## Distributed opportunistic scheduling for ad hoc communications with imperfect channel information,” Submitted to

Venue: | V. CONCLUSION In |

Citations: | 19 - 7 self |

### BibTeX

@INPROCEEDINGS{Zheng_distributedopportunistic,

author = {Dong Zheng and Man-on Pun and Weiyan Ge and Junshan Zhang and H. Vincent Poor},

title = {Distributed opportunistic scheduling for ad hoc communications with imperfect channel information,” Submitted to},

booktitle = {V. CONCLUSION In},

year = {}

}

### OpenURL

### Abstract

Abstract — Distributed opportunistic scheduling is studied for wireless ad-hoc networks, where many links contend for one channel using random access. In such networks, distributed opportunistic scheduling (DOS) involves a process of joint channel probing and distributed scheduling. It has been shown that under perfect channel estimation, the optimal DOS for maximizing the network throughput is a pure threshold policy. In this paper, this formalism is generalized to explore DOS under noisy channel estimation, where the transmission rate needs to be backed off from the estimated rate to reduce the outage. It is shown that the optimal scheduling policy remains to be threshold-based, and that the rate threshold turns out to be a function of the variance of the estimation error and be a functional of the backoff rate function. Since the optimal backoff rate is intractable, a suboptimal linear backoff scheme that backs off the estimated signal-to-noise ratio (SNR) and hence the rate is proposed. The corresponding optimal backoff ratio and rate threshold can be obtained via an iterative algorithm. Finally, simulation results are provided to illustrate the tradeoff caused by increasing training time to improve channel estimation at the cost of probing efficiency. I.

### Citations

1677 |
A Course in Game Theory
- Osborne, Rubinstein
- 1994
(Show Context)
Citation Context ... m = 1, 2, · · · , M. xm∈Am (19) B. Nash Equilibrium for Non-Cooperative Game Treating the rate threshold selection problem as a noncooperative game, we investigate the corresponding Nash equilibrium =-=[31]-=-. 7 . (16) Definition 4.1: A threshold vector x ∗ = {x ∗ 1 , x∗ 2 , · · · , x∗ M } is said to be a Nash equilibrium of game G, if for every link m φm(x ∗ m ,x∗ −m ) ≥ φm(xm,x ∗ −m ), ∀ xm ∈ Am, (20) w... |

745 |
Nonlinear Programming
- Bertsekas
- 2004
(Show Context)
Citation Context ...he fixed point equation (25), it is not feasible to characterize σ∗ using the first order condition. In what follows, we devise an iterative algorithm instead using fractional optimization techniques =-=[2]-=-. Specifically, we define the following functions: » 1 − exp j − 1 αρeff „ 1 σ U(σ, x) � j log(1 + σρeff ˆ λ ′ ) exp(−ˆ λ ′ ) + exp «ff– − 1 × „ « 1 σρeff E1 „ ˆλ ′ + 1 σρeff and V (σ, x) � δ ps + exp... |

504 | How bad is selfish routing
- Roughgarden, Tardos
- 2000
(Show Context)
Citation Context ...operative game, i.e., x∗ co > x∗nco . The proof is relegated to Appendix K. As expected, Proposition 5.1 implies that for M ≥ 2, the efficiency η defined as η � x∗ nco /x∗co is strictly less than one =-=[33]-=-. 10 B. Non-Cooperative Game with Pricing The Nash equilibrium is a solution to the non-cooperative game, where no link can improve its throughput any further through individual effort. Clearly, the n... |

463 | Fair End-to-end Window-based Congestion Control
- Mo, Walrand
- 1998
(Show Context)
Citation Context ...ion of its rate and serves as a measure of satisfaction that user m has from sharing the channel. For example, the reward function (utility function), denoted {Um(r), ∀ m} can take the following form =-=[30]-=-: Um,κ(r) = � wm log r, if κ = 1 wm(1 − κ) −1 r 1−κ , κ ≥ 0, κ �= 1. (14) Then, the optimal strategy for distributed opportunistic scheduling is to characterize the optimal stopping rule N ∗ U for max... |

