Fronthaul Compression for Cloud Radio Access Networks O. Simeone New Jersey Institute of Technology (NJIT) Joint work with S.-H. Park 1 , O. Sahin 2 and S. Shamai 3 3 1 2
Cloud Radio Access Networks Base stations operate as radio units • Baseband processing takes place in the “cloud” • Fronthaul links carry complex (IQ) • baseband signals
Cloud Radio Access Networks Advantages: • Low-cost BSs • Effective interference mitigation via joint baseband processing Key challenge: Effective transfer of the IQ signals on the fronthaul links
Cloud Radio Access Networks • CPRI standard based on ADC/DAC … Need for fronthaul compression
State of the Art • Point-to-point fronthaul compression: – Algorithms [Segel and Weldon] [Samardzija et al ‘12] [Nieman and Evans ’13] – Testbed results [Irmer et al ’11] [Vosoughi et al ‘12]
State of the Art • Multiterminal fronthaul compression: – Uplink: Distributed source coding coding [Sanderovich et al ’09] [del Coso and Simoens ’09] [Zhou and Yu ’11] [Marsch and Fettweis ’11] – Downlink: Multivariate compression [Park et al ’13] • Compute-and-forward: – Uplink [Nazer et al ’09] [Hong and Caire ’11] – Downlink [Hong and Caire ‘12]
Overview • Uplink – Multiterminal compression – Compute-and-forward • Downlink – Multiterminal compression – Compute-and-forward • Performance evaluation • Extensions and conclusions
Overview • Uplink – Multiterminal compression – Compute-and-forward • Downlink – Multiterminal compression – Compute-and-forward • Performance evaluation • Extensions and conclusions
System Model H 1 y ˆ 1 RU1 C y 1 MS 1 1 H i ˆ i C y CU i RU i y i MS N M C N B RU N H B ˆ y N B N B y N B Single-cluster single-hop fronthaul topology
Point-to-Point Fronthaul Compression Control Unit RU 1 C ˆ ul y 1 1 Decompressor Compressor ul y 1 Fronthaul RU 2 ˆ ul y C 2 2 Decompressor Compressor ul y Decoder 2 Fronthaul RU N R ul ˆ y C N N R R ul y Decompressor Compressor N R Fronthaul
Joint Fronthaul Decompression [Sanderovich et al ’09] [del Coso and Simoens ’09] [Zhou and Yu ’11] [Park et al ’13] Control Unit RU 1 C 1 ul y ˆ ul y Compressor (1) (1) Decompressor Fronthaul RU 2 C 2 ul y ˆ WZ WZ ul y (2) (2) Compressor Decoder Decompressor Fronthaul RU N R C N R WZ ul y WZ ( ) N R Compressor ˆ ul y Fronthaul Decompressor ( ) N R
Joint Fronthaul Decompression 010 001 … … 000 ul y 100 ul ˆ y 101 Point-to-point compression
Joint Fronthaul Decompression 010 001 … … 000 ul y 100 ul 101 ˆ y WZ compression … Coset coding at the RU and channel decoding at the CU [Pradhan and Ramchandran ’03]
Compute-and-Forward [Nazer et al ’09] [Hong and Caire ’11] • The MSs use (nested) lattice codes: [B. Nazer]
Compute-and-Forward [Nazer et al ’09] [Hong and Caire ’11] Control Unit RU 1 C 1 Integer ul y Decoder 1 Fronthaul RU 2 C 2 Integer ul y 2 Decoder Decoder Fronthaul RU N R C N R Integer ul y N R Decoder Fronthaul
Numerical Results • Three-cell SISO circular Wyner model CU C C C
Numerical Results 3 bit/s/Hz and =0.4 C 3 Cut-set upper bound 2.5 per-cell sum-rate [bits/s/Hz] Point-to-point compression 2 Single-cell processing 1.5 1 0 5 10 15 20 25 30 MS transmit power [dB]
Numerical Results 3 bit/s/Hz and =0.4 C 3 Cut-set upper bound Joint decompression 2.5 per-cell sum-rate [bits/s/Hz] Point-to-point compression 2 Single-cell processing 1.5 1 0 5 10 15 20 25 30 MS transmit power [dB]
Numerical Results 3 bit/s/Hz and =0.4 C 3 Cut-set upper bound Joint decompression 2.5 per-cell sum-rate [bits/s/Hz] Point-to-point compression 2 Single-cell processing 1.5 Compute-and-forward 1 0 5 10 15 20 25 30 MS transmit power [dB]
Overview • Uplink – Multiterminal compression – Compute-and-forward • Downlink – Multiterminal compression – Compute-and-forward • Performance evaluation • Extensions and conclusions
