Coordinated Multi-Point (CoMP) Adaptive Estimation and Prediction - PowerPoint PPT Presentation

Coordinated Multi-Point (CoMP) Adaptive Estimation and Prediction Schemes using Superimposed and Decomposed Channel Tracking Gencer Cili, Halim Yanikomeroglu, and F. Richard Yu Department of Systems and Computer Engineering, Carleton University, ON, Canada Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 1/19 ICC 2013 June 09, 2013

Introduction Motivation: • Comparison of decomposed and superimposed channel tracking methods in Coordinated Multi-Point (CoMP) networks is not studied in existing literature. Our contributions • Multi-point channel estimation and prediction framework to tackle the CoMP system delays and inaccurate measurements • Comparison between superimposed versus decomposed channel estimation schemes • CoMP adaptive switching/fallback between channel estimation schemes • Balance the clustering accuracy versus the channel estimation computation complexity trade-off • Effects of estimation/prediction filter length increases on CoMP performance are characterized according to users being served by various cluster sizes Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 2/19 ICC 2013 June 09, 2013

CoMP Joint Transmission Procedures CoMP Definition: Dynamic coordination among multiple geographically separated points referred as CoMP cooperating set for downlink transmission and uplink reception 3 Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 3/19 ICC 2013 June 09, 2013

CoMP Joint Transmission Procedures Parameter Assumption or Value 𝐎 𝐃𝐩𝐩𝐪 CoMP coordinating set 𝐎 𝐧𝐟𝐛𝐭 CoMP measurement set 𝐎 𝐊𝐔 CoMP joint transmission set 4 Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 4/19 ICC 2013 June 09, 2013

Threshold Based CoMP Joint Transmission Clustering • Actual measured received power from eNB n by user i at TTI t : 𝑄 𝑆𝑌 𝑜, 𝑢, 𝑗 = 𝑄 𝑈𝑌 𝑜 − 𝑄𝑀 𝑜, 𝑗 − 𝑄 𝐺𝑏𝑒𝑗𝑜𝑕 𝑜, 𝑗, 𝑢 • Received feedback due to estimation error + system delay: 𝑄 𝑆𝑌_𝑓𝑠𝑠 𝑜, 𝑢, 𝑗 = 𝑄 𝑆𝑌 𝑜, 𝑢 − ∆, 𝑗 + 𝑄 𝑓𝑠𝑠 (𝜈, 𝜏) • Threshold based Decision to Form the CoMP Transmission Set:  n ( , ) i t argmax{ P ( , , )} n i t Best RX err n n ∊ 𝑂 𝐾𝑈 𝑗, 𝑢 𝑗𝑔 |𝑄 𝑆𝑌 𝑓𝑠𝑠 (𝑜 𝐶𝑓𝑡𝑢 , 𝑗, 𝑢) − 𝑄 𝑆𝑌 𝑓𝑠𝑠 (𝑜, 𝑗, 𝑢)| ≤ 𝛼 𝑂𝑋−𝐾𝑈 Note: Contents of the joint transmission set will be impacted by reported CSI feedbacks due to multi-point channel estimation errors and system delays !!! Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 5/19 ICC 2013 June 09, 2013

CoMP Performance Metrics - Capacity • Joint PDSCH transmission (TM-9) mitigates the Inter-cell Interference Single Point Transmission CoMP Downlink Transmission 𝑄 𝑡𝑓𝑠𝑤𝑗𝑜𝑕 𝑇𝐽𝑂𝑆 𝐷𝑝𝑁𝑄 = 𝑄 𝑡𝑓𝑠𝑤𝑗𝑜𝑕 + 𝑄 𝑘 + 𝑄 𝑛 𝑇𝐽𝑂𝑆 = 𝐿 𝑗=1 𝑄 𝑗 + 𝑄 𝑂𝑝𝑗𝑡𝑓 𝐿 𝑗=1 𝑄 𝑗 + 𝑄 𝑂𝑝𝑗𝑡𝑓 𝑗≠𝑘,𝑛 Total received Power from CoMP Transmission Set   P ( , ) i t P ( , , ) n i t JT RX  n N JT Perceived Downlink Capacity due to CoMP     P ( , ) i t   JT C i t ( , ) W i t ( , )log  1    2 P ( , , ) n i t P   RX noise    n N N \ ( ) i JT Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 6/19 ICC 2013 June 09, 2013

