Generating Executable Workflows from Solution Plans

The paper presents an extension of solution plans generated by the graph-plan algorithm developed for AI planning. We post process the solution plans in order to generate executable workflows. We make the following main contributions: (i) we generate a Workflow Tree from the solution plans by encapsulating the actions into abstract workflow Blocks, (ii) we develop a methodology that uses the specification of workflow patterns in terms of Abstract State Machines in order to find the appropriate pattern to replace the abstract Block constructs in the workflow tree. (iii) We provide semi-automated support for agents in order to select the most specific pattern that fits their behavior at run-time. We have implemented the approach and our experiments with some real-world case studies show the viability of the proposed approach.