Performance of Round Robin Policies for Dynamic Multichannel Access

Size: px
Start display at page:

Download "Performance of Round Robin Policies for Dynamic Multichannel Access"

Transcription

1 Performance of Round Robin Policies for Dynamic Multichannel Access Changmian Wang, Bhaskar Krishnamachari, Qing Zhao and Geir E. Øien Norwegian University of Science and Technology, Norway, {changmia, University of Southern California, USA, University of California, Davis, USA, Abstract We consider two simple round-robin sensing policies for dynamic multi-channel access in cognitive radio networks one in which channel switching takes place when the primary user is sensed to be, and one in which a channel switching takes place when the primary user is sensed to be sent. Prior work has shown that these policies are each optimal under certain conditions when the primary user occupancy on each channel can be described as an independent two-state Markov chain. In this work, we consider a very general case where the primary user occupancy on each channel is an arbitrary stationary and ergodic two-state process, and derive bounds on their performance. The bounds provide insights into conditions under which these extremely simple policies perform well. I. INTRODUCTION Dynamic spectrum access techniques aim to improve the efficiency of radio-frequency utilization for wireless communications. In these techniques, cognitive secondary users (users that do not have priority access to spectrum) attempt to identify and exploit the opportunities for transmission that arise whenever and wherever primary users (that own or have priority access to spectrum) are not active. We consider the following simple but fundamental dynamic multichannel access problem. There is a single backlogged secondary user. Time is discretized into slots. At the beginning of each slot, the user picks one of multiple orthogonal channels to sense. If that channel is availle (i.e., no primary user activity is detected), the user may proceed to use it, else it has to wait out the rest of the slot. The secondary user must employ some policy to select a channel at each time, taking into account its past observations, so as to maximize its expected long-term throughput. This problem has been considered before in prior work. However, in most of these, the occupancy of the primary user on the channels has been modeled as being Markovian. Under such an assumption, this problem can be formulated as a partially observle Markov decision process (POMDP) []. In the particular case when the primary user occupancy over each channel can be modeled as an independent and identically distributed (i.i.d.) two-state Markov chain, it was shown by Zhao, Krishnamachari and Liu [2] that the myopic policy, in which the secondary user picks the channel that maximizes the immediate reward has a semi-universal structure that obviates the need to know the details of the one-step transition probility matrix. In particular, when the Markov chain describing the primary user occupancy for each channel is positively correlated, the myopic policy is for the secondary user to stick with a channel until it is observed to be busy, in which case it must switch to the channel that was observed the longest ago from the next slot on. When the Markov chain is negatively correlated, the myopic policy is for the secondary user to stick with a channel until it is observed to be free, then switching to a either the channel that was observed the longest ago or the most recent previous channel, depending on whether an odd or even number of steps have elapsed since the last channel switch. It has been shown in [2] that the myopic policy achieves optimality for the two-channel case regardless of the sign of the correlation. Building on this work, Ahmad et al. [3] show that for i.i.d. Markov chains the myopic policy is in fact optimal for any number of channels if they are positively correlated; it is also optimal for 3 channels if they are negatively correlated; however, for 4 channels it is not always optimal. Other extensions of these works have considered imperfect sensing [4], and simultaneous multichannel sensing [5], [6]. Unlike these prior works, we relax in this paper the requirement that the primary user behavior be Markovian. Instead, we consider a more general case when the primary user behavior on each channel can be modeled as an independent stationary and ergodic binary process. We consider two simple round robin policies that are related to the ove-mentioned myopic policy for the case of i.i.d. Markov channels, and develop bounds for the performance of these policies for the general settings considered here. II. CHANNEL MODEL There are N channels, with each channel i described by an independent two-state stationary ergodic process S (i) (n), where n is the time index. The two states can be thought of as the busy () and free () states reflecting the occupancy of the primary user. We define the probility of a state a on channel i as a and the probility of seeing state a followed state b, i.e. P{S (i) (n) = a, S (i) (n + ) = b}, as. We define the following probility mass function (p.m.f.) for n : h (i) (n) = P {S(i) (m + ) = a, S (i) (m + 2) = a,..., S (i) (m + n ) = a, S (i) (m + n) = b S (i) (m ) = b, S (i) (m) = a} ()

2 Channel Channel Fig.. Illustration of the Zero-Switch Round Robin Policy and the One-Switch Round Robin Policy which is the probility distribution of how long the process shall stay in one state immediately after a transition before making its next transition. Then the average time the process stays in a certain state is thus given by: a = {n h (i) (n)} (2) n= Let F (i) (n) be the cumulative distribution function of h(i) (n) and c F (i) (n) be the corresponding complementary cumulative distribution function. We can also define the following p.m.f. for n : rh (i) (n) = P {S(i) (m + ) = a, S (i) (m + 2) = a,..., S (i) (m + n ) = a, S (i) (m + n) = b S (i) (m) = a} (3) which is the probility distribution of the duration that the process shall stay in one state before making its next transition if the process is entered at a random point with an observation of state a. We denote the corresponding cumulative distribution function and complementary cumulative distribution function as r F (i) (n) and c rf (i) (n) respectively. III. ROUND ROBIN POLICIES We assume the secondary user can perform perfect sensing of one channel at each time. We consider two intuitive round-robin channel sensing and access policies for the secondary user: Zero-switch round robin policy (ZSRRP): in this policy, the secondary user stays on a given channel repeatedly accessing it on each round until it observes a zero, then switches to the channel observed the longest ago. One-switch round robin policy (OSRRP): in this policy, the secondary user stays on a given channel until it observes a one, then switches to the channel observed the longest ago. Note that ZSRRP is identical to the myopic policy for positively correlated i.i.d. Markov chains [2], while the OSRRP is a variant of the myopic policy for negatively correlated i.i.d. Markov chains but coincides with it when there are only two channels. Thus the policies we explore in this work are closely related to strategies that have been shown to be optimal for Markov chains under certain conditions in prior work [3]. Channel Fig. 2. previous previous The two error patterns for 2 channels for ZSRRP IV. BOUNDS ON REGRET Let E[T π ] be the expected average throughput for some given policy π. Further, let E[T genie ] be the expected average throughput that can be achieved by an omniscient secondary user, i.e., one that can always pick a free channel at each time slot if one is availle. We define the expected regret E[R π ] as follows: E[R π ] = E[T genie ] E[T π ] (4) Thus, if we can find an upper bound on the regret for a given policy E[R + π ] E[R π ], it can be used to obtain a lower bound on that policy s expected average throughput, as follows: E[T π ] E[T genie ] E[R + π ] (5) We first develop bounds on the regret for both ZSRRP and OSRRP for the case of two channels. The crux of our approach is to identify certain error patterns for each policy, conditions under which they may possibly be misled into missing an opportunity for transmission. We begin with rough bounds, which have the advantage of simplicity in that they consider only two consecutive time slots. We then show how these bounds may be improved by considering corrections that take into account more slots. A. Bounds for Zero-Switch Round Robin Policy Figure 2 shows two error patterns for the ZSRRP policy. The horizontal lines separate the two channels, and the vertical lines separate two adjacent time slots. The symbols showing the state of each channel in a given slot are,, or to indicate a don t care (either or ). The small x on one of the channels in each slot indicates the channel selection and observation made by the secondary user following the ZSRRP policy. Consider the first pattern. In this case, the secondary user is on the first channel at the first time slot, and observes a. Regardless of what happens on the second channel at the first time slot, the ZSRRP policy requires the secondary user to stick with the first channel in the second time slot. However,