453 | Opportunistic beamforming using dumb antennas
- Viswanath, Tse, et al.
- 2002
(Show Context)
Citation Context ... in channel-aware scheduling and channel-aware access control. Channel aware opportunistic scheduling was first developed for the downlink transmissions in multiuser wireless networks (see, e.g., [3]-=-=[8]-=-). Opportunistic scheduling originates from a holistic view: roughly speaking, in a multiuser wireless network, at each moment it is likely that there exists a user with good channel conditions; and b... |

409 | Link-level measurements from an 802.11b mesh network
- Aguayo, Bicket, et al.
(Show Context)
Citation Context ...s to separate link losses caused by fading from those by interference. That is, the PHY layer addresses the issues of fading, and the MAC layer addresses the issue of contention. However, as shown in =-=[1]-=- [2], fading can often adversely affect the MAC layer in many realistic scenarios. The coupling between the timescales of fading and MAC calls for a unified PHY/MAC design for wireless ad-hoc networks... |

388 | A framework for uplink power control in cellular radio systems
- Yates
- 1995
(Show Context)
Citation Context ...worth noting that the convergence proof for this distributed iterative algorithm is nontrivial, and the standard approaches (e.g., using contraction mapping [17] or standard interference functions in =-=[18]-=-) are not applicable here. Indeed, the proof is constructive, and involves an interesting sandwich argument. Observing that the best response strategy requires message passing from neighboring nodes, ... |

326 | A rate-adaptive {MAC} protocol for multi-hop wireless networks
- Holland, Vaidya, et al.
(Show Context)
Citation Context ...el information from potential receivers, thereby serving as a tentative “virtual” base station. A key observation is that in the existing work on rate adaptation for ad hoc communications (see, e.g., =-=[12]-=-, [13], [14]), a link continues transmission after a successful channel contention, no matter whether the channel condition is good or poor. Clearly, this leaves much room for improvement by devising ... |

226 | Opportunistic media access for multirate ad hoc networks
- Sadeghi, Kanodia, et al.
- 2002
(Show Context)
Citation Context ...on from potential receivers, thereby serving as a tentative “virtual” base station. A key observation is that in the existing work on rate adaptation for ad hoc communications (see, e.g., [12], [13], =-=[14]-=-), a link continues transmission after a successful channel contention, no matter whether the channel condition is good or poor. Clearly, this leaves much room for improvement by devising channel-awar... |

199 | Efficient power control via pricing in wireless data networks
- Saraydar, lvlandayam, et al.
- 2002
(Show Context)
Citation Context ...ing, assuming that each link has no knowledge of other links’ channel conditions. There has also been a surge of interest in using game theory to study wireless networks (see, e.g., [23], [24], [25], =-=[26]-=-). We note that a game theoretic formulation on random access protocols has been investigated, and the most relevant to our study are perhaps [27], [28], [29], with one major difference being that non... |

197 |
Stochastic Approximation and Recursive Algorithms and Applications
- Kushner, Yin
- 2003
(Show Context)
Citation Context ...[xm(k) + am(k)�gm(x)I{k ∈ N m }]. Sm(k) � gm(x(k)). (75) (33) can be rewritten as xm(k + 1) = xm(k) + Dm(k) + am(k)Sm(k)I{k ∈ N m } + am(k)wm(x(k))I{k ∈ N m }, (76) where Dm(k) is the projection term =-=[38]-=-. The last two terms correspond to the asynchronism and observation noise, respectively. By showing that the effect of these two terms diminishes as k → ∞, the problem can be converted to a classical ... |

184 |
Providing quality of service over a shared wireless link
- Andrews, Kumaran, et al.
- 2001
(Show Context)
Citation Context ...rest in channel-aware scheduling and channel-aware access control. Channel aware opportunistic scheduling was first developed for the downlink transmissions in multiuser wireless networks (see, e.g., =-=[3]-=--[8]). Opportunistic scheduling originates from a holistic view: roughly speaking, in a multiuser wireless network, at each moment it is likely that there exists a user with good channel conditions; a... |

