R. Rivest, "Network control by bayesian broadcast," Cambridge, MA: MIT, Laboratory for Computer Science, Report MIT/LCS/TM-285.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....other analyses is that we do not assume Poisson arrivals. Instead we model the correlated arrivals by having each of the N stations in the population transmit periodic batches of capacity requests to the HC. This scenario is termed the disaster scenario by members of the IEEE 802.14 committee [2,10] where one assumes that the totality of the stations are all simultaneously powered up. The performance of a simple p persistence CRA in the F CPR protocol as in [13] is studied in [7] however, this paper shows that a more complex CRA can deliver in practice significantly higher stability and ....

....can achieve a stable maximum throughput of e 1 under Poisson arrivals. Note, however, that the method requires all the MTs to use the same retransmission probability and since the arrival rate is not known this must be estimated as well causing increased control complexity in the wireless system [10]. Such complexity is to be avoided since errors in the multicast of the value of p to be used from Figure 4. Comparison of simulated results for adaptive p persistence, fixed p persistence, the basic binary tree algorithm and msSTART. The adaptive p persistence can never reach the maximum L crit ....

R.L. Rivest, Network control by Bayesian broadcast, Report MIT/LCS/TM-285, Laboratory for Computer Science, Cambridge, MA (1985).

.... special issue, 0] of the IEEE Transactions on Information Theory, and in partic ular to [8] and [25] The study of such multi access schemes is by now almost two decades old, but the area continues to attract considerable research activity, witness e.g. 3] 9] 14] 15] 18 , 20] 21] 22 , 24] which have all appeared in print in the last two years. Most of the successful analyses of ALOHA and related collision resolution protocols have relied on the infinite user assumption. liere each node regards itself as tt collection of virtual nodes, one for each arriving packet. Tile ....

R. L. Rivest, "Network Control by Bayesian Broadcast", IEEE Transactions .on Information Theory, Vol. IT-33, NO. 3, PP. 323-328, 1987.

....as the estimated current number of active stations. An active station performs a broadcast at step t with probability 1=A(t) This parameter A(t) is updated after a step by adding a suitable constant among some xed three ones, depending on the respective feedback from the channel. Rivest [94] proposed a speci c additive adaptive protocol in which parameter A(t) is updated as follows: if during step t there is no collision then A(t 1) maxf1; A(t) 1 1 e g otherwise A(t 1) A(t) 1 e 2 1 e . Tsitsiklis [108] proved that this protocol is stable for the arrival rate e ....

R.L. Rivest, Network control by Bayesian broadcast, IEEE Trans. on Information Theory 33 (1987) 323-328.

....The stability of classical ALOHA systems has been studied extensively. For an in nite number of users, it has been found in [Cap78] that the system is unstable. Stability regions have been found for systems with a nite number of users. To combat the instability, decentralized control schemes [Riv87] and con ict resolution schemes [Hay78] have been established. In general, such schemes attempt to avoid successive collisions by retransmitting with some backo policy. Modeling a collision as leading to loss of all packets at receivers is not always practical. The capture phenomenon, for ....

R. Rivest. Network control by Bayesian broadcast. IEEE Transactions on Information Theory, 33:323-328, 1987.

....being adapted to the structure of DR TDMA. The FPBP protocol described in detail in [11] has been designed to satisfy these requirements, and is employed by mobile stations to access uplink control mini slots. This protocol is an extension of Rivet s pseudo Bayesian Aloha stabilization algorithm [12] that provides multiple levels of access priorities and is adapted to the frame structure of DR TDMA. We now summarize the control traffic access algorithm. Suppose that control mini slots are available in frame . is computed each frame using the algorithm presented in Section IV D. There are ....

R. L. Rivest, "Network control by Bayesian broadcast," IEEE Trans. Inform. Theory, vol. IT--33, pp. 323--328, May 1987.

....not feasible, for example due to long delay success failure feedback, or the existence of hidden stations. The potential benefit of using the announced DAP protocol for such networks can be very significant. The idea of adapting the access probability to the channel load was suggested in [8] and [13]. The method suggested in those studies is known as stabilized slotted ALOHA [2] Unfortunately, the strategies suggested in these studies are not applicable on many networks, such as satellite based networks and cellular networks. The reason for this is that these strategies are based on ....

Rivest R.L., Network control by Bayesian broadcast, MIT report No. 285, 1985.

....while being adapted to the structure of DR TDMA. The FPBP protocol described in detail in [11] has been designed to satisfy these requirements, and is employed by mobile stations to access uplink control mini slots. This protocol is similar to Rivet s pseudo Bayesian Aloha stabilization algorithm [12] and is based on the assumption that the probability distribution of the number of control packets waiting for transmission is Poisson. In addition to the standard Aloha algorithm, the FPBP protocol provides multiple levels of access priorities and is adapted to the frame structure of DR TDMA. ....

R. L. Rivest, "Network control byBayesian broadcast," IEEE Transactions on Information Theory,vol. 33, pp. 323--328, May1987.

