental analysis of Memory Scheme; successfully uses the dynamic optimization algorithm on the HW-SW partitioning problem and gets some meaningful results.The main work and contributions of this paper can be summarized as below:1. A new algorithm named"Bi-attractor Multi-population Particle Swarm Algorithm (BMPSO)"has been proposed. BMPSO uses two swarm attractors to perform searching; this method can balance the population diversity and the convergence speed effectively. In otherwise, this paper imports the Memory Scheme to the Multi-population PSO algorithm (MPSO) successfully. This method performs more adaptively in the periodic dynamic environment and can enhance the searching ability greatly in the problem of which has much historical concurrency.2. This paper conducts many experimental analyses on the Memory Scheme and summarizes the two property of Memory Scheme: 1. the effectiveness of the information conversation is the base of the Memory Scheme. 2. the performance of the Memory Scheme has much relation on the correlation of the historical environment.3. This paper researches basically on the SW-HW Co-partitioning problem and uses the MPBIL algorithm on the Dynamic SW-HW Co-partitioning problem. This method can solve the low efficiency problem of traditional dynamic algorithms at the beginning of the dynamic environment and enhance the performance and stably.