3 previous previous Channel Fig. 3. The two error patterns for 2 channels for OSRRP in the second slot there is a on the first channel, and a on the second channel. Thus, the ZSRRP policy misses an opportunity that is availle to the genie. Similarly, in the second error pattern, the secondary user starts from the second channel, but observes a there, causing it to switch to the first channel in the second slot, which has a, so it again misses an opportunity on the second channel, which has a in this second time slot. Not shown are the two symmetric versions of these patterns, with the channel lels reversed. Note that these error patterns must occur for the ZSRRP policy to have missed an opportunity availle to the genie. Therefore quantifying the probility of the occurrence of these patterns yields an upper bound on the regret of the ZSRRP. In the first pattern, in channel, a transition occurs, which can be characterized by e () as defined on the ove. Since the channels are mutually independent and the probility of channel 2 being in state is given by φ (2) based on ove definitions, the overall probility for this pattern to occur is characterized by: e () φ(2) (6) The probility of the second pattern is, similarly: e (2) φ() (7) Equivalent expressions are obtained for the symmetric cases where the channel lels are reversed. The rough upper bound on regret for 2 channels that is thus obtained is given by : E[R + ZSRRP ] = i= It is not hard to see the following fact: Further, φ(i+) + e(i+) e(i+) e(i+) (8) = e(i) = + (9) and φ(i), which denote the probility of seeing a and respectively, can also be expressed in terms of the expect length that the process can stay in each state as follows: We apologize for the use of notation in referring to channel numbers; (i+) res the next channel in the round-robin schedule, allowing for wrap-around. Thus, for two-channels, if i = 2, i + =. = = + + () Using the ove observations, we get the following simplified expression for the bound on ZSRRP regret when the channels are i.i.d.: E[R + ZSRRP ] = 2e φ + 2e φ 2e e = 2 τ + τ (τ + τ ) 2 () Equation () shows that the upper bound on the regret for ZSRRP is inversely proportional to the sum of the expected time spent at each state. This intuitive result suggests that this policy performs well when channel states are positively correlated, which is very much in keeping with the previous findings in the literature ([2], [3]) that it is in fact optimal for positively correlated Markovian channels. B. Bounds for One-Switch Round Robin Policy Figure 3 shows the error patterns under which the OSRRP policy misses opportunities. In the first cases, OSRRP causes the user to switch out from a channel because it observes a, and then miss an opportunity in the next time slot. In the second case, OSRRP causes the user to remain on the same channel because it observes a, and then miss an opportunity in the next time slot. Quantifying the probility of these patterns and their symmetric counterparts (with the channel lels switched) yields a corresponding rough upper bound on the regret of OSRRP: E[R + OSRRP ] = i= e(i+) + e(i+) e(i) (2) Which, when the channels are i.i.d., becomes: E[R + OSRRP ] = 2(e φ + e φ e e ) (3) V. IMPROVED UPPER BOUND ON REGRET We now develop improved upper bounds on the regret for N = 2 by ruling out certain cases when the error patterns described in the previous section are not encountered by the respective policies. A. Improved Bound for ZSRRP To develop improved bounds, we make use of the following straightforward observations out the two policies: Observation : For N = 2 and ZSRRP, a sufficient condition to have channel i selected at time slot n is that at time slot n the other channel was in state and channel i was in state. Observation 2: For N = 2 and OSRRP, the sufficient condition for channel i to be selected at time slot n is

4 Channel Channel Fig. 4. Correct patterns for error pattern for ZSRRP Fig. 6. Correct patterns for error pattern for OSRRP Channel Channel Fig. 7. Correct patterns for error pattern for OSRRP Fig. 5. Correct patterns for error pattern 2 for ZSRRP that at time slot n the other channel was in state and channel i was in state. The bounds derived in the previous section are loose because they quantify the probility of occurrence of error patterns that are not always encountered by the corresponding policy. For example, consider the first error pattern in figure 2. ZSRRP misses an opportunity in the second time slot if the use is in channel at the first time slot. However, consider the scenario depicted on the left part of figure 4. Although the pattern for the second and third slot in this case match error pattern for ZSRRP, because of observation ove, the secondary user following this policy must pick channel 2 in the second slot, not channel, and in doing so does not miss the opportunity in the third slot. Thus even though the two channels have the states indicated in error pattern, a user following ZSRRP does not encounter the pattern and lose an opportunity by it. There are in fact infinitely many such correct patterns, all starting with the same suffix, as shown in figure 4. Similarly, in figure 5, we see that the symmetric version of error pattern 2 occurs (there is a - transition on channel on the second and third slots and a on channel 2 in the third slot), but the user is again constrained to be on the second channel in the second slot because of the channel states in the first time slot, again, as per observation. And again, there is an infinite string of other related longer correct patterns as indicated. Taking these correct patterns into account (i.e., subtracting their probility of occurrence from the probility of occurrence of the basic error patterns) improves the bound on regret to the following: E[R +I ZSRRP ] = E[R+ ZSRRP ] T T 2 T 3 (4) where T = T 2 = T 3 = i= k= i= k= i= k= B. Improved Bound for OSRRP φ (i+) h(i) (k)c rf (i+) (k + ) e(i+) rh (i) (2k + 2) c F (i+) (2k + ) φ (i+) h(i) (2k + 2) c rf (i+) (2k + 3) The improved bound for OSRRP is derived in a similar way as it was done for ZSRRP. The correct patterns in this case for the two error patterns in figure 3 are as depicted by figure 6 and figure 7. We the three correcting terms in this case: S = S 2 = S 3 = i= k= i= k= φ (i+) h(i) (2k + ) c rf (i+) (2k + 2), e(i+) rh (i) (2k + ) c F (i+) (2k), e(i+) c F (i+) i= k= The corresponding improved bound is: (k) r h (i) (k + ). E[R +I OSRRP ] = E[R+ OSRRP ] S S 2 S 3 (5) VI. SIMULATIONS In this section, we evaluate the performance of ZSRRP and OSRRP policies on two-channel systems through simulations and compare them with the bounds derived in the previous sections. For the simulations, we assume that the channels are i.i.d. and the primary occupancy process in each channel can be described as a two-state semi-markov process. We assume that the holding time for state follows a Zipf distribution and that the holding time for state follows a geometric distribution. The p.m.f of a Zipf distribution is given as: p Zipf (n) = n a ζ, (6)

