NetBeam: Network of Distributed Presenter: Carlos Bocanegra - PowerPoint PPT Presentation

NetBeam: Network of Distributed Presenter: Carlos Bocanegra Full-dimension Beamforming Advisor: Prof. Kaushik R. Chowdhury SDRs for Multi-user Heterogeneous Traffic Next GEneration NEtworks and SYStems Lab

Motivation < > 2 1. + N. users, N. devices and data consumption 3. Novel deployments Aerial on sub-6GHz urban oriented Antenna Array Urban blended UAV oriented Pedestrian Resource oriented Management High mobility Entity 2. Boom of new data services - verticals on the ground oriented 4. Wireless efficiency (IoT)

Enablers < > 3 3 !#$%(") (!#$%(") !#$%(") 1. Beamforming or BF " ! ! !

Enablers < > 4 1. Beamforming or BF 2. Distributed and Collaborative Beamforming (DCBF)

Enablers < > 5 1. Beamforming or BF 2. Distributed and Collaborative Beamforming (DCBF) 3. Distributed Antenna Systems (DAS)

Enablers < > 6 *AAS: Active Antenna Aystem 1. Beamforming or BF 2. Distributed and Collaborative Beamforming (DCBF) 3. Distributed Antenna Systems (DAS) 4. Full dimensional BF or 3DBF

Scenario < > 7 System Resource Management 3-D antenna Entity steering DAS deployment : antennas are coordinated and synchronized via RU. z x Heterogeneous Users : 3D locations and demands (SINR/QoS). Distributed Antenna Array ∅ Multi-User comms : Multiple users served using NOMA. Group 1 Group 2 Group 3 y ! Electronic tilt : Each antenna modifies its azimuth/elevation angles. Goal 3-D distributed users How to select azimuth/elevation angles , group antennas and Disimilar traffic demands 𝑅𝑝𝑇 H 𝑅𝑝𝑇 I 𝑅𝑝𝑇 J beamforming weights to minimize the transmit power while ensuring demanded QoS per user. Total power Formulation 𝒊 𝒋 Channel user i Requested SINR for user i Antenna elevation 𝜄 𝜒 Antenna azimuth 𝒙 𝒋 Beamforming weights user i 𝚳 𝒋 QoS for user I (SINR) Achieved SINR Restriction on available power 𝑸𝒖 𝒐 Maximum Transmit Power for user i PROBLEM 1 : What are the angles 𝜄 , 𝜒 that maximize the channel gain per each antenna pair? PROBLEM 2 : (1) solution is NP-Hard. Need constraint relaxation.

Related Work < > 8 1. Full dimensional beamforming and antenna steering [10] N.Seifi et al., “Adaptive Multicell 3-D Beamforming in Multiantenna Cellular Networks,” Tech. Rep. 8, 2016. [12] N. M. Boers and D. Mak, “Impact of orientation and wire placement on received signal strengths,” in IEEE CAMAD , 2016. 3DBF to increase the sum rate via channel decorrelation or higher channel gains. • • The omnidirectionality of antennas is not perfect and can be exploited. 2. Beamforming in DCBF/ DAS systems [17] G. Sun et al., “A Sidelobe and Energy Optimization Array Node Selection Algorithm for Collaborative Beamforming in Wireless Sensor Networks,” IEEE Access , 2017. [18] B. Bejar Haro et al., “Energy efficient collaborative beamforming in wireless sensor networks,” IEEE Transactions on Signal Processing (TSP) , 2014. • Node placement may not be available in some scenarios, i.e. restricted areas. • Do not encompass dissimilar traffic demands (applications). 3. Antenna selection [8] Y. Gao et al., “Massive MIMO Antenna Selection: Switching Architectures, Capacity Bounds, and Optimal Antenna Selection Algorithms,” IEEE TSP, 2018. [15] Kyungchul Kim et al., “Spatial-Correlation- Based Antenna Grouping for MIMO Systems,” IEEE Transactions on Vehicular Technology, 2010. • Helps reducing the power consumption at the transmitter and rises the channel efficiency. • Antenna selection under dissimilar application demands (QoS/SINR).

