Increasing Capacity of Cellular WiMAX Networks by Interference Coordination

Universität Stuttgart INSTITUT FÜR KOMMUNIKATIONSNETZE UND RECHNERSYSTEME Prof. Dr.-Ing. Dr. h. c. mult. P. J. Kühn Increasing Capacity of Cellular WiMAX Networks by Interference Coordination Marc Necker Institute of Communication Networks and Computer Engineering University of Stuttgart, Germany marc.necker@ikr.uni-stuttgart.de ITG FG 5.2.1, Essen November 22, 2007

Outline Introduction and motivation - Requirements and challenges in cellular networks - Introduction to OFDMA networks Interference mitigation techniques - Fractional Frequency (FFR) - Interference Coordination (IFCO) Coordinated Fractional Frequency - Concept and architecture - Algorithm description Performance Evaluation - Comparison with conventional systems

Motivation Scenario Cellullar OFDMA network according to 3GPP Long Term Evolution (LTE) or IEEE 802.16e (WiMAX) Requirements High aggregate throughput serve as many users as possible High cell edge throughput good performance even with weak signal Major problem: Inter-cellular interference

Orthogonal Frequency Division Multiple Access Based on Orthogonal Frequency Division Multiplex (OFDM) - subdivision of frequency spectrum into subcarriers - well suitable for multi-path fading environments f Basis of several emerging cellular standards e.g., 802.16e/m (WiMAX), 3GPP LTE Example: 802.16e MAC Layer ("mobile WiMAX") Frequency-diverse (PUSC zone, FUSC zone) and frequency-selective modes (AMC zone) AMC zone (Adaptive Modulation and Coding) - allocation of consecutive subchannels for the transmission to one terminal - allocations have rectangular shapes allows frequency-selective scheduling well suitable for interference coordination MAC frame Terminal 1 Terminal 3 Terminal 4 Terminal 2 Terminal 6 Terminal 5 T 7 t

Interference in Cellular Networks Major issue in OFDMA: inter-cellular interference

Interference in Cellular Networks 3 Major issue in OFDMA: inter-cellular interference - standard solution: frequency reuse pattern disadvantage: waste of precious frequency resources

Interference in Cellular Networks FFR 1 area 3 area Major issue in OFDMA: 1 inter-cellular area interference 3 area - standard solution: frequency reuse pattern disadvantage: waste of precious frequency resources - optimization: Fractional Frequency (FFR)

Interference in Cellular Networks Major issue in OFDMA: inter-cellular interference - standard solution: frequency reuse pattern - optimization: Fractional Frequency (FFR)

Interference in Cellular Networks Major issue in OFDMA: inter-cellular interference - standard solution: frequency reuse pattern - optimization: Fractional Frequency (FFR) - Usage of directional antennas to lower inter-cellular interference Additional coordination necessary interference coordination (IFCO)

Conventional Fractional Frequency (FFR) 1 Area 3 Area f Partition 3 All Resources Partition 2 Partition 1 1 & reuse 3 areas may or may not be on same frequency range AMC zone t AMC zone t Power levels may or may not be adjusted depending on area Assignment of mobiles to reuse 1 or 3 based on position or SINR

Coordinated Fractional Frequency 1 Area 3 Area f Partition 4 Partition 8 Partition N C All Resources Partition 3 Partition 7... Partition N C 1 Partition 2 Partition 6 Partition N C 2 Partition 1 Partition 5 Partition N C 3 AMC zone AMC zone AMC zone AMC zone virtual frame duration Idea: Reduce interference by optimized and coordinated dynamic choice of reuse partition (semi static or dynamic) interference coordination

System Architecture Coordinator base station Base stations communicate relevant information to central coordinator Central coordinator assigns mobile terminals to resource partitions in a coordinated fashion

Coordination of Resource 3 Partitions Approach - construction of an interference graph G in central coordinator nodes m i M edges e ij E (non-directional) - assignment of resource partitions based on interference graph - communication of resource partitions to base stations Interference graph - based on global knowledge collected from all base stations - edges represent critical interference relations in-between terminals connected terminals should not be served on the same resource (time/frequency slot)