5 TABLE I ACTUAL ACHIEVED THROUGHPUT FOR ZSRRP τ τ =.2 τ =.9 τ =.32 τ =.6 τ = TABLE II ACTUAL ACHIEVED THROUGHPUT FOR OSRRP τ τ =.2 τ =.9 τ =.32 τ =.6 τ = τ s Intersection Contour under Two Policies OSRRP Dominant Region Actual Achieved Performance Rough Bound Improved Bound Markov Channel Model ZSRRP Dominant Region where a is the parameter for Zipf distribution and the ζ function is defined as ζ = i= i. The mean of Zipf a distribution is finite only if a > 2, and it is given by: E{n} = ζ(a ), a > 2. (7) ζ The expected holding time for state can thus be computed according to equation (7). In general, the holding time distribution for a semi-markov process with random starting point is related to the complementary cumulative holding time distribution, by rh sis j (n) = c F sis j (n ) τ sis j. (8) The complementary cumulative holding time distribution with a random start, which is denoted by c rf sis j can be derived as: c rf sis j (n) = n rh sis j (m) (9) m= = n F sis τ j (m ). sis j (2) m= Note that e and e can be expressed by means of stationary distribution φ and the random-start complementary cumulative distribution function c rf (n) as: e = φ c r F () (2) e = φ c r F () (22) In figures 8 and 9, we simulate the actual achieved throughput under ZSRRP and OSRRP and compare them to the various bounds. For both plots, the simulation parameters are as follows. The average holding time for state, τ is set to two different values in each: τ =.2 and τ = 8. And the a parameter for the Zipf holding time distribution is varied from 2.5 to 3., which means by equation (7), τ varies from 2.75 to.37. We can see that the improved bound in general approximate the actual achieved performance well in both cases. Both policies perform worse as a increases, this is Fig.. τ Contour of Intersection Points under Two Policies because as the parameter a gets larger, τ gets smaller compare to τ, reducing the throughput. It is also of interest to know which of the two policies shall perform better under a given circumstance. This can be done by comparing Tle I and Tle II, where we vary the average holding time τ and τ for and process respectively. From the tles, it can be seen that ZSRRP outperforms OSRRP on the upper-right corner of the matrix, while OSRRP is a better algorithm for the lower-left corner of the matrix. In addition, we also plot the contour when the throughput achieved by the two policies cross over, in figure. In particular, we plot the contour under three cases, i.e. the simulated achieved throughput, the throughput indicated by the rough as well as improved bounds ed in the previous two sections. The contours showing the crossover of lower bounds on throughput of the two policies of course is not rigorously an upper or lower bound on the true contour, but nevertheless could be seen as a way to approximate the cross-over contour. Surprisingly, the contour for all the three cases closely match each other, which implies that even the simple rough bounds can help determine the dominant regions for the two policies. In a conventional two-state Markov channel model, it has been proven in [2] that for N = 2, ZSRRP is optimal if p > p and OSRRP is optimal if p < p. Hence it is not hard to see the decision boundary is when p = p, which implies p + p =, and hence we have τ + τ =. (23) Equation (23) is hence the dominant decision region boundary for a Markov model in the case of two channels. This boundary is also shown in figure as a reference. The reason why the lower part of the simulated boundary is closer to the conventional Markov boundary is because as τ gets smaller, the holding time distribution for state becomes close to

6 .9.8 Simulation Results for two i.i.d. semi Markov Channels (τ =.2).9.8 Simulation Results for two i.i.d. semi Markov Channels (τ = 8) Achieved by Genie. Actual Achieved by ZSRRP. Improved Achieved by Genie. Actual Achieved by ZSRRP. Improved.3.2. Fig. 8. ZSRRP bounds and simulated throughput for τ =.2 and τ = Simulation Results for two i.i.d. semi Markov Channels (τ =.2).9.8 Simulation Results for two i.i.d. semi Markov Channels (τ = 8) Achieved by Genie. Actual Achieved by OSRRP. Improved Achieved by Genie. Actual Achieved by OSRRP. Improved.3.2. Fig. 9. OSRRP bounds and simulated throughput for τ =.2 and τ = 8 Fig.. The error patterns for N > 2 under ZSRRP and OSRRP

7 the geometric distribution, so that the semi-markov process behaves increasingly like a conventional Markov process. VII. BOUNDS ON REGRET FOR N > 2 Figure shows the error patterns for ZSRRP when there are more than 2 channels. The situation is quite similar to the N = 2 case, but there are some minor differences. In particular, in case of the second pattern the relevant event probility calculations must differentiate between a appearing on the channel that was andoned, and a appearing elsewhere. The calculations yield the following general bound on regret for ZSRRP for N > 2 channels: where L = L 2 = L 3 = E[R + ZSRRP ] = L + L 2 + L 3 (24) i= i= i= [ N φ(i+), j=,j i φ(i+) ( ], N j=,j i,i+ ) (25) Figure shows the error patterns for OSRRP for N > 2. The corresponding bound on the regret is: where M = M 2 = M 3 = E[R + OSRRP ] = M + M 2 + M 3, (26) i= i= i= [ N j=,j i φ(i+) [ ], N j=,j i,i+ ], φ(i+) (27) It should be noted that, in deriving improved bounds for the N = 2 case, we used the sufficient condition for channel selection. Such techniques can also be applied to improve the bound for N > 2; however, the expressions get more complicated as N becomes large. We omit these detailed calculations. of these policies with respect to an oracle that is aware of all channel realizations in advance using certain characteristic error patterns. The derived bounds are insightful in showing the conditions under which either policy results in low regret. Further, by comparing the bounds for both policies, we are le to gain some insight into the conditions under which one policy outperforms the other. Our results in this paper suggest that these two simple policies may be quite efficient for dynamic multichannel access in practice when the channels are identical. An interesting question to be explored in the future is whether hybrid policies may be similarly useful for the case of non-identical channels (for instance, when some are positively correlated and others are negatively correlated). ACKNOWLEDGMENT The work of C. Wang and G. E. Øien was supported by the NORDITE/NFR(VERDIKT) project CROPS. The work of Q. Zhao was supported by the U.S. Army Research Office under Grant W9NF REFERENCES [] Q. Zhao, L. Tong, A. Swami, and Y. Chen, Decentralized Cognitive MAC for Opportunistic Spectrum Access in Ad Hoc Networks: A POMDP Framework, in IEEE Journal on Selected Areas in Communications (JSAC): Special Issue on Adaptive, Spectrum Agile and Cognitive Wireles Networks, vol. 25, no. 3, pp , April 27. [2] Q. Zhao, B. Krishnamachari, and K. Liu, On Myopic Sensing for Multi- Channel Opportunistic Access: Structure, Optimality, and Performance, IEEE Transactions on Wireless Communications, vol. 7, no. 2, part 2, December 28. [3] S. H. Ahmad, M. Liu, T. Javidi, Q. Zhao and B. Krishnamachari, Optimality of Myopic Sensing in Multi-Channel Opportunistic Access, IEEE Trans. on Information Theory, vol. 55, no. 9, pp , September 29. [4] K. Liu, Q. Zhao, and B. Krishnamachari, Dynamic Multichannel Access with Imperfect Channel State Detection IEEE Transactions on Signal Processing, to appear. [5] K. Liu and Q. Zhao, Channel Probing for Opportunistic Access with Multi-channel Sensing, in Proc. of IEEE Asilomar Conference on Signals, Systems, and Computers, October, 28. [6] S. Ahmad and M. Liu, Multi-channel opportunistic access: a case of restless bandits with multiple plays, in Allerton Conference, October 29, Allerton, IL. VIII. CONCLUSION We have considered the problem of a single secondary user selecting which of multiple channels to access at each time to maximize its expected throughput in the presence of stochastic primary user activity. Unlike most of the prior work that has focused on Markovian primary users, we have developed bounds on the performance of two simple policies when the primary traffic on each channel can be modeled as a more general independent stationary and ergodic two-state process. The crux of our technique is to bound the regret

On the Optimality of Myopic Sensing. in Multi-channel Opportunistic Access: the Case of Sensing Multiple Channels

On the Optimality of Myopic Sensing. in Multi-channel Opportunistic Access: the Case of Sensing Multiple Channels On the Optimality of Myopic Sensing 1 in Multi-channel Opportunistic Access: the Case of Sensing Multiple Channels Kehao Wang, Lin Chen arxiv:1103.1784v1 [cs.it] 9 Mar 2011 Abstract Recent works ([1],

More information

STRUCTURE AND OPTIMALITY OF MYOPIC SENSING FOR OPPORTUNISTIC SPECTRUM ACCESS

STRUCTURE AND OPTIMALITY OF MYOPIC SENSING FOR OPPORTUNISTIC SPECTRUM ACCESS STRUCTURE AND OPTIMALITY OF MYOPIC SENSING FOR OPPORTUNISTIC SPECTRUM ACCESS Qing Zhao University of California Davis, CA 95616 qzhao@ece.ucdavis.edu Bhaskar Krishnamachari University of Southern California

