Adaptive Frequency Reuse in IEEE 802.16m
S80216m-08_702r3
Document Number:
S80216m-08/702r3
Date Submitted:
2008-07-17
Source:
Clark Chen, Hongmei Sun, Hua Yang
Email: {clark.chen, hongmei.sun, hua.yang, shilpa.talwar,
Shilpa Talwar, Vladimir Krasvtov,
vladimir.kravstov, yuval.lomnitz, himayat.nageen,
Yuval Lomnitz, Hujun Yin
hujun.yin} @intel.com
Intel Corporation
Kwanhee Roh, Jaehee Cho, Jeongho Park, E-mail: kwanhee.roh@samsung.com
Sooryong Jung
Samsung Electronics
Venue:
IEEE 802.16m-08/024, Call for Comments and Contributions on Project 802.16m System Description
Document (SDD), Interference Mitigation
Base Contribution: C80216m-08/702r3
Purpose:
For discussion in TGm and adoption of proposed SDD text into the IEEE 802.16m System
Description Document
Notice:
This document does not represent the agreed views of the IEEE 802.16 Working Group or any of its subgroups. It represents
only the views of the participants listed in the “Source(s)” field above. It is offered as a basis for discussion. It is not binding on
the contributor(s), who reserve(s) the right to add, amend or withdraw material contained herein.
Release:
The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any
modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards
publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to
reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this
contribution may be made public by IEEE 802.16.
Patent Policy:
The contributor is familiar with the IEEE-SA Patent Policy and Procedures:
<http://standards.ieee.org/guides/bylaws/sect6-7.html#6> and
<http://standards.ieee.org/guides/opman/sect6.html#6.3>.
1
Outline
•
•
•
•
•
•
Motivation
AFR Requirements
AFR Architecture
AFR Algorithm
Standard Support
Conclusions and recommendations
2
IEEE 16m SRD Requirements
Meet 802.16m downlink Data requirement
3
Interference is serious problem with Reuse 1
Empirical CDF
1
Reuse1
Reuse3/2
Reuse3
0.9
0.8
0.7
F(x)
0.6
0.5
0.4
0.3
0.2
0.1
0
-10
Geometric SINR distribution in hexagon cell structure
•
•
•
•
0
10
20
30
Geometry SINR (dB)
40
50
60
Geometric SINR under different reuse factors
More than 30% of subscribers have average SINR below 0dB
Worst SINR conditions seen at cell-edges
Higher reuse helps improve SINR, but lowers cell capacity
Solution: mixed reuse to tradeoff cell-edge performance with cell-capacity
4
Adaptive Frequency Reuse Concepts
RRM
Interference-aware BS
Coordination
AFR
FFR (semi-static)
MAC
FFR (dynamic)
Interferenceaware
Scheduling
Uplink
Power
Control
Interference Randomization
PHY
Tx Beamforming
Rx Interference
Mitigation
5
AFR Components
• Fractional frequency reuse (reuse partitioning): Restrict usage of
certain frequencies or power level of these frequencies in a sector
to minimize interference. Ex. allocate fraction of frequency subchannels to reuse 3 for cell-edge users
• Interference aware scheduling: Scheduler allocates resources to
users based on FFR partition and interference aware CQI metrics
• Interference-aware BS co-ordination: Base stations dynamically
adjust FFR partition by exchanging information across the network
backbone, to adapt to time-varying user traffic loads & distributions
a) Static - no information exchange
b) Semi-static - on order of 100msec to seconds
c) Dynamic - on order of frame duration
6
AFR Design Requirements
