A Least Mean Loss Algorithm for Scheduling Traffic with Deadlines

[[abstract]]In this thesis we study the problem of scheduling multiple real-time streams with deadlines over a shared wireless channel. We propose to schedule the packets from the queues based on a least mean packet loss (LMPL) policy. The computation on the number of mean lost packets takes into account the transmission errors of respective users and the deadlines of their head of line (HOL) packets. A recursive algorithm is developed to conduct the computation. Examples show that the proposed policy can obtain less lost packets than the earliest deadline first (EDF) or feasible earliest deadline first (FEDF) policy. We also modify the algorithm so that it can be applied to Rayleigh fading channel based on finite state Markov chain (FSMC) model. With the FSMC model, the LMPL algorithm presumes future channel conditions based on the state transition probabilities associated with the FSMC model. Thus, the modified version is more suitable for practical applications