More information

Dynamic spectrum access with learning for cognitive radio

Dynamic spectrum access with learning for cognitive radio 1 Dynamic spectrum access with learning for cognitive radio Jayakrishnan Unnikrishnan and Venugopal V. Veeravalli Department of Electrical and Computer Engineering, and Coordinated Science Laboratory University

More information

Optimality of Myopic Sensing in Multi-Channel Opportunistic Access

Optimality of Myopic Sensing in Multi-Channel Opportunistic Access Optimality of Myopic Sensing in Multi-Channel Opportunistic Access Tara Javidi, Bhasar Krishnamachari, Qing Zhao, Mingyan Liu tara@ece.ucsd.edu, brishna@usc.edu, qzhao@ece.ucdavis.edu, mingyan@eecs.umich.edu

More information

Power Allocation over Two Identical Gilbert-Elliott Channels

Power Allocation over Two Identical Gilbert-Elliott Channels Power Allocation over Two Identical Gilbert-Elliott Channels Junhua Tang School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University, China Email: junhuatang@sjtu.edu.cn Parisa

More information

Multi-Armed Bandit: Learning in Dynamic Systems with Unknown Models

Multi-Armed Bandit: Learning in Dynamic Systems with Unknown Models c Qing Zhao, UC Davis. Talk at Xidian Univ., September, 2011. 1 Multi-Armed Bandit: Learning in Dynamic Systems with Unknown Models Qing Zhao Department of Electrical and Computer Engineering University

More information

Online Learning to Optimize Transmission over an Unknown Gilbert-Elliott Channel

Online Learning to Optimize Transmission over an Unknown Gilbert-Elliott Channel Online Learning to Optimize Transmission over an Unknown Gilbert-Elliott Channel Yanting Wu Dept. of Electrical Engineering University of Southern California Email: yantingw@usc.edu Bhaskar Krishnamachari

More information

Optimality of Myopic Sensing in Multi-Channel Opportunistic Access

Optimality of Myopic Sensing in Multi-Channel Opportunistic Access Optimality of Myopic Sensing in Multi-Channel Opportunistic Access Tara Javidi, Bhasar Krishnamachari,QingZhao, Mingyan Liu tara@ece.ucsd.edu, brishna@usc.edu, qzhao@ece.ucdavis.edu, mingyan@eecs.umich.edu

More information

Channel Probing in Communication Systems: Myopic Policies Are Not Always Optimal

Channel Probing in Communication Systems: Myopic Policies Are Not Always Optimal Channel Probing in Communication Systems: Myopic Policies Are Not Always Optimal Matthew Johnston, Eytan Modiano Laboratory for Information and Decision Systems Massachusetts Institute of Technology Cambridge,

More information

Multi-channel Opportunistic Access: A Case of Restless Bandits with Multiple Plays

Multi-channel Opportunistic Access: A Case of Restless Bandits with Multiple Plays Multi-channel Opportunistic Access: A Case of Restless Bandits with Multiple Plays Sahand Haji Ali Ahmad, Mingyan Liu Abstract This paper considers the following stochastic control problem that arises

More information

Optimality of Myopic Sensing in Multichannel Opportunistic Access

Optimality of Myopic Sensing in Multichannel Opportunistic Access 4040 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 55, NO. 9, SEPTEMBER 2009 Optimality of Myopic Sensing in Multichannel Opportunistic Access Sahand Haji Ali Ahmad, Mingyan Liu, Member, IEEE, Tara Javidi,

More information

A POMDP Framework for Cognitive MAC Based on Primary Feedback Exploitation

A POMDP Framework for Cognitive MAC Based on Primary Feedback Exploitation A POMDP Framework for Cognitive MAC Based on Primary Feedback Exploitation Karim G. Seddik and Amr A. El-Sherif 2 Electronics and Communications Engineering Department, American University in Cairo, New

More information

OPPORTUNISTIC Spectrum Access (OSA) is emerging

OPPORTUNISTIC Spectrum Access (OSA) is emerging Optimal and Low-complexity Algorithms for Dynamic Spectrum Access in Centralized Cognitive Radio Networks with Fading Channels Mario Bkassiny, Sudharman K. Jayaweera, Yang Li Dept. of Electrical and Computer

More information

A State Action Frequency Approach to Throughput Maximization over Uncertain Wireless Channels

A State Action Frequency Approach to Throughput Maximization over Uncertain Wireless Channels A State Action Frequency Approach to Throughput Maximization over Uncertain Wireless Channels Krishna Jagannathan, Shie Mannor, Ishai Menache, Eytan Modiano Abstract We consider scheduling over a wireless

More information

Opportunistic Spectrum Access for Energy-Constrained Cognitive Radios

Opportunistic Spectrum Access for Energy-Constrained Cognitive Radios 1206 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 3, MARCH 2009 Opportunistic Spectrum Access for Energy-Constrained Cognitive Radios Anh Tuan Hoang, Ying-Chang Liang, David Tung Chong Wong,

More information

Cognitive Spectrum Access Control Based on Intrinsic Primary ARQ Information

Cognitive Spectrum Access Control Based on Intrinsic Primary ARQ Information Cognitive Spectrum Access Control Based on Intrinsic Primary ARQ Information Fabio E. Lapiccirella, Zhi Ding and Xin Liu Electrical and Computer Engineering University of California, Davis, California

More information

A Restless Bandit With No Observable States for Recommendation Systems and Communication Link Scheduling

A Restless Bandit With No Observable States for Recommendation Systems and Communication Link Scheduling 2015 IEEE 54th Annual Conference on Decision and Control (CDC) December 15-18, 2015 Osaka, Japan A Restless Bandit With No Observable States for Recommendation Systems and Communication Link Scheduling

More information

Distributed Reinforcement Learning Based MAC Protocols for Autonomous Cognitive Secondary Users

Distributed Reinforcement Learning Based MAC Protocols for Autonomous Cognitive Secondary Users Distributed Reinforcement Learning Based MAC Protocols for Autonomous Cognitive Secondary Users Mario Bkassiny and Sudharman K. Jayaweera Dept. of Electrical and Computer Engineering University of New

More information

Wireless Channel Selection with Restless Bandits

Wireless Channel Selection with Restless Bandits Wireless Channel Selection with Restless Bandits Julia Kuhn and Yoni Nazarathy Abstract Wireless devices are often able to communicate on several alternative channels; for example, cellular phones may

More information

Algorithms for Dynamic Spectrum Access with Learning for Cognitive Radio

Algorithms for Dynamic Spectrum Access with Learning for Cognitive Radio Algorithms for Dynamic Spectrum Access with Learning for Cognitive Radio Jayakrishnan Unnikrishnan, Student Member, IEEE, and Venugopal V. Veeravalli, Fellow, IEEE 1 arxiv:0807.2677v2 [cs.ni] 21 Nov 2008

More information

Optimal and Suboptimal Policies for Opportunistic Spectrum Access: A Resource Allocation Approach

Optimal and Suboptimal Policies for Opportunistic Spectrum Access: A Resource Allocation Approach Optimal and Suboptimal Policies for Opportunistic Spectrum Access: A Resource Allocation Approach by Sahand Haji Ali Ahmad A dissertation submitted in partial fulfillment of the requirements for the degree

More information

arxiv:cs/ v1 [cs.ni] 27 Feb 2007

arxiv:cs/ v1 [cs.ni] 27 Feb 2007 Joint Design and Separation Principle for Opportunistic Spectrum Access in the Presence of Sensing Errors Yunxia Chen, Qing Zhao, and Ananthram Swami Abstract arxiv:cs/0702158v1 [cs.ni] 27 Feb 2007 We