•
•
•
•
•
•
•
•
Support multiple reuse settings: 1, 3, 3/2
Support diversity & contiguous permutation modes
Support hard reuse (AFR-H) and soft reuse (AFR-S)
Flexibility with non-uniform user distributions
Adaptation to time-varying traffic conditions
Exploit channel aware scheduling gains
Robustness to mobile environments
Low system complexity
7
AFR Architecture
Power
P1-3
P1-1
Power
P1-2
P2-2
P2-1
Power
P2-4
P2-6
P3-4
P3-2
P3-3
w2
w3
Reuse M
P1-6
P2-5
P2-3
P3-1
w1
P1-5
P1-4
w12
Reuse N
P2-7
P3-6
P3-7
w13
w123
P3-5
w23
P1-7
Sector 1
Sector 2
Sector 3
Reuse 1
Frequency
•
•
•
•
Reuse partition: N = [K] partitions (to support reuse 1, N, M)
Attributes of partition: A triplet of N-dim vectors: [W, P, C]
– Bandwidth partition W: W  [W1 ,W2 ,W3 ,W12 ,W23 ,W13 ,W123 ]
P  [ P1 , P2 , P3 , P12 , P23 , P13, P123 ]
– Power level P:
C  [C1 , C2 , C3 , C12 , C23 , C13, C123 ]
– System cost C:
Soft reuse achieved by setting power level of each partition group
Optimal resource allocation achieved by setting system-wide Cost for each
partition
8
Various deployments with AFR Architecture
Supported reuse factor
Number of partition groups
/(corresponding size vector)
Reuse 1
1
/
[0, 0, 0, 0,
Reuse 3/2
3
/
[0, 0, 0, W12 , W23 , W13 , 0 ]
Reuse 3
3
/
[W1 , W2 , W3 , 0,
2 mixed
reuse
partitions
Reuse 1 and reuse 3/2
4
/
[0, 0, 0, W12 , W23 , W13 , W123 ]
Reuse 1 and reuse 3
4
/
[W1 , W2 , W3 , 0, 0, 0, W123 ]
3 mixed
reuse
partitions
Reuse 1, reuse 3/2, reuse
3
7
Single reuse factor
system
Hard Reuse
AFR
syste
m
Example:
Reuse 1 and soft
reuse 3 AFR system
Soft Reuse
Example:
Reuse 1 and soft
reuse 3/2 AFR
system
/
0,
0, W123 ]
0, 0,
0 ]
[W1 ,W2 , W3 , W12 ,W23 ,W13 ,W123 ]
Power level of Sector 1
P  [ PLow , PLow , PHigh ,0,0,0, PRe use1 ]
Power level of Sector 2
P  [ PHigh , PLow , PLow ,0,0,0, PRe use1 ]
Power level of Sector 3
P  [ PLow , PHigh , PLow ,0,0,0, PRe use1 ]
Power level of Sector 1
P  [0,0,0, PLow , PLow , PHigh , PRe use1 ]
Power level of Sector 2
P  [0,0,0, PHigh , PLow , PLow , PRe use1 ]
Power level of Sector 3
P  [0,0,0, PLow , PHigh , PLow , PRe use1 ]
9
AFR Algorithm Details
•
•
•
•
Problem definition and theory introduction
Cost and BW partition adaptation
Performance
Summary
10
Problem and Definition
•
Optimal resource allocation problem:
–
–
–
–
•
How power loading level and AFR partition size is adapted?
What measurements are needed?
What information is fed back from SS?
How SS’s are scheduled on different resource types?
Definitions
– Radio Resource Type: FFR partition with different average S/I
– Cost: a real value that is a measurement of system resources used by
a particular resource type
– Normalized SE: a value represents the normalized efficiency
achievable on particular resource type in terms of system resource.
– User Distribution: position and corresponding S/I distribution of users
(SS) in the system
– Fairness Constrain: a pre-defined percentage curve that specifies the
throughput CDF of all SS
* Complete mathematic proof in a separated document
11
Optimal Resource Allocation Theory Solution
• Assumptions
– The user distribution and corresponding average signal/interference
level doesn’t change during the optimization time.
– A user’s average spectrum efficiency is a rising function of their average
SINR at different radio resource types.
