Presentation is loading. Please wait.

Presentation is loading. Please wait.

September 19-20, 2005© SEIL 20051 Progress Summary “Dawn” MURI Review Anthony Ephremides University of Maryland Santa Cruz, CA September 12, 2006.

Similar presentations


Presentation on theme: "September 19-20, 2005© SEIL 20051 Progress Summary “Dawn” MURI Review Anthony Ephremides University of Maryland Santa Cruz, CA September 12, 2006."— Presentation transcript:

1 September 19-20, 2005© SEIL 20051 Progress Summary “Dawn” MURI Review Anthony Ephremides University of Maryland Santa Cruz, CA September 12, 2006

2 © HyNet 20062 OUTLINE Over-Riding Theme: Layer Integration & Theoretical Foundations –Physical layer incorporation into MAC and beyond 1)Capture 2)Alternative Access/Receiver Comparison 3)Multicast Stability/Capacity 4)Tandem Network Stability/Capacity Network Coding

3 © HyNet 20063 (1)On Capture (with J. Wieselthier & G. Nguyen) Essence: –Bridging The Gap Of Networking And Communication-Information-Theoretic Approach To Multiple Access via the SINR Tool –Correcting Previous Analyses For Capture Probability Derivation

4 © HyNet 20064 Random-Access System Collision channel –no capture General Multiple-Access channel –all users “succeed” In-between: Reception in the presence of interference –SINR-based model One or more users can be successful Receiver

5 © HyNet 20065 Capture Probability Capture probability: –C n = Pr{at least one transmission is successful | n simultaneous transmissions} Expected number of successful packets in a slot: –S n = E{number of successful packets | n simultaneous transmissions} Multi-Packet reception capability –Depends on detector Receiver

6 © HyNet 20066 SINR-based Capture Model A packet from user j is successful if and only if b = 0: Perfect capture single detector: largest always successful multiple detectors: all are successful b = ∞: No capture (collision channel) when 2 or more transmit, none are successful P(j) = Power at receiving node from user j b = Threshold that depends on many system parameters (increasing function of rate) Receiver j

7 © HyNet 20067 Earlier Work (Zorzi & Rao, JSAC 1994) t = test user P n (r 0 ) = Pr{SINR(t) > b | r t = r 0 } h(r 0 ) = pdf of r 0 (distance of user to base station)  ( * ) is not valid for b < 1  Implicitly assumes only one signal can satisfy SINR Example: Propagation loss factor  = 4 Fading and shadowing are present *which exceeds 1 when (*)(*)

8 © HyNet 20068 Extend Model to Accommodate All Values of b Observations More than one user can satisfy SINR > b when b < 1  Interesting case C n = Pr{one or more users satisfy SINR condition} = Pr{largest signal satisfies SINR condition} Let user M be the one with the largest signal *Thus, C n = Pr{SINR(M) > b} Since all users are equally likely to be the largest *C n = n Pr{SINR(1) > b, M = 1}

9 © HyNet 20069 Analytical Evaluation of Capture Probability C n = n Pr{SINR(1) > b, M = 1} = Pr{SINR(M) > b} *where M is the user with largest received power Example: *For In general, where F P is the common cdf of the received power levels (which are i.i.d.)

10 © HyNet 200610 Simulation is Needed to Evaluate C n Users uniformly distributed in disk of radius  = 1 –No fading or shadowing: any propagation model can be accommodated Results for b > 1 are same as those obtained by others The model is not realistic!  Valid only in far-field region  Received power approaches ∞ as r approaches 0

11 © HyNet 200611 More-Realistic Physical Model Assume users are uniformly distributed in a circular region of radius  = 10. No fading. Curves for C n are drastically different from those for  Previously described performance is not correct  Overestimates received power when transmitter is close to receiver  = 2  = 4  = 2 instead of

12 © HyNet 200612 Multi-Packet Reception All packets for which SINR > b are successful –Not only the largest S n = n Pr{SINR(1) > b}  = 10

13 © HyNet 200613 A Network with Two Destinations Users uniformly distributed throughout union of 2 circles of radius  One destination receiver in each circle –Separated by distance d Traffic distribution –Each packet has a specific destination (receiver) Does not add to throughput when decoded at “wrong” receiver Adds to interference at both receivers –In intersection of 2 circles Packet is equally likely to be intended for D 1 or D 2 –In rest of region Packet is intended for closer destination

14 © HyNet 200614 C n for Two-Destination Network d = 20 (circles just touching) d = 15 (circles overlap)  = 10 n 1 = n 2 = n (i.e., same number of packets for each destination)  Results demonstrate impact of “broader interference” effect resulting from overlapping user populations

15 © HyNet 200615 (2) Alternative Access- Receiver Comparison (joint work with Jie (Rockey Luo) Essence: –Comparison of Scheduled and Simultaneous (Parallel) Multiple Access Strategies –Two Low-Complexity Schedules: TDMA PMAS –Spectral Efficiency vs. Energy Cost comparison –PMAS Dominates if Multiple Antennas are Available at the Receiver

16 © HyNet 200616 Three Channel Sharing Schemes Optimal (complex) Parallel transmission, Multiuser decoding. TDMA (simple) Sequential transmission PMAS (simple) Parallel transmission, Single user decoding. Joint Decoding R sum =log[1+  i P i /N 0 ] SNR i =P i /(N 0 +  j  i P j ) Single user Decoding Gaussian noise of power N 0 P i is the transmit power of user i. All channels have unit gain.