More information

Near-optimal policies for broadcasting files with unequal sizes

Near-optimal policies for broadcasting files with unequal sizes Near-optimal policies for broadcasting files with unequal sizes Majid Raissi-Dehkordi and John S. Baras Institute for Systems Research University of Maryland College Park, MD 20742 majid@isr.umd.edu, baras@isr.umd.edu

More information

Optimal Sensing-Transmission Structure for Dynamic Spectrum Access

Optimal Sensing-Transmission Structure for Dynamic Spectrum Access Optimal Sensing-Transmission Structure for Dynamic Spectrum Access Senhua Huang, Xin Liu, and Zhi Ding University of California, Davis, CA 9566, USA Abstract In cognitive wireless networks where secondary

More information

Channel Allocation Using Pricing in Satellite Networks

Channel Allocation Using Pricing in Satellite Networks Channel Allocation Using Pricing in Satellite Networks Jun Sun and Eytan Modiano Laboratory for Information and Decision Systems Massachusetts Institute of Technology {junsun, modiano}@mitedu Abstract

More information

arxiv: v3 [cs.lg] 4 Oct 2011

arxiv: v3 [cs.lg] 4 Oct 2011 Reinforcement learning based sensing policy optimization for energy efficient cognitive radio networks Jan Oksanen a,, Jarmo Lundén a,b,1, Visa Koivunen a a Aalto University School of Electrical Engineering,

More information

P e = 0.1. P e = 0.01

P e = 0.1. P e = 0.01 23 10 0 10-2 P e = 0.1 Deadline Failure Probability 10-4 10-6 10-8 P e = 0.01 10-10 P e = 0.001 10-12 10 11 12 13 14 15 16 Number of Slots in a Frame Fig. 10. The deadline failure probability as a function

More information

Optimal Power Allocation Policy over Two Identical Gilbert-Elliott Channels

Optimal Power Allocation Policy over Two Identical Gilbert-Elliott Channels Optimal Power Allocation Policy over Two Identical Gilbert-Elliott Channels Wei Jiang School of Information Security Engineering Shanghai Jiao Tong University, China Email: kerstin@sjtu.edu.cn Junhua Tang

More information

TRANSMISSION STRATEGIES FOR SINGLE-DESTINATION WIRELESS NETWORKS

TRANSMISSION STRATEGIES FOR SINGLE-DESTINATION WIRELESS NETWORKS The 20 Military Communications Conference - Track - Waveforms and Signal Processing TRANSMISSION STRATEGIES FOR SINGLE-DESTINATION WIRELESS NETWORKS Gam D. Nguyen, Jeffrey E. Wieselthier 2, Sastry Kompella,

More information

Introducing strategic measure actions in multi-armed bandits

Introducing strategic measure actions in multi-armed bandits 213 IEEE 24th International Symposium on Personal, Indoor and Mobile Radio Communications: Workshop on Cognitive Radio Medium Access Control and Network Solutions Introducing strategic measure actions

More information

Online Learning with Randomized Feedback Graphs for Optimal PUE Attacks in Cognitive Radio Networks

Online Learning with Randomized Feedback Graphs for Optimal PUE Attacks in Cognitive Radio Networks 1 Online Learning with Randomized Feedback Graphs for Optimal PUE Attacks in Cognitive Radio Networks Monireh Dabaghchian, Amir Alipour-Fanid, Kai Zeng, Qingsi Wang, Peter Auer arxiv:1709.10128v3 [cs.ni]

More information

Opportunistic Channel-Aware Spectrum Access for Cognitive Radio Networks with Interleaved Transmission and Sensing

Opportunistic Channel-Aware Spectrum Access for Cognitive Radio Networks with Interleaved Transmission and Sensing 1 Opportunistic Channel-Aware Spectrum Access for Cognitive Radio Networks with Interleaved Transmission and Sensing Sheu-Sheu Tan, James Zeidler, and Bhaskar Rao Department of Electrical and Computer

More information

Scheduling in parallel queues with randomly varying connectivity and switchover delay

Scheduling in parallel queues with randomly varying connectivity and switchover delay Scheduling in parallel queues with randomly varying connectivity and switchover delay The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters.

More information

Distributed Stochastic Online Learning Policies for Opportunistic Spectrum Access

Distributed Stochastic Online Learning Policies for Opportunistic Spectrum Access 1 Distributed Stochastic Online Learning Policies for Opportunistic Spectrum Access Yi Gai, Member, IEEE, and Bhaskar Krishnamachari, Member, IEEE, ACM Abstract The fundamental problem of multiple secondary

More information

Bayesian Congestion Control over a Markovian Network Bandwidth Process: A multiperiod Newsvendor Problem

Bayesian Congestion Control over a Markovian Network Bandwidth Process: A multiperiod Newsvendor Problem Bayesian Congestion Control over a Markovian Network Bandwidth Process: A multiperiod Newsvendor Problem Parisa Mansourifard 1/37 Bayesian Congestion Control over a Markovian Network Bandwidth Process:

More information

Combinatorial Network Optimization With Unknown Variables: Multi-Armed Bandits With Linear Rewards and Individual Observations

Combinatorial Network Optimization With Unknown Variables: Multi-Armed Bandits With Linear Rewards and Individual Observations IEEE/ACM TRANSACTIONS ON NETWORKING 1 Combinatorial Network Optimization With Unknown Variables: Multi-Armed Bandits With Linear Rewards and Individual Observations Yi Gai, Student Member, IEEE, Member,

More information

Learning Algorithms for Minimizing Queue Length Regret

Learning Algorithms for Minimizing Queue Length Regret Learning Algorithms for Minimizing Queue Length Regret Thomas Stahlbuhk Massachusetts Institute of Technology Cambridge, MA Brooke Shrader MIT Lincoln Laboratory Lexington, MA Eytan Modiano Massachusetts

More information

Markovian Decision Process (MDP): theory and applications to wireless networks

Markovian Decision Process (MDP): theory and applications to wireless networks Markovian Decision Process (MDP): theory and applications to wireless networks Philippe Ciblat Joint work with I. Fawaz, N. Ksairi, C. Le Martret, M. Sarkiss Outline Examples MDP Applications A few examples

More information

Optimal Channel Probing and Transmission Scheduling for Opportunistic Spectrum Access

Optimal Channel Probing and Transmission Scheduling for Opportunistic Spectrum Access Optimal Channel Probing and Transmission Scheduling for Opportunistic Spectrum Access ABSTRACT Nicholas B. Chang Department of Electrical Engineering and Computer Science University of Michigan Ann Arbor,

More information

NOWADAYS the demand to wireless bandwidth is growing

NOWADAYS the demand to wireless bandwidth is growing 1 Online Learning with Randomized Feedback Graphs for Optimal PUE Attacks in Cognitive Radio Networks Monireh Dabaghchian, Student Member, IEEE, Amir Alipour-Fanid, Student Member, IEEE, Kai Zeng, Member,

More information

Bayesian Congestion Control over a Markovian Network Bandwidth Process

Bayesian Congestion Control over a Markovian Network Bandwidth Process Bayesian Congestion Control over a Markovian Network Bandwidth Process Parisa Mansourifard 1/30 Bayesian Congestion Control over a Markovian Network Bandwidth Process Parisa Mansourifard (USC) Joint work

More information

Quickest Anomaly Detection: A Case of Active Hypothesis Testing

Quickest Anomaly Detection: A Case of Active Hypothesis Testing Quickest Anomaly Detection: A Case of Active Hypothesis Testing Kobi Cohen, Qing Zhao Department of Electrical Computer Engineering, University of California, Davis, CA 95616 {yscohen, qzhao}@ucdavis.edu