Creation of Interference Graph cell border mobile terminal m 5 mobile terminal m 10 interference mobile terminal m 12 mobile terminal m 2 Interference by Mobile Terminal m 5 m 8 m 10 m 9 m 42 Interference Level -83 dbm -89 dbm -91 dbm -92 dbm -94 dbm

Creation of Interference Graph cell border Interference by Mobile Terminal mobile terminal m 5 Interference Level Calculation of signal strength of interferers for a particular mobile terminal m j m 2 m 12 m 8 m 20-80 dbm -93 dbm -94 dbm -99 dbm mobile terminal m 10 m 42 m 35-99 dbm -99 dbm interference mobile terminal m 12 mobile terminal m 2 Interference by Mobile Terminal m 5 m 8 m 10 m 9 m 42 m 30 Interference Level -83 dbm -89 dbm -91 dbm -92 dbm -94 dbm -98 dbm

Creation of Interference Graph blocked by interference graph cell border Interference by Mobile Terminal m 2 m 12 m 8 m 20 m 42 mobile terminal m 5 Interference Level -80 dbm -93 dbm -94 dbm -99 dbm -99 dbm interference mobile terminal m 10 Calculation of signal strength of interferers for a particular mobile terminal m j Blocking of strongest interferers such that a desired minimum SIR D S is achieved m 35-99 dbm mobile terminal m 2 mobile terminal m 12 Interference by Mobile Terminal m 5 m 8 m 10 m 9 m 42 Interference Level -83 dbm -89 dbm -91 dbm -92 dbm -94 dbm blocked by interference graph m 30-98 dbm

Creation of Interference Graph blocked by interference graph cell border Interference by Mobile Terminal m 2 m 12 m 8 m 20 m 42 m 35 mobile terminal m 5 Interference Level -80 dbm -93 dbm -94 dbm -99 dbm -99 dbm -99 dbm interference mobile terminal m 10 Calculation of signal strength of interferers for a particular mobile terminal m j Blocking of strongest interferers such that a desired minimum SIR D S is achieved Blocked terminals are connected by edge in interference graph mobile terminal m 2 mobile terminal m 12 Interference by Mobile Terminal m 5 m 8 m 10 m 9 m 42 Interference Level -83 dbm -89 dbm -91 dbm -92 dbm -94 dbm blocked by interference graph m 30-98 dbm

Assignment of Resource Partitions Example of resource mapping f frame frame t Treat resource partitions as colors of graph Resource partitions can be assigned to mobile terminals by coloring of the interference graph - graph coloring is NP hard - large number of heuristics: genetic algorithms, simulated annealing, tabu search, other heuristics (e.g., Dsatur)

Mapping of colors to Resource Partitions Partition N C -1 Farbe c n,3 Partition 3 Farbe c 0,3 Partition 7 Farbe c 1,3 Partition N C -1 Farbe c n,3 Partition 3 Farbe c 0,3... Partition N C -2 Farbe c n,2 Partition N C -3 Farbe c n,1 Partition 2 Farbe c 0,2 Partition 1 Farbe c 0,1 Partition 6 Farbe c 1,2 Partition 5 Farbe c 1,1... Partition N C -2 Farbe c n,2 Partition N C -3 Farbe c n,1 Partition 2 Farbe c 0,2 Partition 1 Farbe c 0,1... Partition N C -4 Farbe c n,0 Partition 0 Farbe c 0,0 Partition 4 Farbe c 1,0 Partition N C -4 Farbe c n,0 Partition 0 Farbe c 0,0 MAC frame MAC frame MAC frame virtual frame duration Virtual frame duration must be adapted to number of colors

Signaling-Time-Diagram coloring valid T C,period Basestation Basestation Basestation T C,delay local state information global coloring Coordinator processing Procedure Communication of all required information to central coordinator Calculation of interference graph Graph Coloring Communication of colors to base stations Mapping of colors to resource partitions coloring valid local state information global coloring processing Important Parameters update period: T C,period delay: T C,delay t t

