KM-based efficient algorithms for optimal packet scheduling problem in celluar/infostation integrated networks

Request models in previous works on the Packet Scheduling (PS) problem in cellular/infostation integrated networks usually have some restrictions which make them less practical in some situations. In this paper, we investigate the packet scheduling problem with our new request model, which is general enough to include previous request models as special cases. We first propose a Greedy algorithm for the PS problem and analyze its approximation ratio. We then transform our PS problem to the mini-slot mini-request version PS problem (MSMR-PS problem) by using the concept of mini-slot and mini-request. Next, by embedding problem information into bipartite graphs, we investigate the relationships between the MSMR-PS problem and the corresponding maximum matching (MM) problem on these graphs. Based on these relations, we propose efficient algorithms to obtain quasi-optimal solutions to the original PS problem by using the Kuhn-Munkres (KM) algorithm to solve the MM problems on the bipartite graphs. These algorithms include two offline algorithms (named as KMUpper and KMLower) and one online algorithm named KMOnline. KMUpper and KMLower return upper and lower bounds for optimal profit of the PS problem, respectively. Simulation results show that our algorithms are more preferable than the other existing algorithms like FIFO, exponential capacity algorithm.