More information

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 4, APRIL

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 4, APRIL IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 4, APRIL 2011 731 Distributed Algorithms for Learning and Cognitive Medium Access with Logarithmic Regret Animashree Anandkumar, Member, IEEE,

More information

Algorithms for Dynamic Spectrum Access with Learning for Cognitive Radio

Algorithms for Dynamic Spectrum Access with Learning for Cognitive Radio Algorithms for Dynamic Spectrum Access with Learning for Cognitive Radio Jayakrishnan Unnikrishnan, Student Member, IEEE, and Venugopal V. Veeravalli, Fellow, IEEE 1 arxiv:0807.2677v4 [cs.ni] 6 Feb 2010

More information

On the Stability and Optimal Decentralized Throughput of CSMA with Multipacket Reception Capability

On the Stability and Optimal Decentralized Throughput of CSMA with Multipacket Reception Capability On the Stability and Optimal Decentralized Throughput of CSMA with Multipacket Reception Capability Douglas S. Chan Toby Berger Lang Tong School of Electrical & Computer Engineering Cornell University,

More information

Online Learning of Rested and Restless Bandits

Online Learning of Rested and Restless Bandits Online Learning of Rested and Restless Bandits Cem Tekin, Mingyan Liu Department of Electrical Engineering and Computer Science University of Michigan, Ann Arbor, Michigan, 48109-2122 Email: {cmtkn, mingyan}@umich.edu

More information

Optimal Transmission Strategies for Dynamic Spectrum Access in Cognitive Radio Networks. Senhua Huang, Xin Liu, and Zhi Ding

Optimal Transmission Strategies for Dynamic Spectrum Access in Cognitive Radio Networks. Senhua Huang, Xin Liu, and Zhi Ding 1 Optimal Transmission Strategies for Dynamic Spectrum Access in Cognitive Radio Networks Senhua Huang, Xin Liu, and Zhi Ding Abstract Cognitive radio offers a promising technology to mitigate spectrum

More information

On the static assignment to parallel servers

On the static assignment to parallel servers On the static assignment to parallel servers Ger Koole Vrije Universiteit Faculty of Mathematics and Computer Science De Boelelaan 1081a, 1081 HV Amsterdam The Netherlands Email: koole@cs.vu.nl, Url: www.cs.vu.nl/

More information

Opportunistic Cooperation in Cognitive Femtocell Networks

Opportunistic Cooperation in Cognitive Femtocell Networks 1 Opportunistic Cooperation in Cognitive Femtocell Networks Rahul Urgaonkar, Member, IEEE, and Michael J. Neely, Senior Member, IEEE Abstract We investigate opportunistic cooperation between secondary

More information

Optimal Sleeping Mechanism for Multiple Servers with MMPP-Based Bursty Traffic Arrival

Optimal Sleeping Mechanism for Multiple Servers with MMPP-Based Bursty Traffic Arrival 1 Optimal Sleeping Mechanism for Multiple Servers with MMPP-Based Bursty Traffic Arrival Zhiyuan Jiang, Bhaskar Krishnamachari, Sheng Zhou, arxiv:1711.07912v1 [cs.it] 21 Nov 2017 Zhisheng Niu, Fellow,

More information

Online Learning Schemes for Power Allocation in Energy Harvesting Communications

Online Learning Schemes for Power Allocation in Energy Harvesting Communications Online Learning Schemes for Power Allocation in Energy Harvesting Communications Pranav Sakulkar and Bhaskar Krishnamachari Ming Hsieh Department of Electrical Engineering Viterbi School of Engineering

More information

Bandit Algorithms. Zhifeng Wang ... Department of Statistics Florida State University

Bandit Algorithms. Zhifeng Wang ... Department of Statistics Florida State University Bandit Algorithms Zhifeng Wang Department of Statistics Florida State University Outline Multi-Armed Bandits (MAB) Exploration-First Epsilon-Greedy Softmax UCB Thompson Sampling Adversarial Bandits Exp3

More information

6196 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 57, NO. 9, SEPTEMBER 2011

6196 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 57, NO. 9, SEPTEMBER 2011 6196 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 57, NO. 9, SEPTEMBER 2011 On the Structure of Real-Time Encoding and Decoding Functions in a Multiterminal Communication System Ashutosh Nayyar, Student

More information

Residual White Space Distribution Based Opportunistic Channel Access Scheme for Cognitive Radio Systems. Technical Report: TR-CSE June 2010

Residual White Space Distribution Based Opportunistic Channel Access Scheme for Cognitive Radio Systems. Technical Report: TR-CSE June 2010 Residual White Space Distribution Based Opportunistic Channel Access Scheme for Cognitive Radio Systems Technical Report: TR-CSE-010-30 June 010 Manuj Sharma, Anirudha Sahoo Department of Computer Science

More information

The Optimality of Beamforming: A Unified View

The Optimality of Beamforming: A Unified View The Optimality of Beamforming: A Unified View Sudhir Srinivasa and Syed Ali Jafar Electrical Engineering and Computer Science University of California Irvine, Irvine, CA 92697-2625 Email: sudhirs@uciedu,

More information

Residual White Space Distribution Based Opportunistic Channel Access Scheme for Cognitive Radio Systems. Technical Report: TR-CSE March 2010

Residual White Space Distribution Based Opportunistic Channel Access Scheme for Cognitive Radio Systems. Technical Report: TR-CSE March 2010 Residual White Space Distribution Based Opportunistic Channel Access Scheme for Cognitive Radio Systems Technical Report: TR-CSE-010-6 March 010 Manuj Sharma, Anirudha Sahoo Department of Computer Science

More information

Distributed Learning in Multi-Armed Bandit with Multiple Players

Distributed Learning in Multi-Armed Bandit with Multiple Players SUBMITTED TO IEEE TRANSACTIONS ON SIGNAL PROCESSING, DECEMBER, 2009. 1 Distributed Learning in Multi-Armed Bandit with Multiple Players arxiv:0910.2065v3 [math.oc] 7 Jun 2010 Keqin Liu, Qing Zhao University

More information

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 5, MAY

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 5, MAY IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 5, MAY 1998 631 Centralized and Decentralized Asynchronous Optimization of Stochastic Discrete-Event Systems Felisa J. Vázquez-Abad, Christos G. Cassandras,

More information

Channel Selection in Cognitive Radio Networks with Opportunistic RF Energy Harvesting

Channel Selection in Cognitive Radio Networks with Opportunistic RF Energy Harvesting 1 Channel Selection in Cognitive Radio Networks with Opportunistic RF Energy Harvesting Dusit Niyato 1, Ping Wang 1, and Dong In Kim 2 1 School of Computer Engineering, Nanyang Technological University

More information

Cooperative Spectrum Sensing for Cognitive Radios under Bandwidth Constraints

Cooperative Spectrum Sensing for Cognitive Radios under Bandwidth Constraints Cooperative Spectrum Sensing for Cognitive Radios under Bandwidth Constraints Chunhua Sun, Wei Zhang, and haled Ben Letaief, Fellow, IEEE Department of Electronic and Computer Engineering The Hong ong

More information

Information in Aloha Networks

Information in Aloha Networks Achieving Proportional Fairness using Local Information in Aloha Networks Koushik Kar, Saswati Sarkar, Leandros Tassiulas Abstract We address the problem of attaining proportionally fair rates using Aloha

More information

STOCHASTIC OPTIMIZATION OVER PARALLEL QUEUES: CHANNEL-BLIND SCHEDULING, RESTLESS BANDIT, AND OPTIMAL DELAY. Chih-ping Li

STOCHASTIC OPTIMIZATION OVER PARALLEL QUEUES: CHANNEL-BLIND SCHEDULING, RESTLESS BANDIT, AND OPTIMAL DELAY. Chih-ping Li STOCHASTIC OPTIMIZATION OVER PARALLEL QUEUES: CHANNEL-BLIND SCHEDULING, RESTLESS BANDIT, AND OPTIMAL DELAY by Chih-ping Li A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY

More information

Quiz 1 EE 549 Wednesday, Feb. 27, 2008

Quiz 1 EE 549 Wednesday, Feb. 27, 2008 UNIVERSITY OF SOUTHERN CALIFORNIA, SPRING 2008 1 Quiz 1 EE 549 Wednesday, Feb. 27, 2008 INSTRUCTIONS This quiz lasts for 85 minutes. This quiz is closed book and closed notes. No Calculators or laptops

More information

3052 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 7, JULY /$ IEEE

3052 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 7, JULY /$ IEEE 3052 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL 52, NO 7, JULY 2006 Capacity of Nearly Decomposable Markovian Fading Channels Under Asymmetric Receiver Sender Side Information Muriel Médard, Senior Member,

More information

Administration. CSCI567 Machine Learning (Fall 2018) Outline. Outline. HW5 is available, due on 11/18. Practice final will also be available soon.

