This thesis describes a fully distributed approach to resource scheduling within a sliding time frame, implemented for a system of agents across multiple JADE platforms. All agents, with operations inside the current window, schedule tasks using recursive propagation and a sorting algorithm. Operations outside the window are not scheduled until either the sliding window has advanced to encompass them or until gaps have opened between tasks inside the sliding window to accommodate them. The distributed sliding window approach to scheduling addresses many of the problems afflicting both centralized systems, including scalability, robustness, and responsiveness to dynamic changes. It also provides full decentralization as compared with other distributed approaches.
展开▼
