Citation Context

... i w i [k℄ + p w i [k℄ 8i; k: Thus, we assume that the arrival process obeys a central limit theorem (CLT). Conditions on the arrival process under which it obeys a CLT are given, for example, in =-=[4]-=-. From the CLT assumption, it is easy to see that w i [k℄ can be upper and lower bounded as follows: i x i [k℄K p x i [k℄ w i [k℄ i x i [k℄ +K p x i [k℄ 8i; k; 12 for appropriate values of f...

Citation Context

...th information and the current channel state (i.e., without knowing channel or arrival statistics). Stable scheduling policies for wireless systems without time-varying channels were first studied in =-=[16]-=-. In fact the model in [16] can also be thought of as a model for a high-speed input-queued switch. Systems with time-varying channels, but limited to the case of ON and OFF channels were studied in [...

Citation Context

... not throughput optimal, but are fair according to some appropriate notion of fairness. Our work and other related work assume that the channel is time-varying and attempt to exploit this feature. In =-=[19]-=-, an interesting technique to induce time variations in channels which may not be inherently time-varying is discussed. The rest of the paper is organized as follows. Section 2 describes the system mo...

Citation Context

...]. In fact the model in [16] can also be thought of as a model for a high-speed input-queued switch. Systems with time-varying channels, but limited to the case of ON and OFF channels were studied in =-=[17, 14]-=-. More general channel models have been studied by others recently [1, 13, 12, 7]. We generalize the class of scheduling policies considered in [17, 14, 1, 13]. Further, we allow imperfect queue lengt...

Citation Context

...knowledge of system statistics. Alternatively, one can formulate a fair resource allocation problem where each user is allocated a certain fraction of the system resources according to some criterion =-=[18, 9]-=-. The approaches in [18, 9] are not throughput optimal, but are fair according to some appropriate notion of fairness. Our work and other related work assume that the channel is time-varying and attem...

Citation Context

...w imperfect queue length information and prove the stability of policies that reduce computational complexity. These class of policies for wireless networks are natural extensions of those studied in =-=[15, 5]-=- for high-speed switches. Our proof uses a quadratic Lyapunov Research supported by NSF Grants ANI-9714685, ITR 00-85929, DARPA grant F30602-00-2-0542 and AFOSR URI F49620-01-10365 1 function argumen...

Citation Context

... 1 [k℄ a 2 [k℄ a N [k℄ Figure 2: Uplink model drained when the channel is in State j. For now, we can simply visualize S j to be a bounded, convex region such as the broadcast channel capacity region =-=[8]-=-. Precise conditions on the allowable set of rates, the channel state process and arrival processes will be given later. In this paper, we consider the following class of scheduling policies: at any t...

Citation Context

...peed input-queued switch. Systems with time-varying channels, but limited to the case of ON and OFF channels were studied in [17, 14]. More general channel models have been studied by others recently =-=[1, 13, 12, 7]-=-. We generalize the class of scheduling policies considered in [17, 14, 1, 13]. Further, we allow imperfect queue length information and prove the stability of policies that reduce computational compl...

Citation Context

... uses a quadratic Lyapunov Research supported by NSF Grants ANI-9714685, ITR 00-85929, DARPA grant F30602-00-2-0542 and AFOSR URI F49620-01-10365 1 function argument along the lines of the proofs in =-=[16, 6]-=-. We also refer the reader to [2, 11] for a geometric approach to scheduling problems. In the context of time-varying wireless channels with many users, our work is an example of exploiting multiuser ...

Citation Context

...peed input-queued switch. Systems with time-varying channels, but limited to the case of ON and OFF channels were studied in [17, 14]. More general channel models have been studied by others recently =-=[1, 13, 12, 7]-=-. We generalize the class of scheduling policies considered in [17, 14, 1, 13]. Further, we allow imperfect queue length information and prove the stability of policies that reduce computational compl...

Citation Context

...]. In fact the model in [16] can also be thought of as a model for a high-speed input-queued switch. Systems with time-varying channels, but limited to the case of ON and OFF channels were studied in =-=[17, 14]-=-. More general channel models have been studied by others recently [1, 13, 12, 7]. We generalize the class of scheduling policies considered in [17, 14, 1, 13]. Further, we allow imperfect queue lengt...

Citation Context

...peed input-queued switch. Systems with time-varying channels, but limited to the case of ON and OFF channels were studied in [17, 14]. More general channel models have been studied by others recently =-=[1, 13, 12, 7]-=-. We generalize the class of scheduling policies considered in [17, 14, 1, 13]. Further, we allow imperfect queue length information and prove the stability of policies that reduce computational compl...

Citation Context

...peed input-queued switch. Systems with time-varying channels, but limited to the case of ON and OFF channels were studied in [17, 14]. More general channel models have been studied by others recently =-=[1, 13, 12, 7]-=-. We generalize the class of scheduling policies considered in [17, 14, 1, 13]. Further, we allow imperfect queue length information and prove the stability of policies that reduce computational compl...

Citation Context

...supported by NSF Grants ANI-9714685, ITR 00-85929, DARPA grant F30602-00-2-0542 and AFOSR URI F49620-01-10365 1 function argument along the lines of the proofs in [16, 6]. We also refer the reader to =-=[2, 11]-=- for a geometric approach to scheduling problems. In the context of time-varying wireless channels with many users, our work is an example of exploiting multiuser diversity to maximize the capacity of...

Citation Context

...w imperfect queue length information and prove the stability of policies that reduce computational complexity. These class of policies for wireless networks are natural extensions of those studied in =-=[15, 5]-=- for high-speed switches. Our proof uses a quadratic Lyapunov Research supported by NSF Grants ANI-9714685, ITR 00-85929, DARPA grant F30602-00-2-0542 and AFOSR URI F49620-01-10365 1 function argumen...

Citation Context

...positive constants ; : There are many examples of stationary 13 stochastic processes that satisfy such a constraint when the arrival process is further peak-rate constrained. We refer the reader to =-=[10]-=- for one such example. The leaky-bucket constraint limits the burstiness of the arrivals, which in turn enables us to upper-bound the difference between x i and i w i ; with high probability, by a l...

Citation Context

...knowledge of system statistics. Alternatively, one can formulate a fair resource allocation problem where each user is allocated a certain fraction of the system resources according to some criterion =-=[18, 9]-=-. The approaches in [18, 9] are not throughput optimal, but are fair according to some appropriate notion of fairness. Our work and other related work assume that the channel is time-varying and attem...

Citation Context

...supported by NSF Grants ANI-9714685, ITR 00-85929, DARPA grant F30602-00-2-0542 and AFOSR URI F49620-01-10365 1 function argument along the lines of the proofs in [16, 6]. We also refer the reader to =-=[2, 11]-=- for a geometric approach to scheduling problems. In the context of time-varying wireless channels with many users, our work is an example of exploiting multiuser diversity to maximize the capacity of...