Planning and Rescheduling

This thesis presents results in two areas. The first area is automated planning. Approaches to the domain-independent automated planning can be divided into two groups: STRIPS-like planning (based on operators) and HTNlike planning (based on hierarchical decomposition). It is widely believed, that the expressivity of STRIPS-like planning is lower than the expressivity of HTN-like planning. This would mean that HTN-like planning can solve more problems than STRIPS-like planning. In this thesis we show, that the expressivity of both approaches is identical and that both approaches can solve all problems solvable by a finite tape Turing machine (i.e. solvable by a common computer). The second part of this thesis addresses rescheduling. If we have a schedule with unmovable deadline, we need to react on late activities by shortening of the remaining schedule part. In this thesis, we show that every rescheduling corresponds to a graph cut or the superposition of several graph cuts. Based on this fact, we designed a new algorithm for the cheapest rescheduling. The algorithm is based on the conversion of the rescheduling problem to the problem of minimal graph cut.