Resource Service Composition and Its Optimal-Selection Based on Particle Swarm Optimization in Manufacturing Grid System

In distributed manufacturing systems, especially in a manufacturing grid (MGrid) system, there are primarily two kinds of manufacturing tasks (or resource service requests): (1) single resource service request task (SRSRTask), which can be completed by invoking only one resource service, and (2) multi-resource service request task (MRSRTask), which is completed by invoking several resource services in a certain sequence. For an SRSRTask, the system searches the resource services that are qualified for its function requirements and chooses the optimal one to execute it. For an MRSRTask, in addition to the search for all qualified resource services according to each subtask, the system selects one candidate resource service for each subtask. Then the system generates a new composite resource service (CRS) and selects the optimal resource service composite path from all possible paths to execute the task with the given multi-objective (e.g., time minimization, cost minimization, and reliability maximization) and constraints. The above problem is defined as multi-objective MGrid resource service composition and optimal-selection (MO-MRSCOS) problem in this paper. The formulation is presented for an MO-MRSCOS problem to minimize execution time and cost, and maximize the reliability. The basic resource service composite modes (RSCM) for CRS are described, and the principles for translating a complicated RSCM into a simple sequence RSCM are presented for simplifying the resolving process and complexity of MO-MRSCOS. A new MGrid resource service composition and optimal-selection method, based on the principles of particle swarm optimization (PSO), is then proposed. The PSO follows a collaborative population-based search, which models based on the social behavior of bird flocking and fish schooling. The case study demonstrates that the proposed method is useful in solving MO-MRSCOS problems. The experimental results and performance comparison show that the proposed method is both effective and efficient.