165 | User-level performance of channel-aware scheduling in wireless data networks
- Borst
- 2003
(Show Context)
Citation Context ...hancing MAC protocol design for wireless LANs and wireless mesh networks. C. Related Work and Organization As noted above, there has been much work on centralized opportunistic scheduling (e.g., [3], =-=[4]-=-, [5], [6], [7], [8], [20]), channel-aware ALOHA (e.g., [9], [10]) and MAC design with rate adaptation (e.g., [12], [13], [14]). Most relevant to our study are perhaps (e.g., [9], [10], [12], [14]). T... |

158 |
Calculus of Variations
- Gelfand, Fomin
- 2000
(Show Context)
Citation Context ...c), i.e., λ ∗ c = arg maxx λc∈A ∗ (λc), (23) where A is the set of the admissible functions (for example, A can be {λc( ˆ λ) : λc( ˆ λ) ≥ 0, ∀ ˆ λ ≥ 0}). Based on the theory of calculus of variations =-=[4]-=-, problem (23) is a variational problem, and the functions λ∗ c (·) are called extremals. However, unlike the canonical calculus of variations problems, in this problem, the functional x∗ is not expli... |

136 |
Information capacity and power control in a single cell multiuser environment
- Knopp, Humblet
- 1995
(Show Context)
Citation Context ...ng MAC protocol design for wireless LANs and wireless mesh networks. C. Related Work and Organization As noted above, there has been much work on centralized opportunistic scheduling (e.g., [3], [4], =-=[5]-=-, [6], [7], [8], [20]), channel-aware ALOHA (e.g., [9], [10]) and MAC design with rate adaptation (e.g., [12], [13], [14]). Most relevant to our study are perhaps (e.g., [9], [10], [12], [14]). The ma... |

128 | A framework for opportunistic scheduling in wireless networks
- Liu, Chong, et al.
- 2003
(Show Context)
Citation Context ...C protocol design for wireless LANs and wireless mesh networks. C. Related Work and Organization As noted above, there has been much work on centralized opportunistic scheduling (e.g., [3], [4], [5], =-=[6]-=-, [7], [8], [20]), channel-aware ALOHA (e.g., [9], [10]) and MAC design with rate adaptation (e.g., [12], [13], [14]). Most relevant to our study are perhaps (e.g., [9], [10], [12], [14]). The main di... |

93 | Goodput Analysis and Link Adaptation for IEEE 802.11a Wireless LANs
- Qiao, Choi, et al.
- 2002
(Show Context)
Citation Context ...ormation from potential receivers, thereby serving as a tentative “virtual” base station. A key observation is that in the existing work on rate adaptation for ad hoc communications (see, e.g., [12], =-=[13]-=-, [14]), a link continues transmission after a successful channel contention, no matter whether the channel condition is good or poor. Clearly, this leaves much room for improvement by devising channe... |

84 |
Non-cooperative uplink power control in cellular radio systems
- Ji, Huang
- 1998
(Show Context)
Citation Context ...uted scheduling, assuming that each link has no knowledge of other links’ channel conditions. There has also been a surge of interest in using game theory to study wireless networks (see, e.g., [23], =-=[24]-=-, [25], [26]). We note that a game theoretic formulation on random access protocols has been investigated, and the most relevant to our study are perhaps [27], [28], [29], with one major difference be... |

81 |
Great expectations: the theory of optimal stopping
- Chow, Robbins, et al.
(Show Context)
Citation Context ... .} be a sequence of real positive numbers, and lim l→∞ xl = x, then R (n) − xl → R (n) − x almost surely. Since |R (n) − xl| ≤ R (n), we have that f(xl) → f(x) by using Dominated Convergence Theorem =-=[15]-=-. Since f(x) decreases from E[R (n)] to 0 and the right hand side of (5) strictly increases from 0 to ∞ as x grows, it follows that (5) has a unique finite solution. ps B. PROOF OF PROPOSITION 3.2 It ... |