Proposed framework - NetBeam < > 9 Distributed Base Stations Resource Distributed Antennas Heterogeneous users Management Entity Antenna steering Beamforming RX Antenna selection Data, control, and signaling Transmitter settings and control Time Sync. Time Sync. 3DBF SDR Time Correction Time Sync. Beamforming TX Channel Estimation MCS selection Feedback interface Beamforming Feedback and control Socket Connection Time correction, Wireless connection programming awareness Channel Estimation

Proposed algorithm - NetBeam < > 10 Ω Sec. VI Λ Sec. V Sec. VII Efficient Digital Antenna Selection EGO-based Antenna 𝒊 𝟐 Beamforming via Matching Orientation 𝑰 ∗ 𝒊 𝟑 Semi-Positive Definite Spatial interpolation: Kriging 𝒊 𝟒 𝑰 T Modified Hungarian alg. 𝑵 ∗ Power minimization at Tx. Prior: - Gaussian fit Binary search space Trial selection: DIRECT-UM User’s app. Demands: SNR 𝒊 𝑶 𝜄 ∗ , 𝜒 ∗ 𝑿 ∗ 𝜄 QRH , 𝜒 QRH 𝑰 T Maximum gain channel NetBeam 𝒊 𝒋 Ω Angular space Channel user i 𝑵 ∗ 𝜄 Λ SNR demands Antenna elevation Assignation matrix Overview 𝑿 ∗ Optimum digital weights 𝑰 ∗ 𝜒 Antenna azimuth Optimum channel *EGO: Efficient Global Optimization

Antenna orientation (AO) < > 11 Distributed Base Stations Resource Distributed Antennas Heterogeneous users Management Entity Antenna steering Beamforming RX Antenna selection Data, control, and signaling Transmitter settings and control Time Sync. Time Sync. 3DBF SDR Time Correction Time Sync. Beamforming TX Channel Estimation MCS selection Feedback interface Beamforming Feedback and control Socket Connection Time correction, Wireless connection programming awareness Channel Estimation

AO – Preliminary studies < > 12 180 Goal: 150 *Angular map : 120 Find the angles that return 𝐵𝑨𝑗𝑛𝑣𝑢ℎ Azymuth and 90 the best RSSI for each user Elevation angles at 60 in the minimum number of the X-Y plane, RSSI 30 trials. at the Z-plane 0 0 15 30 45 60 75 90 𝐹𝑚𝑓𝑤𝑏𝑢𝑗𝑝𝑜 Approach: Sequential decisions problem Gaussian Processes and Kriging A. Model the angular map. Novel DIRECT-UM B. Select next angular tuple to evaluate. C. Reached max trials? -> Return tuple Efficient Global Optimization framework D. Else -> return to A. (EGO) 𝐹𝑚𝑓𝑤𝑏𝑢𝑗𝑝𝑜 𝐹𝑚𝑓𝑤𝑏𝑢𝑗𝑝𝑜 𝐹𝑚𝑓𝑤𝑏𝑢𝑗𝑝𝑜 𝐹𝑚𝑓𝑤𝑏𝑢𝑗𝑝𝑜 180 180 180 180 150 150 150 150 120 120 120 120 𝐵𝑨𝑗𝑛𝑣𝑢ℎ 𝐵𝑨𝑗𝑛𝑣𝑢ℎ 𝐵𝑨𝑗𝑛𝑣𝑢ℎ 𝐵𝑨𝑗𝑛𝑣𝑢ℎ 90 90 90 90 60 60 60 60 30 30 30 30 0 0 0 0 0 15 30 45 60 75 90 0 15 30 45 60 75 90 0 15 30 45 60 75 90 0 15 30 45 60 75 90