CoMP Performance Metrics – Energy Efficiency CoMP Power Consumption Model 2 ) 𝑄 𝑇𝑄−𝐷𝑝𝑁𝑄 = 58 (0.87 + 0.1𝑂 𝐷 + 0.03𝑂 𝐷 Signal Processing Power 𝐷 𝐶𝐼 Backhauling Power 𝑄 𝐶𝐼 = 100𝑁𝑐𝑗𝑢𝑡/𝑡𝑓𝑑 50𝑋 𝐷 𝐶𝐼 = 𝑂 𝑑 2𝑂 𝐷 𝑞𝑟 Additional Data capacity 𝑐𝑗𝑢𝑡/𝑡𝑓𝑑 𝑈 𝑇 for CoMP Backhauling 𝑄 𝑈𝑌 Total Power Consumption 𝑄 𝐷oMP = 𝑂 𝑡 𝑂 + 𝑄 1 + 𝐷 𝐷 1 + 𝐷 𝐶𝐶 + 𝑄 𝐶𝐼 𝑄𝐵 𝑇𝑄 of an eNB using CoMP 𝑄𝐵 𝑓𝑔𝑔 𝑡𝑓𝑑𝑢𝑝𝑠 Power Consumption Parameters = Power amplifiers per sector 𝑄 𝑈𝑌 = DL Transmit Power, 𝐷 𝐷 = Cooling Loss 𝑂 𝑡 = Number of Sectors 𝑂 𝑄𝐵 𝑡𝑓𝑑𝑢𝑝𝑠 𝐷 𝐶𝐶 = Battery Backup 𝑂 𝐷 = Number of points in Joint Transmission 𝑞 = pilot density 𝑟 = CSI signalling 𝑈 𝑇 = Symbol Period 𝑄𝐵 𝑓𝑔𝑔 = Power Amplifier Efficiency Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 7/19 ICC 2013 June 09, 2013

CoMP Performance Metrics – Energy Efficiency 𝐷𝑏𝑞𝑏𝑑𝑗𝑢𝑧 (𝑐𝑗𝑢𝑡/𝑡𝑓𝑑) 𝑐𝑗𝑢𝑡 𝑄𝑝𝑥𝑓𝑠 𝐷𝑝𝑜𝑡𝑣𝑛𝑞𝑢𝑗𝑝𝑜(𝐾𝑝𝑣𝑚𝑓𝑡/𝑡𝑓𝑑) = Energy Efficiency = 𝐾𝑝𝑣𝑚𝑓 Time Varying Energy Efficiency Joint Transmission CoMP Operation ( 𝑂 𝐷 ≥ 2 ) 𝐹𝐹(𝑗, 𝑢) 𝐷(𝑗, 𝑢) = 𝑄 𝐷𝑝𝑁𝑄 + 𝑂 𝐾𝑈(𝑗,𝑢) − 1 𝑄 𝐷𝑝𝑁𝑄 − 𝑄 𝐶𝑏𝑡𝑓 𝐹𝐹(𝑗, 𝑢) = 𝐷(𝑗, 𝑢) Single Point Transmission ( 𝑂 𝐷 = 1 ) 𝑄 𝐶𝑏𝑡𝑓 Notes: 1) 𝑄 𝐶𝑏𝑡𝑓 has 𝑄 𝐶𝐼 = 0 since there is not need for multi-point CSI transfer to serving cell 2) 𝑄 𝑇𝑄−𝐷𝑝𝑁𝑄 = 58W since 𝑂 𝐷 = 1 Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 8/19 ICC 2013 June 09, 2013