77 |
Probability and measure (3rd Ed
- Billingsley
- 1995
(Show Context)
Citation Context ...up E RNT − x N∈Q ∗ � �� τ · N + T ps � � (b) ≤ E sup TRn − x n ∗ � ��� τ · n + T , (55) ps where (a) is by the definition of N ∗ , and (b) can be obtained using the same technique as in Fatou’s Lemma =-=[34]-=-. It is clear that for any x1 ≤ x2, � � � ��� τ E sup R (n)T − x1 · n + T n ps � � � ��� τ ≥ E sup R (n)T − x2 · n + T . n ps It follows from (55) that x ∗ ≤ � E[R2 ] 2 δ . ps C. PROOF OF PROPOSITION ... |

71 | Dynamic rate control algorithms for HDR throughput optimization
- Borst, Whiting
(Show Context)
Citation Context ...gn for wireless LANs and wireless mesh networks. C. Related Work and Organization As noted above, there has been much work on centralized opportunistic scheduling (e.g., [3], [4], [5], [6], [7], [8], =-=[20]-=-), channel-aware ALOHA (e.g., [9], [10]) and MAC design with rate adaptation (e.g., [12], [13], [14]). Most relevant to our study are perhaps (e.g., [9], [10], [12], [14]). The main differences betwee... |

64 | Exploiting Multiuser Diversity for Medium Access Control in Wireless Networks
- Qin, Berry
- 2003
(Show Context)
Citation Context ...wireless communications has recently emerged as a promising technique for improving spectral efficiency in wireless networks. Most existing studies along this line require centralized scheduling (see =-=[5]-=- and the references therein), and little work has been done on developing distributed algorithms to harvest diversity gains for ad hoc communications. Unlike centralized cases, in ad hoc communication... |

63 |
Convex Analysis and Optimization
- Bertsekas
- 2003
(Show Context)
Citation Context ... compact convex set for any m and the utility function φm(·) is quasiconcave on Am. Recall that a function f : Rn → [0, ∞) is quasi-concave if the sublevel sets Sc = {x|f(x) ≥ c} are convex for all c =-=[35]-=-. To this end, rewrite φm(x,x−m) ≥ c as ˆφm(x) � ps,m � ∞ x (r − c)dFm(r) − cδ −c � ps,j (1 − Fj(xj)) ≥ 0. (59) j�=m Then, it suffices to show that for any given c, ˆ φm(x) is quasiconcave in [0, ∞). ... |

60 | Exploiting decentralized channel state information for random access
- Adireddy, Tong
- 2002
(Show Context)
Citation Context ...el aware random access has been investigated for the uplink transmissions in a many-to-one network, where channel probing can be realized by broadcasting pilot signals from the base station. Notably, =-=[9]-=-, [10] study opportunistic ALOHA under a collision model, with a basic idea being that in every slot each user transmits with a probability based on its own channel condition. While recent work [11] d... |

47 | Optimal channel probing and transmission scheduling for opportunistic spectrum access
- Chang, Liu
- 2007
(Show Context)
Citation Context ... channel condition, regardless of whether the channel condition is poor or good. Along a different avenue, opportunistic channel probing for single-user multichannel systems has been studied in [21], =-=[22]-=-, where the basic idea is to opportunistically probe and select a transmission channel among multiple channels between the transmitter node and the receiver node. In contrast, in this study, we consid... |

45 |
Exploiting medium access diversity in rate adaptive wireless lans
- Ji, Yang, et al.
- 2004
(Show Context)
Citation Context ...y, [9], [10] study opportunistic ALOHA under a collision model, with a basic idea being that in every slot each user transmits with a probability based on its own channel condition. While recent work =-=[11]-=- does not assume a base station in a wireless LAN, the transmitter node still needs to collect channel information from potential receivers, thereby serving as a tentative “virtual” base station. A ke... |

44 |
bandwidth sharing algorithms based on game theory frameworksfor wireless ad-hoc networks
- Fang, Bensaou, et al.
- 2004
(Show Context)
Citation Context ...istributed scheduling, assuming that each link has no knowledge of other links’ channel conditions. There has also been a surge of interest in using game theory to study wireless networks (see, e.g., =-=[23]-=-, [24], [25], [26]). We note that a game theoretic formulation on random access protocols has been investigated, and the most relevant to our study are perhaps [27], [28], [29], with one major differe... |