• Theory Solution
– Given a user distribution  and a fairness constraint , for every power
loading factor , there is a optimal resource allocation solution
achievable, that yields the highest average SE, with a unique (W , C ),
where the system partition is proportional to SS’s relative bandwidth
request on different resource types, and all SS are allocated resources
that yield maximum normalized SE amongst all resource types.
opt
opt
• Reference:
–
Problem of optimal radio resource allocation, by Vladimir Kravstov, Intel (unpublished doc)
12
AFR Scheme Details – Theory Solution Exemplary
• Intuitive example: AFR system with Hard Reuse 3
– Cost of reuse-3 is 3 while cost of reuse-1 is 1
• Each resource block in Reuse 3 occupies 3 times system resource compared
with resource block in Reuse 1
– SS should be allocated with FFR partition that yields highest nSE
• Only SS that can get more than 3 times SE gain should use Reuse 3
– The optimal Reuse partition would be decide by all SS’ choice of reuse
partition and their corresponding bandwidth required (fairness constrain)
• Key learning
– Cost of different reuse partition is different
• Must find optimal cost vector online
– Reuse Partition should be proportional to SS’ relative bandwidth request
• Must adapt online
* Complete mathematic proof in a separated document
13
AFR Implementation and Procedure
• Given SS distribution and S/I, find the optimal {P, W, C}
• Power loading
– predefined according to engineering experience or network planning
• Initial Partition
– Enable SS measurement on different AFR partition
• Initial Cost
C  [C1 , C2 , C3 , C12 , C23 , C13, C123 ]
– Can start from any real value or choose on engineering experience
• Two steps adaptation
– cost adaptation and channel partition adaptation.
– the cost adaptation is a necessary step for partition adaptation.
* Complete mathematic proof in a separated document
14
AFR Scheme Details – Cost adaptation
•
•
•
•
•
‘Market Price Iteration’ algorithm
to find the ‘optimal’ C
incrementally
Decreases the cost values if there
are too many/few bandwidth
requests from SS for the
corresponding channel partitions
Target is to achieve balanced
bandwidth request on three reuse3 partitions
Theory proves the convergence
and unique of cost vector
BS autonomous decision process
in cost adaptation, and deals good
with un-balanced SS distribution
among BS
SS
BS
Initial Price
Updated cost vector
Broadcast
SINR measurement
for different reuse
partition
Estimate the SE of
different subchannel
Calculate the nSE by:
SE/cost
Preferred subchannel and reuse
choice
Preferred feedback
by CQI
Collect bandwidth
request of different
reuse
15
AFR Scheme Details – BW Partition adaptation
•
•
•
•
•
•
Require the cost converged in each
BS individually
BS report all SS preference of
different reuse partition to RRM
RRM take system wide fairness rule in
addition to the relative bandwidth
request
RRM (function) unit takes charge of
BW partition changes for all
neighboring BS in a system
Partition adaptation is relatively slow,
from tens of minutes or hours, on
carrier’s choice, depending on system
dynamic
Cost/Partition adaptation approaches
optimal system operation point
RRM Unit
BS
Cost converged of
diff reuse partition
Report signalling
Statistic bandwidth
request of different
reuse partition
Collect all bandwidth
request on different
reuse partition
Notice the same
trend of change in
system
Adjust system wide
AFR configuration
Signalling
All BS change AFR
configuration in sync
16
AFR Scheme Details – Soft Reuse with AFR Architecture
w123
w1
w2
w3
1,2,
3
4
1
2
3
4
2
1,3,
4
1,2,
3
4
2
1,3,
4
3
3
1,2,
3
4
2
1,3,
4
1,2,
4
1,2,
3
4
1,2,
4
2
1,3,
4
3
1,2,
3
4
1,2,
4
2
1,3,
4
3
3
1,2,
3
4
1,2,
4
1,2,
3
4
2
1,3,
4
1,2,
4
2
1,3,
4
3
3
1,2,
4
1,2,
4
High Power
