A general framework integrating techniques for scheduling under uncertainty

For last years, a number of research investigations on task planning and scheduling under uncertainty have been conducted. This research domain comprises a large number of models, resolution techniques, and systems, and it is difficult to compare them since the existing terminologies are incomplete. However, we identified general families of approaches that can be used to structure the literature given three perpendicular axes. This new classification of the state of the art is based on the way decisions are taken. In addition, we propose a generation and execution model for scheduling under uncertainty that combines these three families of approaches. This model is an automaton that develops when the current schedule is no longer executable or when some particular conditions are met. The third part of this thesis concerns our experimental study. On top of ILOG Solver and Scheduler, we implemented a software prototype in C++ directly instantiated from our generation and execution model. We present new probabilistic scheduling problems and a constraintbased approach combined with simulation to solve some instances thereof