31 | Asynchronous stochastic approximations
- Borkar
- 1998
(Show Context)
Citation Context ...tic iterative algorithm based on local observations only. In light of the asynchronous feature of the online algorithm, we appeal to recent results on asynchronous stochastic approximation algorithms =-=[19]-=- and establish its convergence under some regularity conditions. Finally, since the equilibrium point does not maximize the social utility, we examine the efficiency loss in terms of the throughput in... |

27 |
Stochastic approximation and its applications
- Chen
- 2002
(Show Context)
Citation Context ...ergence of {x(t)} by appealing to the mean ordinary differential equation (ODE) method. The convergence of the original sequence {x(k)} then follows by using the similar argument as in Section 1.3 in =-=[37]-=-. Towards this end, define Dm(k) � [xm(k) + am(k)�gm(x(k))I{n ∈ N m }] b 0 − [xm(k) + am(k)�gm(x)I{k ∈ N m }]. Sm(k) � gm(x(k)). (75) (33) can be rewritten as xm(k + 1) = xm(k) + Dm(k) + am(k)Sm(k)I{k... |

22 |
Jointly optimal transmission and probing strategies for multichannel wireless systems
- Guha, Munagala, et al.
- 2006
(Show Context)
Citation Context ...urrent channel condition, regardless of whether the channel condition is poor or good. Along a different avenue, opportunistic channel probing for single-user multichannel systems has been studied in =-=[21]-=-, [22], where the basic idea is to opportunistically probe and select a transmission channel among multiple channels between the transmitter node and the receiver node. In contrast, in this study, we ... |

20 |
Pathwise optimality and state space collapse for the exponential rule
- Shakkottai, Srikant, et al.
- 2002
(Show Context)
Citation Context ...tocol design for wireless LANs and wireless mesh networks. C. Related Work and Organization As noted above, there has been much work on centralized opportunistic scheduling (e.g., [3], [4], [5], [6], =-=[7]-=-, [8], [20]), channel-aware ALOHA (e.g., [9], [10]) and MAC design with rate adaptation (e.g., [12], [13], [14]). Most relevant to our study are perhaps (e.g., [9], [10], [12], [14]). The main differe... |

11 |
Random access in wireless ad hoc networks as a distributed game
- Borkar, Kherani
- 2004
(Show Context)
Citation Context ...wireless networks (see, e.g., [23], [24], [25], [26]). We note that a game theoretic formulation on random access protocols has been investigated, and the most relevant to our study are perhaps [27], =-=[28]-=-, [29], with one major difference being that none of these works exploit time-varying channel conditions for scheduling. The rest of the paper is organized as follows. Section II gives a brief introdu... |

9 |
Optimal stopping and applications
- Ferguson
- 2006
(Show Context)
Citation Context ...e optimal stopping rule for channel probing, in order to maximize the throughput. In this paper, we obtain a systematic characterization of this tradeoff by appealing to optimal stopping theory [15], =-=[16]-=-, and explore channelaware distributed scheduling to exploit multiuser diversity and time diversity for wireless ad-hoc networks in an opportunistic manner. We shall tackle this problem from the follo... |

9 | The effect of channel estimation error on the throughput of broadcast channels
- Vakili, Sharif, et al.
- 2006
(Show Context)
Citation Context ... great interest to study DOS under noisy channel estimation. In centralized scheduling schemes, it has been shown that the estimated signal-to-noise ratio (SNR) is always larger than the “actual SNR” =-=[6]-=-. Thus, if the data were transmitted using the estimated rate, there would always be an outage. To reduce the outage probability, the transmission rate has to back off from the estimated rate. Therefo... |

8 |
Exploiting multiuser dieversity for medium access control in wireless networks
- Qin, Berry
- 2003
(Show Context)
Citation Context ...are random access has been investigated for the uplink transmissions in a many-to-one network, where channel probing can be realized by broadcasting pilot signals from the base station. Notably, [9], =-=[10]-=- study opportunistic ALOHA under a collision model, with a basic idea being that in every slot each user transmits with a probability based on its own channel condition. While recent work [11] does no... |

