Abstract : We address the problem of building an interruptible real-time system using non-interruptible components. Some artificial intelligence techniques offer a tradeoff between computation time and quality of results, but their run-time must be determined when they are activated. These techniques, called contract algorithms, introduce a complex scheduling problem when there is uncertainty about the amount of time available for problem-solving. We show how to optimally sequence contract algorithms to create the best possible interruptible system with or without stochastic information about the deadline. These results extend the foundation of real-time problem-solving and provide useful guidance for embedding contract algorithms in applications.
Shlomo Zilberstein, François Charpillet, Philippe Chassaing. Optimal Sequencing of Contract Algorithms. Annals of Mathematics and Artificial Intelligence, Springer Verlag, 2003, 39 (1), pp.1-18. ⟨10.1023/A:1024412831598⟩. ⟨inria-00099720⟩