Online Learning Algorithms for Dynamic Scheduling Problems

This paper presents game theoretic and reinforcement learning approach to dynamic scheduling. In dynamic scheduling problems all input is not available at the starting of scheduling. Input arrives online at the time of scheduling. So scheduling algorithm should be capable of scheduling the piece of input available till the time and schedule it as the input arrives. We used game theoretic and reinforcement learning approaches for dynamic scheduling. The learning approaches are useful when the scheduling requests repeat themselves for long duration. We developed the mechanism for scheduling in which all process work selfishly to maximize their own reward. Processes use reinforcement learning approach to select the suitable action. Finally the scheduling is performed on the basis of their actions.