AFR-S: AFR with reuse 1 and soft reuse 3
17
AFR Scheme Details – AFR Performance
Scheme
Channel Model
Gross SE
(bps/Hz/cell)
AMC baseline
PedB-3kmph
5.84
AMC AFR-S
Dist-baseline
Dist AFR-S
5.93
PedB-3kmph
Gain on SE
Gain on cell
edge user
throughput
667k
1.6%
3.68
4.52
Cell edge
user throughput
689k
standard PF
3.3%
315k
22.8%
595k
Note
standard PF
standard PF
88.9%
standard PF
• With standard proportional fair scheduling, AFR soft reuse provides
– limited gain on both cell edge and cell capacity in localized (AMC-like)
resources
– 89% gain on cell edge with +23% gain on cell capacity in distributed
resources
* Initial data and due to change according to updates
18
AFR Scheme Details – AFR Performance (continued)
Scheme
Channel Model
SE
(bps/Hz/Cell)
baseline
PedB-3kmph
5.84
gain on SE
cell edge user
Throughput
(kbps)
gain on cell edge
user throughput
667
Note
standard PF
AFR-S
5.93
1.6%
689
3.3%
standard PF
AFR-S
5.71
-2.2%
772
15.7%
Weighted PF
AFR-S
5.29
-9.4%
819
22.8%
Weighted PF
AFR-S
5.18
-11.2%
843
26.4%
Weighted PF
• AFR can further improve cell edge user throughput at the expense of
cell capacity loss by changing scheduling policy
19
AFR Summary
• Theory indicates the optimal (power loading, partition, cost)
combination that AFR system achievable
• AFR architecture approaches the optimal solution point with selflearning process
• AFR architecture well deal with the dynamic facts in real system with
un-predictable SS distribution and changing propagation
environment
• AFR architecture is simple, scalable, and has light overhead
20
Standard Support
•
•
•
•
Symbol Structure
Signaling Support
Measurement and feedback
Inter-BS Co-ordination
21
AFR Implementation Details
1) FFR partition needs to be inherent in Symbol Structure
(2) Renumber PRUs
for each Freq
partition
Localized
Distributed
Subcarrier
permutation
00
01
02
03
04
05
06
07
08
09
...
Localized
Freq.
Part3
Distributed
Resource groups
Inter-cell (semi static)
(4) Distribute
subcarriers to
subchannels (LRU)
(3) Distribute PRUs to
localized (LLRU) and
distributed (DRU) resources
Freq. Part2
(Outer)
Permutation
of PRU to
Freq.
Partitions
Reordered PRU
Physical frequency
Freq. Part1
(1)
Distribute PRU to Freq
Partitions
Subcarrier
permutation
Localized
Single resource
Intra-cell (potentially dynamic)
Symbol structure with localized and distributed resources per FFR partition
22
AFR Implementation Details
2) Downlink Signaling
– AFR system configuration needs be broadcast in BCH/SFH or DL MAP
– AFR information includes Bandwidth Partition, Power Level, and System
Cost of each partition
– Needed for each SS in initial entry process and measurement
3) CQI Indication:
– CQI feedback is needed to support frequency selective scheduling
• For example, best-M CQI feedback scheme
– CQI is interference-aware, ex. post-processing SINR after LMMSE
23
AFR Implementation Details
4) Measurements:
– Long-term measurements: Geometric SINR on different reuse
partitions is needed to select best partition for each SS
• Can be computed from AFR-friendly preamble
– Short-term measurements: Instantaneous SINR (localized or
distributed) on different partitions is needed to support frequency
selective scheduling
• Can be measured from dedicated pilots in symbol structure design
• Symbol structure should support boosted pilots proportional to data
for soft reuse
24
AFR Implementation Details
5) Inter-BS Co-ordination
– BS ex-change information for system BW adaptation by:
•
•
•
•
Bandwidth partition: W,
Power level: P
Cost vector: C
Load on different FFR partition, etc
– A RRM (function) unit should decide whether the BW partition
should be adapted according to fairness constrain and
performance metric (system SE and/or edge throughput)
– The adaptation of system wide AFR configuration can be