Administration. CSCI567 Machine Learning (Fall 2018) Outline. Outline. HW5 is available, due on 11/18. Practice final will also be available soon. Administration CSCI567 Machine Learning Fall 2018 Prof. Haipeng Luo U of Southern California Nov 7, 2018 HW5 is available, due on 11/18. Practice final will also be available soon. Remaining weeks: 11/14,

More information

Quality of Information Aware Scheduling in Task Processing Networks

Quality of Information Aware Scheduling in Task Processing Networks Resource Allocation on Wireless Networks and Wireless Networks - Communication, Cooperation and Competition Quality of Information Aware Scheduling in Task Processing Networks Rahul Urgaonkar, Ertugrul

More information

Adaptive Shortest-Path Routing under Unknown and Stochastically Varying Link States

Adaptive Shortest-Path Routing under Unknown and Stochastically Varying Link States Adaptive Shortest-Path Routing under Unknown and Stochastically Varying Link States Keqin Liu, Qing Zhao To cite this version: Keqin Liu, Qing Zhao. Adaptive Shortest-Path Routing under Unknown and Stochastically

More information

ABSTRACT CROSS-LAYER ASPECTS OF COGNITIVE WIRELESS NETWORKS. Title of dissertation: Anthony A. Fanous, Doctor of Philosophy, 2013

ABSTRACT CROSS-LAYER ASPECTS OF COGNITIVE WIRELESS NETWORKS. Title of dissertation: Anthony A. Fanous, Doctor of Philosophy, 2013 ABSTRACT Title of dissertation: CROSS-LAYER ASPECTS OF COGNITIVE WIRELESS NETWORKS Anthony A. Fanous, Doctor of Philosophy, 2013 Dissertation directed by: Professor Anthony Ephremides Department of Electrical

More information

Capacity of the Discrete Memoryless Energy Harvesting Channel with Side Information

Capacity of the Discrete Memoryless Energy Harvesting Channel with Side Information 204 IEEE International Symposium on Information Theory Capacity of the Discrete Memoryless Energy Harvesting Channel with Side Information Omur Ozel, Kaya Tutuncuoglu 2, Sennur Ulukus, and Aylin Yener

More information

Learning of Uncontrolled Restless Bandits with Logarithmic Strong Regret

Learning of Uncontrolled Restless Bandits with Logarithmic Strong Regret Learning of Uncontrolled Restless Bandits with Logarithmic Strong Regret Cem Tekin, Member, IEEE, Mingyan Liu, Senior Member, IEEE 1 Abstract In this paper we consider the problem of learning the optimal

More information

Exploiting Channel Memory for Joint Estimation and Scheduling in Downlink Networks

Exploiting Channel Memory for Joint Estimation and Scheduling in Downlink Networks Exploiting Channel Memory for Joint Estimation and Scheduling in Downlink Networks Wenzhuo Ouyang, Sugumar Murugesan, Atilla Eryilmaz, Ness B. Shroff Department of Electrical and Computer Engineering The

More information

HITTING TIME IN AN ERLANG LOSS SYSTEM

HITTING TIME IN AN ERLANG LOSS SYSTEM Probability in the Engineering and Informational Sciences, 16, 2002, 167 184+ Printed in the U+S+A+ HITTING TIME IN AN ERLANG LOSS SYSTEM SHELDON M. ROSS Department of Industrial Engineering and Operations

More information

On Distribution and Limits of Information Dissemination Latency and Speed In Mobile Cognitive Radio Networks

On Distribution and Limits of Information Dissemination Latency and Speed In Mobile Cognitive Radio Networks This paper was presented as part of the Mini-Conference at IEEE INFOCOM 11 On Distribution and Limits of Information Dissemination Latency and Speed In Mobile Cognitive Radio Networks Lei Sun Wenye Wang

More information

MARKOV DECISION PROCESSES (MDP) AND REINFORCEMENT LEARNING (RL) Versione originale delle slide fornita dal Prof. Francesco Lo Presti

MARKOV DECISION PROCESSES (MDP) AND REINFORCEMENT LEARNING (RL) Versione originale delle slide fornita dal Prof. Francesco Lo Presti 1 MARKOV DECISION PROCESSES (MDP) AND REINFORCEMENT LEARNING (RL) Versione originale delle slide fornita dal Prof. Francesco Lo Presti Historical background 2 Original motivation: animal learning Early

More information

Fairness and Optimal Stochastic Control for Heterogeneous Networks

Fairness and Optimal Stochastic Control for Heterogeneous Networks λ 91 λ 93 Fairness and Optimal Stochastic Control for Heterogeneous Networks sensor network wired network wireless 9 8 7 6 5 λ 48 λ 42 4 3 0 1 2 λ n R n U n Michael J. Neely (USC) Eytan Modiano (MIT) Chih-Ping

More information

Decentralized Multi-Armed Bandit with Multiple Distributed Players

Decentralized Multi-Armed Bandit with Multiple Distributed Players Decentralized Multi-Armed Bandit with Multiple Distributed Players Keqin Liu, Qing Zhao Department of Electrical and Computer Engineering University of California, Davis, CA 95616 {kqliu, qzhao}@ucdavis.edu

More information

Wideband Fading Channel Capacity with Training and Partial Feedback

Wideband Fading Channel Capacity with Training and Partial Feedback Wideband Fading Channel Capacity with Training and Partial Feedback Manish Agarwal, Michael L. Honig ECE Department, Northwestern University 145 Sheridan Road, Evanston, IL 6008 USA {m-agarwal,mh}@northwestern.edu

More information

Energy Harvesting Multiple Access Channel with Peak Temperature Constraints

Energy Harvesting Multiple Access Channel with Peak Temperature Constraints Energy Harvesting Multiple Access Channel with Peak Temperature Constraints Abdulrahman Baknina, Omur Ozel 2, and Sennur Ulukus Department of Electrical and Computer Engineering, University of Maryland,

More information

Morning Session Capacity-based Power Control. Department of Electrical and Computer Engineering University of Maryland

Morning Session Capacity-based Power Control. Department of Electrical and Computer Engineering University of Maryland Morning Session Capacity-based Power Control Şennur Ulukuş Department of Electrical and Computer Engineering University of Maryland So Far, We Learned... Power control with SIR-based QoS guarantees Suitable

More information

Operations Research Letters. Instability of FIFO in a simple queueing system with arbitrarily low loads

Operations Research Letters. Instability of FIFO in a simple queueing system with arbitrarily low loads Operations Research Letters 37 (2009) 312 316 Contents lists available at ScienceDirect Operations Research Letters journal homepage: www.elsevier.com/locate/orl Instability of FIFO in a simple queueing

