A Joint Optimization Algorithm for Dispatching Tasks in Agent-Based Workflow Management Systems

Workflow problems generally require the coordination of many workers; machines and computers. Agents provide a natural mechanism for modelling a system where multiple actors operate, but they do not explicitly support coordination schemes. Efficient task allocation to these actors is a fundamental coordination prerequisite. A competent allocation policy should address both system performance issues and users’ quality demands. Since these factors are often contradictory, an efficient solution is hard to be identified. In this study, we suggest a task delegation strategy that jointly optimizes system performance (as expressed by workload balancing) and quality demands (as expressed by minimum task overlapping). A consistent modelling approach allows us to transform data of both these factors into a matrix format. The next step is to exploit the Ky-fan theorem and the notion of generalized eigenvalues to optimally solve the task allocation problem. A simple scheduling policy and an experimental setup were applied to test the efficiency of the proposed algorithm.