Although there has been a lot of progress in knowledgebased scheduling [5, 4], there is still a need for schedule improvement and repair through interaction with a human scheduler. There are several reasons for this. First, a user's preferences on the schedule are context dependent (e.g., may depend on the state of the scheduling environment at a particular time). Also, interactions among preferences and effective tradeoff very often depend on the particular schedule produced. This means that generally a user of the scheduling system can't fully specify his/her preferences a priori before getting the scheduling results from the system. By looking over the obtained schedule results, the user often thinks of additional preferences. Consider, for example a situation where a human scheduler does not like to use MACHINE-A which is substitutable for MACHIN -B but is of lower quality than MACHINE-B for processing ORDER-X. The reason high quality results are desired is that ORDER-X belongs to a quite important client.
Automated systems for planning and scheduling the activities of individuals, and of coordinated groups, must scale into increasingly complex, dynamic, stochastic, and even adversarial application domains where plan/schedule failure can be costly or catastrophic. Substantial research continues to be directed towards techniques for robust/probabilistic planning and scheduling, monitoring and repairing plans and schedules, continual distributed planning, etc. This work typically assumes that agents' objectives stay relatively stable over time, but plans and schedules need to adjust to deviations from predicted execution paths to account for violated assumptions and to exploit emergent opportunities. There is also growing interest in applications where agents' objectives (tasks, intentions, preferences...) can evolve over time. Even when a plan/schedule is progressing exactly as expected, the agents might reprioritize their tasks or alter how they intend to pursue their objectives.
The performance of anytime algorithms can be improved by simultaneously solving several instances of algorithm-problem pairs. These pairs may include different instances of a problem (such as starting from a different initial state), different algorithms (if several alternatives exist), or several runs of the same algorithm (for non-deterministic algorithms). In this paper we present a methodology for designing an optimal scheduling policy based on the statistical characteristics of the algorithms involved. We formally analyze the case where the processes share resources (a single-processor model), and provide an algorithm for optimal scheduling. We analyze, theoretically and empirically, the behavior of our scheduling algorithm for various distribution types.
Mammalian skeletal muscle harbors tissue-specific stem cells that are triggered to replace damaged fibers after injury. Genetic ablation of satellite cells in the mouse results in a failure to regenerate muscle, which indicates that these cells are the major (and possibly only) mediators for repair of skeletal muscle. Further evidence for the central role of satellite cells in muscle regeneration comes from transplantation experiments with genetically marked cells, which demonstrate that satellite cells are highly proliferative myogenic precursors capable of self?renewal and the resumption of quiescence, properties deemed important in a cell population responsible for muscle repair. Considerable in vitro evidence, derived from cultured fibers and myoblasts, is suggestive of a role for asymmetric division in generating both a self-renewing "immortal" stem cell and a differentiation-competent progenitor cell that proliferates and ultimately replaces damaged muscle. However, asymmetric division of satellite cells has not been documented in vivo.