Minimizing Power Consumption of Source Encoding and Radio Transmission in CDMA Systems Xiaoan Lu Department of Electrical and Computer Engineering Polytechnic University, Brooklyn, NY Background ! Power control is essential to CDMA wireless networks " Relax the near far problem. " Improve the quality of service. " Increase the channel capacity. " Increase the battery life of the mobile terminal. ! Power control as an optimization problem " Minimize the total transmission power " maintain a required SINR threshold SINR: the signal to interference and noise ratio , depends on transmission power of all users. 2
Motivation (2/ 1) ! Previous power control focused on voice signal ! Video signal is integrated into the new generation wireless communication system " Constraint: PSNR or distortion (MSE), not SINR ! Lossy compression D s ! Erroneous transmission D t (D s + D t = D 0 ) ! PSNR: peak signal-to-noise ratio, MSE: mean squared error D s D s1 D s2 R s1 R s2 Bit rate (a) Distortion(compression) (b) Distortion(transmission) (c) (d) 3 ∝ Power bit rate x SI NR Motivation (2/ 1) ! Previous power control focused on voice signal ! Video signal is integrated into the new generation wireless communication system " Constraint: PSNR or distortion (MSE), not SINR " Minimize: transmission power + signal processing power " Parameters: { bit rate, compression complexity} , { transmission power} 4 ∝ Power bit rate x SI NR
Motivation (2/ 2) ! Proposed solution: D ynamically R econfigurable E nergy A ware M ultimedia I nformation T erminal (DREAM-IT) " Adapting operating parameters of all components simultaneously and dynamically to minimize total power consumption ! This subproject focuses on power allocation between video source coding, and radio transmission " Parameter: bit rate R , s , i β compressio n complexity i , transmissi on power P t , i 5 System description N ∑ + γ ( ) P P Minimize t , i s , i voice : SINR base 0 = i 1 ≤ station subject to D D tot , i i , 0 N ∑ D 0 Minimize P t , i P = i 1 subject to γ ≥ γ i 0 transmitter P transmitter P 1 t tN ...... baseband signal baseband signal processing processing terminal N terminal 1 { } N ∑ = β = + β γ = Adapt to minimize ,subject to c R , , P D ( R , , ) D P ( P P ) i s i i , t , i tot , i s , i i i i , 0 tot s , i t , i = ! the uplink of a CDMA cell i 1 ! video transmission β R : bit rate, : compressio n complexity P transmissi on power , , , : s i i t i 6
One terminal Component Parameters Distortion Power D s (R s , β ): lossy compression Video compressor Bit rate R s P s Complexity β p L ( γ ) : packet error rate Channel encoder D t ( β ,p L ): (1) transmission Transmitter Transmission P t power P t error (2) Error propagation γ : signal to interference and noise ratio , SINR , depends on transmission power of all users. N ∑ + = + ≤ ( ) Minimize subject to P P D D D D t , i s , i tot , i s , i t , i i , 0 = Decoded i 1 7 Conceptual illustration power total signal processing bit rate ! Special case example: , fixed individually, one user D D s t " Signal compression Bit rate , power " Transmission Bit rate , power " Total Bit rate , power ? 8
Why optimize jointly? ! Separate optimization c 1 " Operating parameters for terminal , is c i i i decided by base station + user bit rate R s , i = β c 3 c compressio n complexity c 2 i i transmissi on power P t , i ! One user’s signal is other users’ interference " All users interact with others " Local minima may not be global optimum c 1 ! Optimize jointly: Base station + all terminals " Full search: good for a small number of users c 3 " Iterative algorithm: converge? c 2 " Our approach: ! Simplified models + Lagarangian method ! Two-step fast algorithm 9 Simplified models ! Power consumption " Distortion= f(compression, transmission) = β ! From compression ( ) ( ) D D R D β , , , , s i s R s i s i [ ] ! Total distortion β γ = − β + σ 2 D ( R , , ) 1 p D ( , R ) p tot , i s , i i i L , i s , i i s , i L , i s , i " P , Source compression power s i β ! Increase linearly with complexity i " Transform coding: transform block size " H.263 encoder(periodic INTRA update, full ME search): INTER rate ! Independent of bit rate R , s i { } = β , , c R P Adapt i s i i , t , i N ∑ = + tot = to minimize ( ) subject to D D P P P , 0 tot s , i t , i i = 1 i 10
Method ! Lagarangian Multiplier method { } N ∑ = β + λ β − ( , , P ) ( , , P ) J c P R D R D s , i tot , i s , i i t i tot , i s , i i t i , 0 ∂ ∂ = ∂ 1 β γ = i J J J ( , , ) = = = D R D ! Equations: , , , 0 0 0 tot , i s , i i i i , 0 ∂ ∂ β ∂ R P , s i β i λ t , i ! Unknowns: R , , P , s , i i t , i i Γ = γ γ * * * ! ( ,..., ) is used to re-parameterize the equations 1 N ∂ J γ = R ~ 0 γ s , i i γ ~ ∂ R β ~ γ R ~ R s , i i tot = s , i s , i i D D i i ∂ β ~ γ λ ~ γ , i 0 N equations, J = i i 0 i i ∂ ∂ ∂ β with solutions J J = = 0 0 i γ γ ∂ λ * * ∂ ( ,..., ) P 1 N i γ γ t , i P ( ,..., ) t , i 1 N 11 Simulation ! Transform coding 1.4 first user ! Gauss-Markov source Source rate (bits/sample) second user 1.2 ! Two users 1 " 1 st user moves around 2 nd user closer " 2 nd user stands still 0.8 1 st user closer 0.6 0.4 0 0.2 0.4 0.6 0.8 1 distance (km) 2 nd user ! As distance increases, more compression is needed (lower source rate) ! The “better” users (with a small distance) needs to compress less 1 st user 12
Comparison ! our adaptive algorithm vs. fixed schemes (both users have same parameters) " Fixed simple compression (good for small distance) " Fixed complex compression (good for large distance) ! significant power saving 1 fixed compression (simple) Total power (user 1 + user 2) 0.8 0.6 Power 0.4 fixed compression (complex) 0.2 adaptive compression 0 0 0.2 0.4 0.6 0.8 1 13 distance (km) Two-step approach ! Computation " Dimensions ! Bit rate: M R base station ! complexity: M β ! SINR: M γ " Full search ! { R s , Β, Γ Β, Γ } Β, Γ Β, Γ × × N ( ) M M M β γ R Choose the optimum complexity β β β set to minimize the * * * " Two-step (can be { , ,..., } 1 2 N total power consumption, β further reduced) q ( ) i , max γ i β i β * * * * ( ) corresponding and R ( ) ! { Β Β } Β Β i , s i i are together taken as the optimum β γ i β × × γ + R ( ) ( ) N , N M M ( M ) s i i i β R operating parameters. 14
Conclusions ! Minimize total power consumption while maintaining the video quality at the receiver " mobile users sending video to a base station in one CDMA cell " Video compression power + radio transmission power are considered " An analytical solution based on simplified models " A two-step fast algorithm ! Results " Operating parameters depend on the distance " “better” users compress less. " Adaptive solution leads to significant power savings 15
Recommend
More recommend
