Advanced interference coordination techniques in heterogeneous cellular networks

1
Advanced interference coordination techniques
in heterogeneous cellular networks
Speaker Tingfang Ji

• Collaborator Naga Bhushan, Mohammad Jaber
  Borran, Aamod Khandekar, Ritesh Madan, and Ashwin
  Sampath

ITA Workshop 2010
2
Introduction

• Definition heterogeneous network
• Extension of cellular networks
• Base stations are not homogeneous
• Different transmit power
• Different topology (above rooftop, below rooftop,
  indoor)
• Different access policy (open access, closed
  access)
• Why are heterogeneous networks interesting?
• Heterogeneous networks provide flexible coverage
  enhancement
• Pico/femto/relay cells provide coverage in areas
  with insufficient macro coverage
• Traffic growth is outpacing macro network growth
• Heterogeneous networks offload traffic from macro
  networks
• Lower /bit cost

3
Challenges and Solutions

• Severe interference issue
• Example
• Closed femto cells only serves subscribed users
• A macro cell mobile could get very close to a
  femto cell but not receiving service form the
  femto cell
• Interference from the femto could be much higher
  than the macro signal
• Interference from the non-member mobile could be
  much higher than a member mobile signal.
• Similar issues exist for pico cells and relays
  with aggressive load balancing
• Solutions
• Interference cancellation
• Power control
• TDM/FDM partitioning
• Cooperative beamforming
• Coherent joint transmission

Femtos
Macro
4
Receiver Techniques

• Interference cancellation
• In theory, interference cancellation could be
  used to peel off dominant interferers.
• In practice, unicast data from interfering cell
  is hard to decode due to scrambling, adaptive
  modulation and coding, HARQ retransmissions.
• In practice, interference cancellation is very
  effective for broadcast information
• Acquisition
• Synchronization and broadcast signals from
  dominant interferer could be decoded and
  cancelled
• 3GPP 4G (OFDM) system have synchronization
  signals interfering with each other. Hence
  cancellation of interfering synchronization
  signals would allow the acquisition of the
  desired cell.
• Pilot
• Common pilot are used for channel estimation
• Pilots from interfering cell could also be
  decoded and cancelled
• Data
• After all the overhead channels have been
  cancelled, only unicast data is left unprotected.

5
Sensitivity to Topology and Fairness

• Tradeoff between different schemes changes
  drastically with topology
• Fairness requirements also changes the tradeoff

Rate region for two interfering links for high
interference case
Rate region for two interfering links for medium
interference case.
6
Maximizing Sum Rate is Not the Ultimate Goal

• Utility function models perceived value of
  allocated rate to a user

Utility log(R1) 5log(R2) Optimal rate
pair via power control
Utility 2log(R1) log(R2) Optimal rate
pair achieved by time-sharing user 1 with power
controlled point of user1, user 2
7
Semi-static Resource Partitioning

• Optimization problem

• ?j,r the fraction of resource r that is assigned
  to user j (by the scheduler at the serving node
  S(j)), r 1, , Nr denotes the resource, ? the
  spectral efficiency of a link

• p (pi,r) where pi,r?Pi, i 1, , N denotes the
  transmitting node, P is the transmitting power

• Joint optimization of the resource partitioning
  and load balancing
• Allow mobile to connect to a very weak cell
  thats lightly loaded
• Used in conjunction of interference mitigation

8
Capacity and Fairness Improvements with
Interference Mitigation

• Iterative algorithm was defined to solve the
  optimization problem
• Assumptions 3GPP LTE (OFDM) system 57 macro
  cells (40 Watts), embedded with 228 pico cells (1
  Watt), 25 mobiles/cell

9
Dynamic Interference Avoidance

• Short term interference avoidance (e.g., on a
  packet-by-packet basis)
• Enables fast coordination for bursty and
  latency-sensitive traffic.
• Optimizes capacity and user experience.
• Uses simple over-the-air or over-the-backhaul
  message.
• Define Priority Metrics
• QoS, HTTP pkt_delayrate (i.e., assume same
  delay target across cells)
• Best effort rate/avg. rate
• Simulations
• Apartment building with 25 units, on average 5
  units have femto cells, one mobile in each femto
  cell.
• Mixed background downloading and HTTP traffic,
  75 HTTP, 25 background download.
• HTTP
• pkt inter-arrival times geometric with
• mean 1 ms
• pkt Size 6KB
• call size is Pareto distributed
• min 30 KB, max 10 MB, mean 200 KB
• reading time geometric with mean 4 seconds

10
Cooperative Beamforming

• The baseline
• Every active femto schedules its mobile at every
  scheduling instance
• Eigen-beamforming with equal power distribution
  across layers
• Dynamic rank selection based on the maximum
  spectral efficiency.
• CB scheme
• Spatial coordination information is sent from
  mobile to the neighboring cell
• Compressed CSI and utility
• Neighboring femto choose from two options
• Coordinated silencing
• Signal-to-leakage ratio (SLR) beamforming
• Optimize local utility
• Significant gain at tail with slight loss at mean

11
Other Practical Constraints

• Coherent processing (joint transmission) is
  optimal at a high cost
• Backhaul load could be orders of magnitude higher
• Backhaul latency requirement is high
• Sum power optimization is useful, but peak power
  constraint has to be enforced at each node
• Can not move power between different nodes,
  especially considering heterogeneous networks
  with vastly different power level
• Orthogonalization in frequency provides limited
  protection
• Adjacent channel interference (Tx leakage and Rx
  sensitivity) is regulated by FCC and standards
  bodies
• It is often found to be insufficient in dominant
  interference scenarios
• Spectrum availability
• TDD spectrum requires tight coordination between
  adjacent carriers
• FDD full duplex devices requires large frequency
  separate due to self-dense (Tx, RX desense)

12
Conclusions

• Heterogeneous networks provides capacity/coverage
  gain compared to conventional macro networks,
  while introducing severe interference due to the
  low power nodes and restricted association
• Investigated an array of techniques
• Interference cancellation
• Resource partitioning and adaptive association,
• Dynamic packet-by-packet negotiation for QoS and
  capacity optimization
• Coordinated beamforming,
• Lesson learnt
• Optimality depends heavily on topology, fairness
  and QoS
• Joint optimization of resource allocation and
  association provides significant gain
• RF planning is difficult for unplanned networks,
  distributed iterative solutions are more robust
  and efficient