Performance Evaluation Scenario Event-driven simulation model implemented using IKR SimLib Hexagonal scenario described before with wrap-around mobility model - 9 mobile terminals per cell sector - 30 km/h, random direction mobility model Traffic model - greedy traffic sources in downlink direction - throughput measured at IP level Detailed MAC and Physical layer model with path loss and shadowing Metrics: Aggregate sector throughput does not take into account fairness towards cell edge users 5 % quantile of the individual throughputs of all mobiles - terminals close to cell center have high throughput - terminals close to cell edge have low throughput corresponds to throughput of terminals close to cell edge

Throughput Performance 500 450 5% throughput quantile [kbit/s] 400 350 300 250 Frequency reuse 3 system 200 150 800 900 1000 1100 1200 1300 1400 1500 aggregate sector throughput [kbit/s] 3 system achieves good aggregate performance and good cell edge performance

Throughput Performance 500 450 5% throughput quantile [kbit/s] 400 350 300 250 200 Frequency reuse 3 system Frequency reuse 1 system 150 800 900 1000 1100 1200 1300 1400 1500 aggregate sector throughput [kbit/s] 1 system achieves better aggregate performance but falls short with respect to cell edge performance

Throughput Performance 500 450 5% throughput quantile [kbit/s] 400 350 300 250 200 Frequency reuse 3 system Frequency reuse 1 system Fractional Frequency 150 800 900 1000 1100 1200 1300 1400 1500 aggregate sector throughput [kbit/s] Conventional Fractional Frequency, locally coordinated - achieves great increase in aggregate performance - falls short with respect to cell edge performance

Throughput Performance 500 450 Coordinated Fractional Frequency 5% throughput quantile [kbit/s] 400 350 300 250 200 Frequency reuse 3 system Frequency reuse 1 system Fractional Frequency 150 800 900 1000 1100 1200 1300 1400 1500 aggregate sector throughput [kbit/s] Coordinated Fractional Frequency - achieves good increase in aggregate and cell edge performance - allows to trade off cell edge and aggregate performance on a high level

Impact of Signaling Delays 340 320 T C,delay = 0 ms 5% throughput quantile [kbit/s] 300 280 260 240 220 200 180 0 500 1000 1500 2000 2500 3000 3500 4000 T C,period [ms] Increased signaling delay T C,period - leads to graceful degradation of cell edge performance - has much less impact on aggregate performance (not shown here)

Impact of Signaling Delays 5% throughput quantile [kbit/s] 340 320 300 280 260 240 220 T C,delay = 0 ms T C,delay = 1000 ms T C,delay = 2000 ms T C,delay = 4000 ms 200 180 0 500 1000 1500 2000 2500 3000 3500 4000 T C,period [ms] Increased signaling delays T C,period and T C,delay - lead to graceful degradation of cell edge performance - have much less impact on aggregate performance (not shown here)

Area Throughput 3 Coordinated FFR kbit/s 4000 3500 3000 2500 2000 1500 1000 500 0 140 y[pixel] 120 100 80 60 160 kbit/s 4000 3500 3000 2500 2000 1500 1000 500 0 140 120 y[pixel] 160 100 80 60 40 40 20 20 0 0 0 20 40 60 80 100 x[pixel] 120 140 160 0 20 40 60 80 100 x[pixel] 120 140 160 TC,period = 2s, TC,delay = 1s DS,o = 0dB, DS,i = 20dB Big increase close to base stations Good coverage at cell edge with coordinated FFR Institut für Kommunikationsnetze und Rechnersysteme Universität Stuttgart

Conclusion Frequency spectrum is one of the most precious resources operators strive to get maximum performance out of limited spectrum Possible solutions - denser planning of base station grid high additional cost - deployment of advanced algorithms, such as interference coordination capacity improvements achievable by much lower cost Coordinated Fractional Frequency - algorithm for distributed and dynamic interference coordination - low complexity scheme based on central coordinator - performance improvements of about 50% (compared to 3) communication with central coordinator in intervals in the order of 500 ms with respect to aggregate throughput (maintaining cell edge throughput) with respect to cell edge throughput (maintaining aggregate throughput)