1 / 36

Dynamic Data Compression for Wireless Transmission

Explore channel state dynamic data compression for wireless transmission over a fading channel. Goal: Minimize total average power expenditure. Example system with packet arrivals and compression decisions. Sponsored by DARPA and NSF.

thomasjoseph
Download Presentation

Dynamic Data Compression for Wireless Transmission

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. 1 Compression 2 N S(t) = Channel State Dynamic Data Compression for WirelessTransmission over a Fading Channel Michael J. Neely University of Southern California CISS 2008 *Sponsored in part by the DARPA IT-MANET Program, NSF OCE-0520324, NSF Career CCF-0747525

  2. 1 Compression 2 N S(t) = Channel State -Random Packet Arrivals A(t) -Stochastic Channel S(t) -Data must be Compressed, Stored, and Transmitted Both Compression and Transmission expend power! Goal: Minimize Total Average Power Expenditure Ptot = Pcomp + Ptran (signal processing consumes power)

  3. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function = Random bit size output after compression

  4. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  5. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  6. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  7. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  8. Compression Operation: 1 Y(A(t), K(t)) 2 1.1 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example: A(t) = 2, K(t)=k Y(A(t), K(t)) = (1.1)B bits (random output) Pcomp(t) = .2 mW (random output)

  9. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 2: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  10. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 2: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  11. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 2: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  12. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 2: A(t) = 2, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  13. Compression Operation: 1 Y(A(t), K(t)) 2 0.9 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 2: A(t) = 2, K(t)=k Y(A(t), K(t)) = (0.9)B bits (random output) Pcomp(t) = .3 mW (random output)

  14. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 3: A(t) = 3, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  15. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 3: A(t) = 3, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  16. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 3: A(t) = 3, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  17. Compression Operation: 1 Y(A(t), K(t)) 2 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 3: A(t) = 3, K(t)=k Y(A(t), K(t)) = ?? (random output) Pcomp(t) = ?? (random output)

  18. Compression Operation: 1 Y(A(t), K(t)) 2 1.6 N Timeslotted System: t = {0,1, 2, 3,…} A(t) = # packet arrivals at time t (fixed packets size B bits, A(t) {0, 1, …, N} ) K(t) = Compression Decision Option at time t (K(t) {0, 1, …, K}) Y(A(t), K(t)) = Compression Function Example 3: A(t) = 3, K(t)=k Y(A(t), K(t)) = (1.6)B bits (random output) Pcomp(t) = .4 mW (random output)

  19. Good rate m Med Bad power P Transmission Operation: 1 Y(A(t), K(t)) 2 N S(t) = Channel State S(t) = Current Channel State on slot t Example: S(t) {“Good”, “Med”, “Bad”} S(t) {attenuation, 0 < S(t) < 1} Ptran(t) = Power Decision on slot t (P = set of power options) Example:P {0 < p < Pmax} or P {0, Pmax/2, Pmax} m(t) = C(Ptran(t), S(t)) = transmission rate on slot t rate-power curve

  20. Good rate m Med Bad power P Transmission Operation: 1 Y(A(t), K(t)) 2 N S(t) = Channel State S(t) = Current Channel State on slot t Example: S(t) {“Good”, “Med”, “Bad”} S(t) {attenuation, 0 < S(t) < 1} Ptran(t) = Power Decision on slot t (P = set of power options) Example:P {0 < p < Pmax} or P {0, Pmax/2, Pmax} m(t) = C(Ptran(t), S(t)) = transmission rate on slot t rate-power curve

  21. Ptot = Pcomp + Ptran 1 Y(A(t), K(t)) 2 Q(t) ~ bits N Queueing Dynamics: Q(t+1) = max[Q(t) - m(t), 0] + R(t) m(t) = C(Ptran(t), S(t)) , R(t) = Y(A(t), K(t)) Goal: Design a joint strategy for choosing compression and transmission decisions K(t), Ptran(t) over time to support all traffic and minimize total average power. [Also want low delay!]

  22. Intuition:Consider a trivial case where all of the following hold… Static Channel: S(t) = Constant Linear rate-power curve: C(P) = aP Raw Data Rate small: E{A(t)}B < mmax= aPmax If all 3 hold, easy to show Ptran proportional to bits transmitted, and decision is trivial: choose K(t) = k {0,…,K} that minimizes: f(a, k) + m(a,k)/a If one or more of the above fail, decision is non-trivial!

  23. Prior Experimental Work in this case: Barr, Asanovic “Energy Aware Lossless Data compression” [2003] 2. Sadler, Martonosi “Data Compression algorithms for energy-constrained devices” [SenSys2006] Data Compression an important problem! [Richard Baraniuk’s Plenary talk was totally cool!] Intelligent Compression Can Significantly Improve Power Expenditure.

  24. Defining Optimality: The h*(r) and g*(r) functions Assume: - # arrivals A(t) iid over slots, pA(a) = Pr[A(t) = a] - S(t) iid over slots, ps = Pr[S(t) = s] Distributions pA(a) and ps are potentially unknown to controller Min bit rate out of compressor Max time avg. transmission rate (assume rmin < rmax)

  25. Defining Optimality: The h*(r) and g*(r) functions Assume: - # arrivals A(t) iid over slots, pA(a) = Pr[A(t) = a] - S(t) iid over slots, ps = Pr[S(t) = s] Distributions pA(a) and ps are potentially unknown to controller Intuitively: There are two reasons to compress data: To stabilize queue, we may need to compress (if raw input rate > rmax ) ii) Power spent compressing can reduce trans. power.

  26. Defining Optimality: The h*(r) and g*(r) functions h*(r) = minimum Pcomp to yield compressor output rate r. g*(r) = minimum Ptran to yield avg. transmission rate r. Min h such that: Min g such that: (optimizing over all stationary randomized opportunistic policies) h*(r) g*(r) rate r rmax rate r rmin

  27. Theorem (characterizing minimum avg. power): Any stabilizing policy yields avg. power such that: Pcomp + Ptran P* > av Where P* is the min avg. power and satisfies: av Minimize: h*(r) + g*(r) Subject to: rmin < r < min[rmax, BE{A(t)}] Total avg power rate r rmin min[rmax, BE{A(t)}]

  28. Want a dynamic control algorithm to minimize power… Use Joint Lyapunov Stability + Performance Optimization Technique for Stochastic Network Optimization: -Georgiadis, Neely, Tassiulas [NOW F&T in Networking, 2006] -Neely [Phd thesis 2003, Infocom 2005, IT 2006] -See also related technique in Stolyar [Queueing Systems 2005] Lyapunov Function: L(Q) = (1/2) Q2 Lyapunov Drift:D(Q(t)) = E{L(Q(t+1)) - L(Q(t)) | Q(t)} Technique: Every slot, observe Q(t), take control action to minimize (for a given constant V>0): D(Q(t)) + VE{Cost(t)| Q(t)} Theorem: 1) E{Q} < O(V) 2) E{Cost(actual) - Cost(optimal)} < O(1/V)

  29. 1 Y(A(t), K(t)) 2 N S(t) = Channel State The Dynamic Compression and Transmission Algorithm: Compression: On slot t, observe A(t). Choose K(t) such that: Transmission: On slot t, observe S(t). Choose Ptran(t) such that: Q Q

  30. 1 Y(A(t), K(t)) 2 N S(t) = Channel State Q Concluding Theorem (Performance): For any V>0 we have: B = (1/2)[mmax2 + B2E{A2}]

  31. 1 Y(A(t), K(t)) 2 N S(t) = Channel State Q P* av Concluding Theorem (Performance): For any V>0 we have: Avg Power Avg Delay V V

  32. 1 Y(A(t), K(t)) 2 N S(t) = Channel State Q P* av Concluding Theorem (Performance): For any V>0 we have: Avg Power Avg Delay V V

  33. 1 Y(A(t), K(t)) 2 N S(t) = Channel State Q P* av Concluding Theorem (Performance): For any V>0 we have: Avg Power Avg Delay V V

  34. 1 Y(A(t), K(t)) 2 N S(t) = Channel State Q P* av Concluding Theorem (Performance): For any V>0 we have: Avg Power Avg Delay V V

  35. 1 Y(A(t), K(t)) 2 N S(t) = Channel State Q P* av Concluding Theorem (Performance): For any V>0 we have: Avg Power Avg Delay V V

  36. 1 Y(A(t), K(t)) 2 N S(t) = Channel State Q P* av Concluding Theorem (Performance): For any V>0 we have: Avg Power Avg Delay V V

More Related