....as the estimated current number of active stations. An active station performs a broadcast at step t with probability 1=A(t) This parameter A(t) is updated after a step by adding a suitable constant among some fixed three ones, depending on the respective feedback from the channel. Rivest [86] proposed a specific additive adaptive protocol in which parameter A(t) is updated as follows: if during step t there is no collision then A(t 1) maxf1; A(t) Gamma 1 1 e g otherwise A(t 1) A(t) 1 e Gamma2 1 e . Tsitsiklis [99] proved that this protocol is stable for the arrival ....

R.L. Rivest, Network control by Bayesian broadcast, IEEE Trans. on Information Theory 33 (1987) 323--328.

....and to retransmit backlogged packets with a fixed probability q r in subsequent slots, is unstable for anyvalue of the arrival rate. Thus to implement a stabilized slotted Aloha with priority classes, we derive an algorithm similar to the pseudo Bayesian Aloha stabilization algorithm presented in [14] and [13] Let new packets be regarded as backlogged immediately after their arrivals at the respective mobile terminals. They will attempt transmission in subsequent slots until success with a probability determined by their priority class and the estimated backlogged state of the system. ....

R. L. Rivest, "Network control byBayesian broadcast," IEEE Transactions on Information Theory,vol. 33, pp. 323-- 328, May 1987.

....be guaranteed only when the value of p is dynamically adjusted to follow the number of contending users 2 , that is, when p = 1=N a [4] In general the exact number of contending users is not known and an estimator to adapt p is required. We use the pseudo Bayesian estimator proposed by Rivest [6]. In this adaptive p implementation, unlike conventional Aloha, new arrivals also apply the probability p before transmitting and thus are not distinguished from old contenders. 2.2 Tree Algorithm In a tree algorithm, all users involved in a collision are divided in n sub groups. Each user ....

....(curve i) corresponds to an ideal implementation that adjusts p according to the exact number of contending users. This implementation bounds the performance for a real implementation of p persistence using an estimator. The second implementation (curve e) uses the estimator proposed by Rivest in [6] to adapt the p value (implemented as in [7] Curves f and b give the results for the tree scheme with free access and blocked access respectively both operating with a single interleave in breadth priority order (as described in Section 3.2 and [11] Access delay is defined as the time from the ....

R. L. Rivest, "Network Control by Bayesian Broadcast" Technical Report MIT/LCS/TM287, MIT Lab for Computer Science, Cambridge, Massachusetts, 1985.

....so as to keep G close to 1. The papers [63] and [64] independently gave the first proofs that finite mean delay can be achieved for the canonical model (with Poisson arrivals, corresponding to infinite N ) Several other stabilization algorithms were given, including an interesting one of Rivest [65] based on Bayesian estimation. The famous exponential back off algorithm, that basically reduces the retransmission probability of a packet by a factor of 2 every time the packet experiences a collision, preoccupied the minds of many researchers for a while. It was initially conjectured that ....

R. Rivest, "Network control by Bayesian broadcast," IEEE Transactions on Information Theory, pp. 323-328, May 1987.

....n ap;pr (k) with p pr (0) 1 (a 0 ; a 1 ; a c ) i e 1 Gamma2=e 1 Gamma1=e j ; 1; i e Gamma1=e 1 Gamma1=e j (1:519; 1; 0:559) 6) 3.3 Pseudo Bayesian Algorithm In contrast to the algorithms in sec. 3.1 and 3. 2, the pseudo Bayesian algorithm, proposed by Rivest in [6], estimates the number of active stations of the next time slot. Let n(k 1) denote the estimate of active stations and (k 1) the arrival rate of new packets in access window k 1. Then, the pseudo Bayesian algorithm adapted to PDO is with n(0) 1 n(k 1) max ae (k) 7) n(k) ....

R. L. Rivest. Network Control by Bayesian Broadcast. IEEE Transactions on Information Theory, Vol. IT--33, No. 3, pp. 323--328, May 1987.

No context found.

R. Rivest, "Network control by bayesian broadcast," Cambridge, MA: MIT, Laboratory for Computer Science, Report MIT/LCS/TM-285.

No context found.

R. L. Rivest. Network control by bayesian broadcast. IEEE Transactions on Information Theory, IT-33(3):323--328, May 1987.

No context found.

R. L. Rivest, "Network control by Bayesian broadcast," IEEE Trans. Inform. Theory, vol. IT-33, pp. 323--328, 1987.

No context found.

R. Rivest, "Network control by Bayesian broadcast," IEEE Trans. Inform. Theory, vol. IT-33, pp. 323--328, May 1987.

No context found.

Rivest R.L., Network control by Bayesian broadcast, MIT report No. 285, 1985.

No context found.

R. Rivest, "Network control by Bayesian broadcast," IEEE Transactions on Information Theory, vol. 33, pp. 323--328, 1987.

No context found.

R. Rivest, Network Control by Bayesian Broadcast. IEEE Transactions on Information Theory, IT-33, pp. 323--328, 1987.

No context found.

R. Rivest, Network Control by Bayesian Broadcast. IEEE Transactions on Information Theory, IT-33, pp. 323--328, 1987.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC