Cloud-Based Cell Associations Aly El Gamal Department of Electrical and Computer Engineering Purdue University ITA Workshop, 02/02/16
2 / 23 Cloud Communication Global Knowledge / Control available at Central nodes
3 / 23 Cloud-Based Interference Management Enabling centralized approaches: 1 Learning the network / channel state information 2 Transmission schedules 3 Cell association decisions Application in Multi-RAT Environments
4 / 23 Linear Interference Networks M 1 Tx1 Rx1 ˆM1 M 2 Tx2 Rx2 ˆM2 M 3 Tx3 Rx3 ˆM3 Generic Time Varying Channel
5 / 23 Linear Interference Networks M 1 BS1 MT1 M 1 M 2 BS2 MT2 M 2 M 3 BS3 MT3 M 3 BS: Base Station MT: Mobile Terminal
6 / 23 Cloud-Based Cell Associations M 1 BS1 MT1 M 1 M 2 BS2 MT2 M 2 M 3 BS3 MT3 M 3 Each Mobile Terminal can be associated with N Base Stations
7 / 23 Degrees of Freedom (DoF) DoF = sum capacity lim SNR log SNR Objective: Determine Per User DoF as a function of N. DoF(K, N) PUDoF(N) = lim K K What is the optimal cell association?
8 / 23 What we know For Uplink: PUDoF(N) = 1, N 2 For Downlink: PUDoF(N) = 2N 2N + 1
9 / 23 Uplink: Achieving Full DoF M 1 BS1 MT1 M 1 M 2 BS2 MT2 M 2 M 3 BS3 MT3 M 3 Associating each MT with two BSs connected to it Interference-free Degrees of Freedom
10 / 23 Downlink: Exploiting Global Topology Knowledge M 1 BS1 MT1 M 1 M 2 BS2 MT2 M 2 M 3 BS3 MT3 M 3 M 4 BS4 MT4 M 4 M 5 BS5 MT5 M 5 Achieving 4 5 Per User DoF PUDoF(N) = 2N 2N+1
11 / 23 Average Uplink-Downlink DoF M 1 BS1 MT1 M 1 M 2 BS2 MT2 M 2 M 3 BS3 MT3 M 3 M 4 BS4 MT4 M 4 M 5 BS5 MT5 M 5 Downlink Associations Uplink Associations N = 3 PUDoF = 1+ 4 5 2 = 9 10
12 / 23 Average Uplink-Downlink DoF M 1 BS1 MT1 M 1 M 2 BS2 MT2 M 2 M 3 BS3 MT3 M 3 M 4 BS4 MT4 M 4 M 5 BS5 MT5 M 5 PUDoF(N) = 4N 3 4N 2
13 / 23 Further Questions 1 General network topologies 2 When to simplify into optimizing for uplink / downlink only 3 Constrain average number of cell associations
14 / 23 Next: Transmit Cooperation with no CSIT We know that flexible cell association is useful 1 We do not know whether cooperative transmission is useful 1 Xinping Yi and David Gesbert, Topological Interference Management with Transmitter Cooperation, Dec. 2014
15 / 23 No CSIT: Linear Interference Networks Theorem Transmitter cooperation with no CSIT does not increase the asymptotic per user DoF in linear interference networks PUDoF(N > 1) = PUDoF(1) = 2 3
16 / 23 Converse Proof Channel is time varying with joint pdf Once message is transmitted, it appears at all connected receivers Coordinated Multi-Point transmission cannot be used to cancel interference
17 / 23 Next Tasks Can transmitter cooperation help in any network topology Characterize DoF for general network topologies Extend to other backhaul constraints
18 / 23 Coordinated Learning of Network Topology Earlier work for the braodcast problem 2 Cloud communication can enable some of these ideas 2 Noga Alon, Amotz Bar-Noy, Nathan Linial, David Peleg, On the complexity of radio communication, 1987,1991
19 / 23 Coordinated Learning of Network Topology Lemma For any x 1,, x s n, there exists a prime p s log n such that, x i x j mod p, i, j {1,, s} s : Connectivity parameter n : total number of users
20 / 23 Coordinated Learning of Network Topology Lemma For any x 1,, x s n, there exists a prime p s log n such that, x i x j mod p, i, j {1,, s} 1 Let p 1,, p m be the prime numbers in {1,, s log n} 2 m phases of transmission 3 in i th phase, x j transmits in slot x j mod p i
21 / 23 Coordinated Learning of Network Topology Lemma For any x 1,, x s n, there exists a prime p s log n such that, x i x j mod p, i, j {1,, s} 1 Let p 1,, p m be the prime numbers in {1,, s log n} 2 m phases of transmission 3 in i th phase, x j transmits in slot x j mod p i O(s 2 log 2 n) Communication rounds
22 / 23 Converse? Mimic Probabilistic method for broadcast channel? Slight variation of Group Testing?
23 / 23 Conclusions Cloud-Based Wireless Networks: Enabling centralized approaches new questions and conclusions Value of flexible cell association Coordinated transmission / reception Different learning strategies Benefit with no CSIT / Ad-hoc networks?