Optimal flow control allocation policies in communication networks with priorities

M(⩾2) transmitting stations sending packets to a single receiver over a slotted time-multiplexed link are considered. The objective is to characterize dynamic policies that minimize the discounted and long-term average costs due to holding packets at the M stations. This flow control problem is modeled as a Markov decision process and treated by dynamic programming methods. The authors derive properties of optimal (discounted) policies that reduce the computational complexity of the flow control algorithm, and prove two properties of optimal policies. These properties are used to show that the optimization reduces to the calculation of optimal allocations for a finite number of states. Some special cases are considered