CoPrimarySharingLSP_PPT_BSingh_IEEEICC_2015
-
Upload
bikramjit-singh -
Category
Engineering
-
view
133 -
download
0
Transcript of CoPrimarySharingLSP_PPT_BSingh_IEEEICC_2015
A! Aalto UniversityComnet
Co-primary inter-operator spectrum sharingover a limited spectrum pool using repeatedgames
Bikramjit Singh1, Konstantinos Koufos1, Olav Tirkkonen1 andRandall Berry2
1Aalto University, Finland and 2Northwestern University, USA
IEEE ICC London, 9.6.2015
A! Motivation: Dense Networks⇒ New challenges for flexible, efficient spectrum use
• small cells added to increase capacity in hot spots
• multiple indoor operators in the same geographic area
⇒ Prominent load asymmetry• number of UEs per BS can vary a lot
• orthogonal spectrum sharing between the operators becomes inefficient
IEEE ICC 2015 2 (13) London, 9.6.2015
A! Co-primary Spectrum Sharing• More spectrum is needed for 5G cellular systems⇒ Co-primary sharing: Multiple operators share spectrum
• Closed set of players with similar legal rights to spectrum
• Only these players have access to spectrum considered.
⇒ Limited spectrum pool (LSP)
• Operators have equal legal rights to the shared spectrum
• Baseline: shared spectrum is used by all
⇒ Mutual renting
• Operators have exclusive right to own spectrum, may permit others to use
• Baseline: each operator uses its own spectrum
IEEE ICC 2015 3 (13) London, 9.6.2015
A! System Model I• Operator x constructs a number (utility) Ux to describe QoS of its
user population• proportionally fair, cell edge throughput, ...
• Operators do not have to maintain same utility nor be aware of otheroperator utility
• Operators are self-interested entities• They do not want to reveal their CSI, performance, or load
• Added functionality over state-of-the-art• Operator may ask its UE to mea-
sure aggregate signal level fromopponent network
• UEs able to separate between ownand other operator’s interference
• Simple extension of LTE handovermeasurements
Serving BS
Interfering BS
IEEE ICC 2015 4 (13) London, 9.6.2015
A! Coordination Protocol
• We assume direct connection between spectrum controllingelements in the radio access networks of the operators
• We assume no monetary transactions• Instead, to enable spectrum sharing, coordination protocol
• Operators may ask/grant opponents for spectrum usage favor• Protocol consists of bids and results (agreements, disagreements)• Results hold for a certain fixed time slot.
IEEE ICC 2015 5 (13) London, 9.6.2015
A! Coordination Protocol: Possible States
Network states
Symmetric load
Asymmetric load
IEEE ICC 2015 6 (13) London, 9.6.2015
A! Spectrum Sharing Game based on Favors
• Given a coordination protocol, spectrum sharing becomes anon-cooperative game between operators• The protocol determines the rules
• The players are rational and self-interested
• The game is Bayesian: opponent state is not known• All you know about the value of a favor to the opponent is:
a) sometimes he asks for it, sometimes notb) sometimes he grants it to you, sometimes not
• Spectrum sharing game for LSP with K carriers• Actions: ak—ask opponent to stop using k = 1, . . . K carriers,
gk—propose to stop using k = 1, . . . K carriers,n—do nothing
• Outcome: If one plays gl and other ak with l ≥ k ⇒ k favours exhanged.Otherwise nothing happens.
IEEE ICC 2015 7 (13) London, 9.6.2015
A! Repeated Game
• In one-shot game, rational player always asks, never grants
• Operators share spectrum for a long time• an almost infinite sequence of games
⇒ Repeated game:• Strategy depends on network state, network state statistics, and history of
interactions
• Book-keeping of spectrum usage favors given and taken
• Operators are free to decide whether they play the game or not
⇒ A rudimentary economy on RAN-level• A favor is the currency
IEEE ICC 2015 8 (13) London, 9.6.2015
A! Decision Making Process• Heuristic strategy for long-term reciprocity• Operator has estimated probability distribution of gain/loss when
asking/granting a carrier
0 1 2 3 4 5 6 7 80
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Utility
CD
F
Gain, FG2
Gain, FG1
Loss, FL2
Loss, FL1
• In every time slot, Operator x estimates utility gain/loss forasking/granting up to K favors
• Decision to ask/grant a favor on k carriers based onthreshold-based tests• Gain threshold, θX,j; Loss threshold, λX,j
IEEE ICC 2015 9 (13) London, 9.6.2015
A! Optimal Gain/Loss thresholds
• Decision thresholds set to maximize excess utility∼UX over the
one-shot game NE
• Constraint: long-term reciprocity
Maximize :θA,k,λA,k ∀k
∼UA,
Subject to :
K∑k=1
kP askA,kP
grantB,k =
K∑k=1
kP askB,kP
grantA,k ,
• History of interactions taken into account: own ask/grantthresholds depend on opponent ask/grant probabilities
IEEE ICC 2015 10 (13) London, 9.6.2015
A! Simulation Setting• Two operators each with 2 BSs each in a single-story building• Spectrum divided into 4 parts
• Two parts in limited spectrum pool• one private part each
• Users generated according to Poisson point process• Operator mean load 5 UEs• fully overlapping coverage areas
• Operators maximize proportionally fair utility
Operator A
Operator B
120 MHz
IEEE ICC 2015 11 (13) London, 9.6.2015
A! Simulation results• Learning phase
• operators construct the distributions and set decision thresholds
• Steady state distributions are constructed over many network states
• Operation phase• load pdfs assumed constant over both learning and operation phase
• Gain at 50-th percentile of UEs rate CDF is 25%
0 10 20 30 40 50 60 700
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Mbps
CD
F
One−shot game
Coordination protocol
IEEE ICC 2015 12 (13) London, 9.6.2015
A! Summary
• Repeated game framework for spectrum sharing fornon-cooperative operators
• Operators share spectrum pool consisting of multiple carriers• Spectrum sharing through spectrum usage favors
• Operators construct utility gain/loss distributions and obtaindecision thresholds
• Based on gain and loss thresholds, operators ask and grantspectrum usage favors
• Favors provide a rudimentary currency to trade losses and gainsin the time domain• Effective when operator load varies in time
IEEE ICC 2015 13 (13) London, 9.6.2015