17 © HyNet 200617 Spectral Efficiency Comparison approximately half spectral efficiency 10 transmitters, time-invariant channel, channel gains randomly generated, transmitters know channel states 10 transmitters, time-variant channel, independent Rayleigh fading, transmitters do not know channel states approximately half spectral efficiency Comparing the three schemes: Optimal, TDMA and PMAS

18 © HyNet 200618 Comparison on System Slopes 10 transmitters, time-variant channel, Rayleigh fading, CDI at the transmitter K: number of transmitters M: number of receive antennas We can exploit space resource at the receiver!

19 © HyNet 200619 Superiority of PMAS over TDMA General Results : PMAS is better than TDMA ! In low power regime, let # of terminals  ∞ R PMAS  R OPT / 2 R OPT, R PMAS  # receive antennas R TDMA  constant Multiantenna sampling creates a multi dimensional image. Nature helps separating signals (multiuser multiplex gain) Orthogonal channel sharing is inefficient in exploiting multiuser multiplex gain.

20 © HyNet 200620 Summary The dominance of TDMA is due to its simplicity. 1. Orthogonal channel sharing is inefficient in exploiting multiuser multiplex gain. Such inefficiency can be significant in the low power regime. 2. Overcoming this inefficiency with simple alternative is the key challenge. We showed that such alternative exists in low power regime multiple access systems. 3.

21 © HyNet 200621 (3) Multicast Stability/Capacity (joint work with B. Shrader and Y. Sagduyu) Essence: –Extension of Collision Channel Results –Stability Region Capacity Region –Simple Network coding Scheme Extends Stability Region to the Capacity Region

22 © HyNet 200622

23 © HyNet 200623

24 © HyNet 200624

25 © HyNet 200625

26 © HyNet 200626

27 © HyNet 200627

28 © HyNet 200628 (4) Tandem Network Stability/Capacity (via Network coding) (joint work with Y. Sagduyu) Essence: –Joint Access/Network coding Broadcasting over Wireless Tandem Networks –Focus on Stable Throughput –Extensions (Cooperative/Competitive Strategies, Energy Optimization)

29 © HyNet 200629 Tandem network: Multiple sources. Error-free transmissions & Mostly broadcasting. Case 1: Assume continuously generated packet traffic, i.e. saturated packet queues. –With or without Network Coding, Find Maximum Throughput Region. Case 2: Allow packet queues to empty. –With or without Network Coding, Find Maximum Stable Throughput Region. Model and Objectives i,j : average rate (packets/s) i j

30 © HyNet 200630 Three separate queues at node i: Plain Routing: Network Coding: Tandem Wireless Network Model (Saturated Queues) Scheduled Access: Group 1: 1, 4, 7, …, Group 2: 2, 5, 8, …, Group 3: 3, 6, 9, … Activate node group m over disjoint fractions of time t m, m  {1,2,3}. Random Access: Node i transmits (new or collided) packets with fixed probability p i. (Crucial Point) 123 n -1 n 456 Q i 1 stores source packets node i generates. Q i 2 stores relay packets from right neighbor of node i Q i 3 stores relay packets from left neighbor of node i + or Qi1Qi2Qi3Qi1Qi2Qi3 Qi1Qi2Qi3Qi1Qi2Qi3 Qi3Qi3 Qi1Qi1 Qi2Qi2 (Linear combination)

31 © HyNet 200631 Achievable Throughput Region under Scheduled Access  i r and  i l : total rates of packets arriving at node i from right and left neighbors. i : throughput rate from node i to destinations M i. Throughput rates satisfy: Achievable throughput region A includes s.t.: For n = 3, achievable throughput region A is:

32 © HyNet 200632 Allow packet queues to empty. –Packet underflow possible: node can wait to perform Network Coding or proceed with Plain Routing. –Consider two dynamic strategies based on instantaneous queue contents: Strategy 1 : Every node attempts first to transmit relay packets and transmits a source packet only if both relay queues are empty. Strategy 2: Every node attempts first to transmit a source packet and transmits relay packets only if the source queue is empty. –Strategy 2 expands the stability region STR(S) to the boundary of TR(A). Stable Throughput Region under Scheduled Access

33 © HyNet 200633 Deriving the achievable and stable throughput regions A and S is difficult. –Throughput regions depend on the transmission schedules t. Consider alternative measures: Assume saturated queues (or non-saturated queues together with strategy 2.). Alternative Optimization Measures Minimum transmitted throughput “Sum”-delivered throughput Find best schedule t to maximize min or  over  A or  S.

34 © HyNet 200634 Throughput Optimization Results For broadcasting with M i = N – {i}, i  N :  min = 0 for optimal values of .  Network coding doubles .  No improvement in min, as n increases.  Objectives of maximizing min and  under broadcast communication cannot be achieved simultaneously.

35 © HyNet 200635 Extension to Random Access Assume saturated queues (otherwise, the problem involves interacting queues). –Nodes randomize between waiting or transmitting source packets or relay packets. –Source packet transmissions: A: Transmit new source packets at any time slot (no feedback - possible loss) B: Transmit source packets until they are received by both neighbors (feedback + repetition) C: Transmit linear combinations of source packets (feedback + coding) : Packet remains in queue Q i 1. Packet enters queue Q i 2. Packet enters queue Q i 3.

36 © HyNet 200636 Wrap-Up Four Inter-Related Sets of Contributions Physical Layer Role in MA Network coding in Wireless Multicast Environments Stable Throughput Region Focus

37 © HyNet 200637 Plans Convert Throughput Results (packets/s) to Spectral Efficiencies (bits/s) Scheduling versus Parallel Transmission (with or without Network Coding) Packet Erasure Channel Models to be Broadened Connect Delay Analysis to Back-Pressure Algorithm


Download ppt "September 19-20, 2005© SEIL 20051 Progress Summary “Dawn” MURI Review Anthony Ephremides University of Maryland Santa Cruz, CA September 12, 2006."

Similar presentations


Ads by Google