1/12IEEE ICC 2014 (Sydney, Australia)
Quality-Aware Millimeter-Wave Device-to-Device
Multi-Hop Routing for 5G Cellular Networks
Joongheon (Joon) Kim and Andreas F. Molisch
Department of Electrical Engineering, University of Southern California
IEEE International Conference on Communications (ICC)
Sydney, Australia, June 2014
2/12IEEE ICC 2014 (Sydney, Australia)
Introduction
β’ Millimeter (Mm-Wave) transmission has been actively studied for 5G cellular systems
β’ Objective: Increasing capacity based on ultra-wide channel bandwidth
β’ Thus, next generation phones will be equipped with mm-wave RF.
β’ Question
If device-to-device (D2D) video streaming is performed over the mm-wave enabled phones,
What kinds of algorithms are required?
β’ Multi-hop routing mechanisms are required due to its propagation characteristics.
β’ Therefore,
β’ A Quality-Aware Millimeter-Wave Multi-Hop Routing Algorithm is investigated.
3/12IEEE ICC 2014 (Sydney, Australia)
Preliminaries
A Reference Network Model
Relay
ππ
ππ
Wall
Multi-Hop Routing is required in Mm-Wave D2D tocombat non-line-of-sight (NLOS) situations
Multi-Hop Routing is required in
Mm-Wave D2D toenable long-distance transmission
Why Multi-Hop Routing is required for Mm-Wave D2D Communications?
ππ ππ
ππ
ππ
ππ
4/12IEEE ICC 2014 (Sydney, Australia)
Quality-Aware Mm-Wave D2D Multi-Hop Routing
Mathematical Modeling
Objective Function
π πβππ
ππ ππ πβπ£π πMaximize:
ππ π π
Flow from ππ to π originated by ππ
π
β¦β¦πππβπππ ππβππ
πππππβππ
ππ
ππ
πππβπ π
ππ
The Quality Function of the
Flow Originated by π π
Maximize the sum of the qualities of all give flows
Summation of the Qualities
of All Flows
(ππ is a set of sources)
Source ππ Session π Destination π π
Two Types of Quality Functions
Flow amount Flow amount
Quality Quality
ππ β ππ β
π€π π€π
ππππ₯π π ππππ₯
π π
Linear
FormNonlinear
(Concave)
Form
5/12IEEE ICC 2014 (Sydney, Australia)
Quality-Aware Mm-Wave D2D Multi-Hop Routing
Constraint #1: Device Constraints
ππ β¦β¦ ππ
πΏπ£πβπ£π = 1,0,
if π£π sends data to π£πotherwise
ππ β¦β¦ π π
π πβ π£
πΏπ πβπ£ = 1, βπ π
Each source π π should
send data to the one of
the other nodes:
π£β ππ
πΏπ£βππ= 1, βππ
Each destination ππ should
receive data from the one of
the other nodes:
ππ
In intermediate nodes,
If it receives data, it should transmit the data,
and visa versa, i.e.,
π£πβ π£π
πΏπ£πβπ£π = π£πβ π£π
πΏπ£πβπ£π
6/12IEEE ICC 2014 (Sydney, Australia)
Quality-Aware Mm-Wave D2D Multi-Hop Routing
Constraint #2: Relay Constraints
β¦ β¦
π£β ππ
πΏπ£βππ β€ ππ πΉπ π₯
The number of incoming flows is limited by
the number of receiver RF ππ πΉπ π₯ chains:
Relay
ππ
ππβ π£
πΏππβπ£ β€ ππ πΉππ₯
The number of outgoing flows is limited by
the number of transmitter RF ππ πΉππ₯ chains:
Each antenna has a connection to one device only.