8 |
Game theoretic cross-layer transmission policies in multipacket reception wireless networks
- Ngo, Krishnamurthy
- 1911
(Show Context)
Citation Context ...cheduling, assuming that each link has no knowledge of other links’ channel conditions. There has also been a surge of interest in using game theory to study wireless networks (see, e.g., [23], [24], =-=[25]-=-, [26]). We note that a game theoretic formulation on random access protocols has been investigated, and the most relevant to our study are perhaps [27], [28], [29], with one major difference being th... |

7 | Optimal random access in networks with two-way traffic
- Altman, Borkar, et al.
(Show Context)
Citation Context ...study wireless networks (see, e.g., [23], [24], [25], [26]). We note that a game theoretic formulation on random access protocols has been investigated, and the most relevant to our study are perhaps =-=[27]-=-, [28], [29], with one major difference being that none of these works exploit time-varying channel conditions for scheduling. The rest of the paper is organized as follows. Section II gives a brief i... |

6 |
Cross layer exploitation of MAC layer diversity in wireless networks
- Cao, Raghunathan, et al.
- 2006
(Show Context)
Citation Context ... separate link losses caused by fading from those by interference. That is, the PHY layer addresses the issues of fading, and the MAC layer addresses the issue of contention. However, as shown in [1] =-=[2]-=-, fading can often adversely affect the MAC layer in many realistic scenarios. The coupling between the timescales of fading and MAC calls for a unified PHY/MAC design for wireless ad-hoc networks, in... |

6 | Reverse engineering MAC
- Tang, Lee, et al.
- 2006
(Show Context)
Citation Context ...ss networks (see, e.g., [23], [24], [25], [26]). We note that a game theoretic formulation on random access protocols has been investigated, and the most relevant to our study are perhaps [27], [28], =-=[29]-=-, with one major difference being that none of these works exploit time-varying channel conditions for scheduling. The rest of the paper is organized as follows. Section II gives a brief introduction ... |

6 |
Optimal Stopping and Applications. available at http://www.math.ucla.edu/˜tom/Stopping/Contents.html
- Ferguson
- 2006
(Show Context)
Citation Context ...annel can be perfectly estimated. In [7], we have studied DOS with perfect CSI. Specifically, we have shown that the problem can be cast as a maximal rate of return problem in optimal stopping theory =-=[3]-=-, where the rate of return is the average network throughput, x, and is determined by the stopping time N: x = E[RNT] , (2) E[TN] where Tn � ∑ n j=1 Kjτ +T is the total system time including the conte... |

4 | A statistical version of prophet inequalities
- Assaf, Goldstein, et al.
- 1996
(Show Context)
Citation Context ...can be shown that the structure of the optimal scheduling strategy remains the same, except that the random “reward” Rn is replaced with its conditional expectation, ¯ Rn � E [Rn|F ′ n] [3, Page 1.3] =-=[1]-=-. More specifically, define Q ′ � {N ≥ 1 : {N = n} ∈ F ′ n , E[TN] < ∞}. (16) Analogously, define Q ′′ � {N ≥ 1 : {N = n} ∈ F ′′ n, E[TN] < ∞}, (17) where F ′′ n is the σ-field generated by {( ¯ Rj, K... |

2 | A distributed stochastic approximation approach for max-min fair rate control of flows in packet networks
- Arbraham, Kumar
- 2006
(Show Context)
Citation Context ...y ∈ [0, b]. (37) This lemma follows from the fact that the derivative of g(·) is bounded. 3) Stochastic Convergence of the Algorithm: Define the stepsize as k� ai(k) = a(i, I{k ∈ N i }). l=1 Based on =-=[32]-=-, [19], we impose the following conditions. B1) The sequence {a(i, k)} satisfy ∞� ∞� a(i, k) = ∞, and a(i, k) 2 < ∞. k=1 and for β ∈ (0, 1), ∀ i, j lim k→∞ �⌊βk⌋ l=1 �k l=1 a(i, l) k=1 a(i, l) = 1, an... |