More information

The Non-Bayesian Restless Multi-Armed Bandit: A Case of Near-Logarithmic Strict Regret

The Non-Bayesian Restless Multi-Armed Bandit: A Case of Near-Logarithmic Strict Regret 1 The Non-Bayesian Restless Multi-Armed Bandit: A Case of Near-Logarithmic Strict Regret arxiv:19.1533v1 [math.oc] 7 Sep 11 Wenhan Dai, Yi Gai, Bhaskar Krishnamachari and Qing Zhao Department of Aeronautics

More information

Application-Level Scheduling with Deadline Constraints

Application-Level Scheduling with Deadline Constraints Application-Level Scheduling with Deadline Constraints 1 Huasen Wu, Xiaojun Lin, Xin Liu, and Youguang Zhang School of Electronic and Information Engineering, Beihang University, Beijing 100191, China

More information

WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES

WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES Donald E. K. Martin 1 and John A. D. Aston 2 1 Mathematics Department, Howard University, Washington,

More information

HDR - A Hysteresis-Driven Routing Algorithm for Energy Harvesting Tag Networks

HDR - A Hysteresis-Driven Routing Algorithm for Energy Harvesting Tag Networks HDR - A Hysteresis-Driven Routing Algorithm for Energy Harvesting Tag Networks Adrian Segall arxiv:1512.06997v1 [cs.ni] 22 Dec 2015 March 12, 2018 Abstract The work contains a first attempt to treat the

More information

Performance and Convergence of Multi-user Online Learning

Performance and Convergence of Multi-user Online Learning Performance and Convergence of Multi-user Online Learning Cem Tekin, Mingyan Liu Department of Electrical Engineering and Computer Science University of Michigan, Ann Arbor, Michigan, 4809-222 Email: {cmtkn,

More information

Continuous-Model Communication Complexity with Application in Distributed Resource Allocation in Wireless Ad hoc Networks

Continuous-Model Communication Complexity with Application in Distributed Resource Allocation in Wireless Ad hoc Networks Continuous-Model Communication Complexity with Application in Distributed Resource Allocation in Wireless Ad hoc Networks Husheng Li 1 and Huaiyu Dai 2 1 Department of Electrical Engineering and Computer

More information

Part I Stochastic variables and Markov chains

Part I Stochastic variables and Markov chains Part I Stochastic variables and Markov chains Random variables describe the behaviour of a phenomenon independent of any specific sample space Distribution function (cdf, cumulative distribution function)

More information

Almost Optimal Dynamically-Ordered Channel Sensing and Accessing for Cognitive Networks

Almost Optimal Dynamically-Ordered Channel Sensing and Accessing for Cognitive Networks TRANSACTIONS ON MOBILE COMPUTING Almost Optimal Dynamically-Ordered Channel Sensing and Accessing for Cognitive Networks Bowen Li, Panlong Yang, Jinlong Wang, Qihui Wu, Shaojie Tang, Xiang-Yang Li, Yunhao

More information

Exploiting Channel Memory for Joint Estimation and Scheduling in Downlink Networks A Whittle s Indexability Analysis

Exploiting Channel Memory for Joint Estimation and Scheduling in Downlink Networks A Whittle s Indexability Analysis 1 Exploiting Channel Memory for Joint Estimation and Scheduling in Downlink Networks A Whittle s Indexability Analysis Wenzhuo Ouyang, Sugumar Murugesan, Atilla Eryilmaz, Ness B Shroff Abstract We address

More information

On the Throughput, Capacity and Stability Regions of Random Multiple Access over Standard Multi-Packet Reception Channels

On the Throughput, Capacity and Stability Regions of Random Multiple Access over Standard Multi-Packet Reception Channels On the Throughput, Capacity and Stability Regions of Random Multiple Access over Standard Multi-Packet Reception Channels Jie Luo, Anthony Ephremides ECE Dept. Univ. of Maryland College Park, MD 20742

More information

Analysis of random-access MAC schemes

Analysis of random-access MAC schemes Analysis of random-access MA schemes M. Veeraraghavan and Tao i ast updated: Sept. 203. Slotted Aloha [4] First-order analysis: if we assume there are infinite number of nodes, the number of new arrivals

More information

Greedy weighted matching for scheduling the input-queued switch

Greedy weighted matching for scheduling the input-queued switch Greedy weighted matching for scheduling the input-queued switch Andrew Brzezinski and Eytan Modiano Laboratory for Information and Decision Systems Massachusetts Institute of Technology {brzezin,modiano}@mit.edu

More information

Combining Opportunistic and Size-Based Scheduling in Wireless Systems

Combining Opportunistic and Size-Based Scheduling in Wireless Systems Combining Opportunistic and Size-Based Scheduling in Wireless Systems Pasi Lassila TKK Helsinki University of Technology P.O.Box 3, FI-15 TKK, Finland Pasi.Lassila@tkk.fi Samuli Aalto TKK Helsinki University

More information

Stability of the Maximum Size Matching

Stability of the Maximum Size Matching Stability of the Maximum Size Matching Mohsen Bayati Dept. of Electrical Engineering Stanford University Stanford, CA 9405-950 bayati@stanford.edu Neda Beheshti Dept. of Electrical Engineering Stanford

More information

arxiv: v1 [math.ho] 25 Feb 2008

arxiv: v1 [math.ho] 25 Feb 2008 A Note on Walking Versus Waiting Anthony B. Morton February 28 arxiv:82.3653v [math.ho] 25 Feb 28 To what extent is a traveller called Justin, say) better off to wait for a bus rather than just start walking

More information

Stochastic Contextual Bandits with Known. Reward Functions

Stochastic Contextual Bandits with Known. Reward Functions Stochastic Contextual Bandits with nown 1 Reward Functions Pranav Sakulkar and Bhaskar rishnamachari Ming Hsieh Department of Electrical Engineering Viterbi School of Engineering University of Southern

More information

Can Decentralized Status Update Achieve Universally Near-Optimal Age-of-Information in Wireless Multiaccess Channels?

Can Decentralized Status Update Achieve Universally Near-Optimal Age-of-Information in Wireless Multiaccess Channels? Can Decentralized Status Update Achieve Universally Near-Optimal Age-of-Information in Wireless Multiaccess Channels? Zhiyuan Jiang, Bhaskar Krishnamachari, Sheng Zhou, Zhisheng Niu, Fellow, IEEE {zhiyuan,

More information

Online Supplement to Delay-Based Service Differentiation with Many Servers and Time-Varying Arrival Rates

Online Supplement to Delay-Based Service Differentiation with Many Servers and Time-Varying Arrival Rates Online Supplement to Delay-Based Service Differentiation with Many Servers and Time-Varying Arrival Rates Xu Sun and Ward Whitt Department of Industrial Engineering and Operations Research, Columbia University

More information

Queue Length Stability in Trees under Slowly Convergent Traffic using Sequential Maximal Scheduling

Queue Length Stability in Trees under Slowly Convergent Traffic using Sequential Maximal Scheduling 1 Queue Length Stability in Trees under Slowly Convergent Traffic using Sequential Maximal Scheduling Saswati Sarkar and Koushik Kar Abstract In this paper, we consider queue-length stability in wireless

More information

Semi-Markovian User State Estimation and Policy Optimization for Energy Efficient Mobile Sensing

Semi-Markovian User State Estimation and Policy Optimization for Energy Efficient Mobile Sensing Semi-Markovian User State Estimation Policy Optimization for Energy Efficient Mobile Sensing ABSTRACT User context monitoring using sensors on mobile devices benefits end-users unobtrusively in real-time

More information