7/12IEEE ICC 2014 (Sydney, Australia)
Quality-Aware Mm-Wave D2D Multi-Hop Routing
Constraint #3: Flow Constraints
β¦ β¦
Relay
ππThe amount of
Incoming Traffic
The amount of
Outgoing TrafficThe amount of
Incoming Traffic
The amount of
Outgoing Traffic
The amounts of incoming traffic and outgoing traffic should be same:
ππ
In each device π£π,
In each relay ππ,
π£πβ π£π
ππ£πβπ£π
π π = π£πβ π£π
ππ£πβπ£π
π π , βπ π
π£πβ ππ
ππ£πβππ
π π = ππβ π£π
πππβπ£π
π π , βπ π
8/12IEEE ICC 2014 (Sydney, Australia)
Quality-Aware Mm-Wave D2D Multi-Hop Routing
Constraint #3: Flow Constraints (Continued), Capacity Calculation
πππβππ
ππππ ππ
Limited by Link Capacity: πΆ π£π,π£π= π΅ β log2 1 + πππ
ππ πππππ,ππ΅ β πππππ π,ππ΅
ππ πππππ,ππ΅ = πΈπΌπ π + πΊπ π₯ + πΏ π
β’ πΈπΌπ π: 47 dBm in 38GHz
β’ πΊπ π₯: Rx antenna gain (25 dBm in relays, 13.3 dBm in phones)
β’ πΏ π : path loss model which is formulated as
where π0 = 5m (unit distance), Ξ» is wavelength, π is path-loss
coefficient, ππ is a shadowing (Gaussian) random variables.
πΏ π = 20 log104ππ0
Ξ»+10π log10
π
π0+ππ
πππππ π,ππ΅ = 10 log10 ππ΅ππ β π΅ + πΉπ
β’ ππ΅ππ: noise power spectral density (-174dBm/Hz)
β’ πΉπ: Rx noise figure (set to 6 dB)
9/12IEEE ICC 2014 (Sydney, Australia)
Quality-Aware Mm-Wave D2D Multi-Hop Routing
Mathematical Optimization Formulation
Quality-Aware Mm-Wave D2D Multi-Hop Routing
π πβππ
ππ ππ πβπ£π πMaximize:
Subject to
π πβ π£
πΏπ πβπ£ = 1, βπ π π£β ππ
πΏπ£βππ= 1, βππ
π£πβ π£π
πΏπ£πβπ£π = π£πβ π£π
πΏπ£πβπ£π
π£β ππ
πΏπ£βππ β€ ππ πΉπ π₯
ππβ π£πΏππβπ£ β€ ππ πΉ
ππ₯
π£πβ π£π
ππ£πβπ£π
π π = π£πβ π£π
ππ£πβπ£π
π π , βπ π
π£πβ ππ
ππ£πβππ
π π = ππβ π£π
πππβπ£π
π π , βπ π
πππβππ
ππ β€ πͺ ππ,ππ
Max-Min Multi-Hop Flow Routing
πMaximize:
where πΈ β€ πππβππ
ππ
Even though max-min multi-hop flow routing is widely
used for quality-aware applications, it cannot
consider the differentiated quality functions of the
given individual flows.
This formulation is mixed integer disciplined convex
programming where the given integers are 0-1 binary
(i.e., πΏπ£πβπ£π = 0,1 ), i.e., branch-and-bound is
widely used in literatures to obtain optimal solutions.
10/12IEEE ICC 2014 (Sydney, Australia)
Performance Evaluation
Parameters, Settings, and Results
β’ Parameters
β’ Carrier frequency: 38 GHz
β’ In 25 dBi Rx antenna (for relays),β’ π is 2.20 in LOS and 3.88 in NLOS
β’ π is 10.3 in LOS and 14.6 in NLOS
β’ In 13.3 dBi Rx antenna (for phones),β’ π is 2.21 in LOS and 3.18 in NLOS
β’ π is 9.40 in LOS and 11.0 in NLOS
β’ Settings
β’ 20 number of phones; 5 number of relays
β’ Each relay has 4 Tx RF and 4 Rx RF
β’ 4 sessions with various quality functions
Parameters and Settings Performance Evaluation
β’ The proposed algorithm (differentiated quality consideration (DQC)) is
compared with max-min scheme routing (MmF).
β’ Average throughput of DQC & MmF, i.e.,πΈ ππ·ππΆ|ππ & πΈ ππππΉ|ππdepending on link failure probability ππ
DQC presents 33% better average
throughput compared to max-min
flow routing.
11/12IEEE ICC 2014 (Sydney, Australia)
Conclusions and Future Work
β’ We propose a millimeter-wave multi-hop routing protocol for 5G cellular systems:
β’ Assisted by multi-antenna relays
β’ Quality-Awareness is introduced
β’ Differentiated quality metrics for individual flows are taken account (better performance than max-min routing)
β’ 33% performance improvement compared to max-min flow routing
β’ Future research direction
β’ Conducting further research for the other 5G frequency, i.e., 28 GHz, as well.
12/12IEEE ICC 2014 (Sydney, Australia)
Q&A
β’ For more questions,
please email to