Toward Unobtrusive Service Composition in Smart Environments based on AI Planning

Marquardt, Florian and Uhrmacher, Adelinde M. (2009) Toward Unobtrusive Service Composition in Smart Environments based on AI Planning. In: 23th Workshop Planning, Scheduling, Design, and Configuration (PuK 2009), September 15-18, Paderborn, Germany. Proceedings,, pp. 228-239.

Full text not available from this repository.
Official URL: http://www.puk-workshop.de/puk2009/KI2009_WS_Proce...

Abstract

Using AI planning for service composition in smart environments requires fast and reliable implementations of planning algorithms. Experimentally the behavior of several planners were investigated with respect to the number of operators, of state variables, of true propositions in the initial world, of goals, and of preconditions and effects by executing several millions of experiment runs on about 55.000 generated problem configurations. Based on the experimental results we outline a heuristic that in addition leverages the knowledge of phase shifts. When a planner’s execution time crosses a certain threshold, it is canceled. It is shown that this strategy leads to a significant speedup without losing too many solutions and thus, presents a promising step toward realizing unobtrusive service composition in smart environments.

Item Type: Conference or Workshop Item (Paper)