System Model RU1 H 1 C 1 MS 1 1 ,..., M M C CU N M i RU i C N B MS N M H RU N N M B Single-cluster single-hop fronthaul topology
Point-to-Point Fronthaul Compression Control Unit C s M x Compressor 1 x 1 Channel 1 1 RU 1 1 1 encoder 1 Precoding s x M C Channel N N Compressor N N M x B B M RU N N encoder N M B N B B
Joint Fronthaul Compression [Park et al ’13] Control Unit C s M x 1 x 1 Channel 1 1 RU 1 1 encoder 1 Joint Precoding compression s x M C Channel N N N N M x B B M RU N N encoder N M B B
Joint Fronthaul Compression • Multivariate compression produces compressed signals with correlated quantization noises q x E As q H 1 y H As H z 1 1 1 z 1 1 1 1 q 1 C 2 1 RU 1 H 1,1 MS CU RU 2 H C 1,2 2 x E As q H 2 2 2 Ω Ω 1,1 1,2 CN 0 H , H H 1 1 Ω Ω 2,1 2,2 can be reduced by controlling Ω Ω H 1,2 2,1
Joint Fronthaul Compression x 2 x 1 Point-to-point compression
Joint Fronthaul Compression x 2 x 1 Point-to-point compression
Joint Fronthaul Compression x 2 x 1 Multivariate compression
Joint Fronthaul Compression Successive estimation-compression implementation [Park et • al ’13]: x x (1) (1) RU π(1) Compressor x ˆ x x (2) (2) MMSE (2) RU π(2) Compressor estimation x ˆ x x ( ) N ( ) MMSE N B B ( ) RU π(N B ) N Compressor B estimation
Compute-and-Forward • Reverse compute-and-forward (RCoF) [Hong and Caire ‘12] Control Unit C s M x 1 x 1 Channel 1 1 RU 1 1 encoder 1 Integer precoding s x C M N Channel N N B N M x B M RU N encoder N M N B B
Numerical Results • Three-cell SISO circular Wyner model CU C C C
Numerical Results • Three-cell SISO circular Wyner model ( and ) P 0.5 20 dB Cut-set upper bound 6 Joint compression 5 Linear precoding per-cell sum-rate [bits/s/Hz] 4 Point-to-point compression 3 2 Single-cell processing 1 0 0 2 4 6 8 10 12 14 C [bits/s/Hz]
Numerical Results • Three-cell SISO circular Wyner model ( and ) P 20 dB 0.5 Cut-set upper bound 6 DPC precoding Joint compression 5 Linear precoding per-cell sum-rate [bits/s/Hz] 4 Point-to-point compression 3 2 Single-cell processing 1 0 0 2 4 6 8 10 12 14 C [bits/s/Hz]
Numerical Results • Three-cell SISO circular Wyner model ( and ) P 20 dB 0.5 Cut-set upper bound 6 DPC precoding Joint compression 5 Linear precoding per-cell sum-rate [bits/s/Hz] 4 Compute-and-forward Point-to-point compression 3 2 Single-cell processing 1 0 0 2 4 6 8 10 12 14 C [bits/s/Hz]
Overview • Uplink – Multiterminal compression – Compute-and-forward • Downlink – Multiterminal compression – Compute-and-forward • Performance evaluation • Extensions and conclusions
Simulation Set-up • In each macro-cell, pico-BSs and MSs are uniformly K N distributed.
Simulation Set-up F Frequency reuse pattern with reuse factor for 1-cell cluster 1/ 3 • [Wang and Yeh ’11]
Numerical Results • Cell-edge throughput versus average spectral efficiency – Uplink, 1-cell cluster, 3 pico-BS, 5 MSs, ( , ) (9,3)bps/Hz, 10, 0.5, 1/ 3 N K C C T F macro pico max 3600 =2.0 3400 5%-ile rate (cell-edge throughput) [kbps] =1.0 3200 3000 1.6x 2800 =0.5 2600 2400 =0.25 2200 Point-to-point compression 2000 Multiterminal compression 1800 0.85 0.9 0.95 1 1.05 1.1 spectral efficiency [bps/Hz]
Numerical Results • Cell-edge throughput versus average spectral efficiency – Downlink, 1-cell cluster, 1 pico-BS, 4 MSs, ( , ) (3,1)bps/Hz, 5, 0.5, 1/ 3 N K C C T F macro pico max 4000 3500 =0.5 5%-ile rate (cell-edge throughput) [kbps] 3000 =1.5 2x 2500 2000 1500 1000 Point-to-point compression =0.25 500 Multiterminal compression 0 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 spectral efficiency [bps/Hz]
Overview • Uplink – Multiterminal compression – Compute-and-forward • Downlink – Multiterminal compression – Compute-and-forward • Performance evaluation • Extensions and conclusions
Multiterminal Compression with Imperfect CSI [Park et al ‘13]
Recommend
More recommend