dynamic, semi-static or static
25
Conclusions and Recommendations
Conclusions
– Adaptive Frequency Reuse (AFR) can effectively improve cell edge
performance while keep or gain in SE
– Adaptive Frequency Reuse (AFR) provides very flexible solution to
dynamic deployment environments which approaches the optimal
resource allocation in system
– Symbol structure, Signaling design, and Measurement and feedback
support are needed to implement AFR
We would like to recommend the following to be incorporated in SDD
to support AFR
–
–
–
–
Support Adaptive Frequency Reuse partition framework
Support Symbol structure design that accommodates AFR framework
Support Downlink Signaling necessary for AFR
Support Measurements and feedback necessary for AFR, including
preamble and dedicated pilots
26
Proposed Text
•
•
•
11.x Inter-cell interference mitigation
11.x.1 Flexible Frequency Reuse
Flexible Frequency Reuse can be used to help users suffering from severe inter-cell interference. With
FFR enabled, subcarriers across the whole frequency band are grouped into [K] frequency partitions.
Associated with partitions are the parameters of power and cost. The usage of frequency partitions is
limited in each BS such that parts of frequency partitions can be unused or have lower/higher
transmission power. There is a system cost associated with partitions which is a measurement of the
system resource used by the partitions due to its different transmission power, in terms of
interference caused to other cells. For example, the cost of a partition is high if this partition is
restricted in other sectors to create higher reuse pattern (3, 3/2), or if the partition uses higher
transmission power and causes interference to neighbor cells. To control system-wide interference, the
BS can adjust partition and power in coordination with other BSs.
27
Proposed Text (Cont’d)
The following diagram shows an example of frequency partition and power
control to mitigate the interference in DL. The system may have [K] frequency
partitions in mixture of reuse-1, reuse-M, reuse-N according to deployment needs.
Power
P1-3
P1-1
Power
P1-2
P2-2
P2-1
Power
P2-4
P2-6
P3-4
P3-2
P3-3
w2
w3
Reuse M
P1-6
P2-5
P2-3
P3-1
w1
P1-5
P1-4
w12
Reuse N
P2-7
P3-6
P3-7
w13
w123
P3-5
w23
P1-7
Sector 1
Sector 2
Sector 3
Reuse 1
Frequency
Figure x: Example of Flexible Frequency Reuse
28
Proposed Text (Cont’d)
11.x.2 Control signaling for inter-cell interference mitigation
BS can transmit necessary information through signaling channel or message to
MS to support Flexible Frequency Reuse. The necessary information includes
frequency partition and additional partition information such as power and cost.
MS can transmit measurements through signaling channel or message to BS to
support Flexible Frequency Reuse. The necessary information includes CINR
measurements (long-term and instantaneous) or their representatives. The
feedback from MS across the frequency partitions should be minimized.
BS can coordinate and exchange information with neighbor BSs to support
Flexible Frequency Reuse through the backbone network. The necessary
information includes frequency partition, cost and interference level.
29
Q&A
30
Backup - System Simulation Assumptions
Number of cells
19 cells, with 3 sector per cell
(with wrapping around)
Number of sectors per cell
3
Site-to-site distance
1.5km
Carrier Frequency
2.5GHz
User Number
10 users/sector
Permutation mode
AMC, PUSC
MIMO
2x2 with rank adaptation (STBC/SM) with
MMSE aware receiver
Repetition
ON
Target PER
0.1
Strong interference number
14
Channel model
PedB 3kmph
CQI feedback
Full feedback with 5ms, 10ms, 15ms delay
HARQ
Chase-combing with 4 retransmission with 4
frames of retransmission delay
RB size
48 sub-carriers x 6 symbols
Frame length
5 ms
Scheduler
PF
31