CoMP Performance Metrics Summary Technical challenges of CoMP systems due to channel estimation errors and system delays: 1. Lowered Joint Transmission Cluster Degree • Inaccurate multi-point CSI feedbacks can exclude a potential joint transmission point from the CoMP cluster unnecessarily. • This decreases both the energy efficiency of the access network and the user perceived quality of service in terms of received downlink data rates. 2. Expanded Joint Transmission Cluster Degree • Inclusion of an incorrect point in the CoMP joint transmission cluster increases the downlink data rates slightly; however, this causes significant bits/Joule energy efficiency losses since the increased power consumption of the access network is not compensated by an equal amount of downlink capacity gain for the served UEs. 9 Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 9/19 ICC 2013 June 09, 2013

    w j l ( ) L w k ( ) j k Decomposed Channel Impulse Response (CIR) Estimation • Track each multipath delay tap of every CoMP measurement set member individually • Perform channel estimation separately for each path • Smoothened CIR at each path is then merged to report CSI feedback for all points  M 1  UE ˆ        w j ( ) w k ( ) j k h ( , t ) w m h t ( ) ( m , ) n i , l l  m 0 𝐽 𝑁 𝑉𝐹 𝑦𝑁 𝑉𝐹 ) −1 𝑠 ℎ ∆𝑢, 𝜐 𝑚 ] 𝐼 2 ℎ 𝑜,𝑗 𝑢, 𝜐 𝑚 = [(𝑆 ℎ ∆𝑢, 𝜐 𝑚 + 𝜏 𝑜𝑝𝑗𝑡𝑓 𝒊 𝑢,…,𝑢−𝑁 𝑉𝐹 +1;𝜐 𝑚 𝑈 𝒊 𝑢,…,𝑢−𝑁+1;𝜐 𝑚 = ℎ 𝑢, 𝜐 𝑚 … . ℎ(𝑢 − 𝑁 𝑉𝐹 + 1, 𝜐 𝑚 ) 𝑆 ℎ ∆𝑢 = 𝑁 𝑉𝐹 − 1, 𝜐 𝑚 = 𝐹 ℎ 𝑢 − 𝑁 𝑉𝐹 + 1, 𝜐 𝑚 ℎ 𝑢, 𝜐 𝑚 ∗ 10 10 10 Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 10/19 ICC 2013 June 09, 2013

Superimposed Channel Impulse Response (CIR) Estimation • Track the superimposed time-varying CIR coefficients instead of separate CIR realizations at each path • Yields less accurate CSI estimates compared to decomposed multipath tracking • Multi-point channel estimation complexity of the UE is decreased significantly • Smaller auto-correlation filter and channel estimation filter input buffers ℎ ∆𝑢 ] 𝐼 2 𝐽 𝑁 𝑉𝐹 𝑦𝑁 𝑉𝐹 ) −1 𝑠 ℎ 𝑜,𝑗 𝑢 = [(𝑆 ℎ ∆𝑢 + 𝜏 𝑜𝑝𝑗𝑡𝑓 𝒊 𝑢,…,𝑢−𝑁 𝑉𝐹 +1 𝑈 𝑀 𝑀 𝒊 𝑢,…𝑢−𝑁+1;𝜐 𝑚 = ℎ 𝑢, 𝜐 𝑚 , … , ℎ 𝑢 − 𝑁 𝑉𝐹 + 1, 𝜐 𝑚 𝑚=1 𝑚=1 𝑀 𝑆 ℎ ∆𝑢 = 𝑆 ℎ ∆𝑢, 𝜐 𝑚 𝑒𝜐 𝑚 𝑚=1 11 11 11 Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 11/19 ICC 2013 June 09, 2013

Performance Analysis of Multi-point Channel Estimation Schemes Decomposed multi-point channel estimation Superimposed multi-point channel estimation • • Better performance for users being served Energy efficiency and capacity degradation by large clusters for users being served by large clusters • • Increased computational burden on the UE Faster channel estimation computation firmware to track each path separately time and decreased CSI feedback delay 12 12 12 Carleton University: G. Cili, H. Yanikomeroglu, F. R. Yu 12/19 ICC 2013 June 09, 2013