AO – Gaussian Processes < > 13 Modeling the angular map 𝜗 j ~𝑂(0, 𝛾 wH ) AWGN characterization: 𝑄(𝑢 j |𝑧 j )~𝑂(𝑢 j , 𝛾 wH z 𝜀) 𝑦 j 𝑧 j 𝑢 j channel AWGN 𝜸 w𝟐 𝒖 𝒐 𝑢 j Channel characterization: 𝑄(𝑧 j |𝑦 j )~𝑂(𝑧 j , 0 z 𝑳) 180 150 𝐵𝑨𝑗𝑛𝑣𝑢ℎ 𝑳 𝑦 j , 𝑦 „ : Kernel or 120 90 covariance 𝑦 j 60 𝑧 j 30 0 0 15 30 45 60 75 90 Channel characterization: 𝑄(𝑢 j )~ ‚ 𝑄 𝑢 𝑧 𝑄 𝑧 𝑒𝑧 𝐹𝑚𝑓𝑤𝑏𝑢𝑗𝑝𝑜 𝑂(𝑧 j , 0 z 𝑳) 𝑢 j = 𝑧 j + 𝜗 j = 𝑔(𝑦 j ) + 𝜗 j 𝑂(𝑢 j , 𝛾 wH z 𝜀) 𝑔(𝑦 j ): The wireless channel 𝑫 𝑦 j , 𝑦 jwH = 𝑳 𝑦 j , 𝑦 jwH + 𝛾 wH z 𝜀 j„ 𝜗 j : Additive White Gaussian Noise (AWGN) 𝜈 𝐷 𝐷 ∗ Find a precise estimate of 𝒖 𝒐 Posterior distribution (1 observation): 𝑄(𝑧|𝑧 † )~𝑂 , ‰ 𝜈 † 𝐷 ∗ 𝐷 † keeping n low 𝒖 𝒐 → {𝒚 𝟏 , 𝒚 𝟐 , 𝒚 𝟑 ,…, 𝒚 𝒐w𝟐 }

AO – Gaussian Processes regression (Kriging) < > 14 Modeling the angular map Multi-observation: Gaussian Process Regression (GPR) 𝒛 𝒐 | {𝒚 𝟏 , 𝒚 𝟐 , 𝒚 𝟑 ,…, 𝒚 𝒐w𝟐 } ? 𝑦 j 𝑧 j 𝑢 j channel AWGN 𝑧 j = 𝑥 H 𝑦 H + 𝑥 H 𝑦 H + ⋯ + 𝑥 H 𝑦 H + 𝑕 + 𝜁 j = 𝐗 𝐔 𝐲 + 𝐡 + 𝜻 180 Kriging-based prediction using 4 trials 150 0.022 𝐵𝑨𝑗𝑛𝑣𝑢ℎ 120 Real measurement Trials (known) 0.02 90 𝑢 › ? Prediction (mean) Empirical channel gain 𝑦 j 60 Uncertainty (variance) 0.018 30 0.016 0 0 15 30 45 60 75 90 𝐹𝑚𝑓𝑤𝑏𝑢𝑗𝑝𝑜 0.014 𝑢 j = 𝑧 j + 𝜗 j = 𝑔(𝑦 j ) + 𝜗 j 0.012 0.01 𝑔(𝑦 j ): The wireless channel 𝑋 H > 𝑋 J > 𝑋 ” > 𝑋 I 0.008 𝜗 j : Additive White Gaussian Noise (AWGN) 0 20 40 60 80 100 120 140 160 180 𝑒 H Azimuth antenna steering 𝑒 J 𝑒 I 𝑒 ” Find a precise estimate of 𝒛 𝒐 keeping n low Kernel selection – The exponential semivariogram 𝒛 𝒐 → {𝒚 𝟏 , 𝒚 𝟐 , 𝒚 𝟑 ,…, 𝒚 𝒐w𝟐 } 𝐿 ℎ 𝑑 † , 𝑏 † ~ 𝑑 † 1 − exp(− ℎ I I ) 𝑏 †