Ryoichi Shinkuma received B.E., M.E., and Ph.D. degrees in communications engineering from Osaka University in 2000, 2001, and 2003, respectively. In 2003, he joined the Communications and Computer Engineering Faculty of the Graduate School of Informatics, Kyoto University, where he is currently an associate professor. He was a visiting scholar in the Wireless Information Network Laboratory, Rutgers University from Fall 2008 to Fall 2009. His research interest is mainly cooperation in heterogeneous networks. He received the Young Researchers’ Award from IEICE in 2006 and the Young Scientist Award from Ericsson Japan in 2007. He also received the TELECOM System Technology Award from the Telecommunications Advancement Foundation in 2016 and the best tutorial paper award from the IEICE Communications Society in 2019. He was the Chairperson of the Mobile Network and Applications Technical Committee of the IEICE Communications Society from 2017 to 2019. He is a senior member of IEEE.
2. 2
Trend: diversification of connected `things'
2000
2010
2005
2015
2020
My research started
Drones (Micro UAVs)
Feature
phones
Servers PCs
IoT
Diversified
Smartphones
Wearable
devices
Connected cars
My research
started in 2003
Before 2000, only
servers and PCs are
connected
After 2005, many kinds
of smart devices are
connected
In coming years,
everything will be
connected
3. 3
Benefits of heterogeneous networks
Diversity of connection
Many nodes for
cooperation,
congested
network
High-speed
but short-
range
connection
Wide-range
but low-
speed
connection
Few nodes for
cooperation,
no congestion
Diversity of environment
Diversity of resources
High-speed
communicatio
n resource
Powerful
computing
resource
Diversity of data
Collect and
provide fine-
grained data in
dedicated area
Collect and
provide bird's-
eye view and
wide-range data
4. 4
One entity can control all nodes
How to optimize?
e.g. Maximum flow problem
Cooperation in homogenous network
s t
b
a
e
g
f
c
d
Maximize flow from s to t
5. 5
More than just variety of parameters:
Definition of `heterogeneous'
• Two or more kinds of nodes
• Each node owned and controlled by
different entities
Much more difficult to
optimize
than homogeneous networks
7. 7[(Invitied) R. Shinkuma and K. Yamori, IEICE Technical report, CQ2012-69, Nov. 2012 (in Japanese)
Utility brought
about by
cooperation, Ui
Cost caused by
cooperation, Ci
Ui + Ri > Ci ?
No motivation
Incentive
mechanism
S Ui
S Ri
(< S Ui – e)
Yes
No
How to solve the problem of cooperation in
heterogeneous networks
Conventional approach: simplify and introduce game theory [Han'12]
My approach: understand and create a new model [Shinkuma'12]
Incentive
reward
allocated to
contribution, Ri
8. 8
Sharing
• Communication resources (bandwidth) and energy resources to
forward other's data
• Computing resources and energy resources to process other's task
Delaying
requests from peak time to off-peak time
Moving
to location where more efficient resource is available
Actions for cooperation in heterogeneous
networks
9. 9
[H. Kubo, R. Shinkuma, and T. Takahashi, IEICE Trans. Inf. & Syst., vol. E93-D,
no. 12, pp. 3260-3268, Dec. 2010]
[M. Yoshino, R. Shinkuma, and T. Takahashi, Proc. IEEE GLOBECOM 2008, vol.
12, pp. 5042-5046, Dec. 2008]
Modeling of cost
Forwarding data for others causes battery
cost
Experimental evaluation
• Emulator developed by our lab
• 58 subjects
• Results of qR :
Social network
Relay network
• Friends: 41.2%
• Friends of friends: 55.4%
• Strangers: 60.8%
Modeling of cost caused by cooperative
forwarding
Psychological factor related to
social relationship?
10. 10[Y. Inagaki and R. Shinkuma, IEEE Access, vol. 6, pp. 23191-23201, April 2018]
Adamic-Adar Index
( kz: no. of degree of z )
Katz Index
( Al
xy: number of paths of length l between x
and y )
Prior work: only friends, friends of
friends, and strangers
Extract from structure of
social networks, validated
by experiment using
Stanford Brightkite dataset
My work: more differentiated
social relationship metrics for
cooperative forwarding
Differentiate social relationships (1)
11. 11
Prior work: only pre-determined social
relationships
[R. Shinkuma, Y. Sugimoto, and Y. Inagaki, Springer Soft Computing, Nov. 2017]
Convert time domain
to frequency domain
Extract from intercontact
logs like message
exchanges, Bluetooth
connections, validated by
experiment using MIT
Friends-and-Family dataset
per 1 day
per 10 min
My work: more dynamically changed
social relationships
Differentiate social relationship (2)
12. 12
Bandwidth Exchange (BE) in wireless relay
network:
Giving a portion of radio frequency
band when asking for forwarding
• Maximize product of utilities
• Ensure efficiency and fairness
Nash Bargaining Solution (NBS) based
incentive mechanism
Problem formulation
・ : utility of node i
・ : relay-request probability from i to j
・ : probability of choosing strategy x
・ : cooperation probability of i for j
Incentive mechanism for cooperative
forwarding
13. 13
Decision-making model
Numerical evaluation
• Rayleigh fading channel
• Orthogonal frequency division
multiplexing (OFDM) system
• Results:
NBS achieves best fair throughput
Geometricmeanof
throughput:
No. of nodes
Savedtransmissionpower
[dB]
• NBS: long-term profit
• Myopic: short-term profit
• Altruistic: always cooperate
Incentive mechanism for cooperative
forwarding (2)
NBS saves 4 times transmission
power
14. 14
Means of giving incentive reward for
cooperative forwarding
[D. Zhang, R. Shinkuma, and N. Mandayam, IEEE Trans. Wireless Communications, vol. 9, no. 6, pp. 2055-2065, Jun. 2010]
[T. Nishio, R. Shinkuma, T. Takahashi, and N. Mandayam, IEICE Trans. Commun., vol. E95-B, no. 6, pp. 1944-1952, Jun. 2012]
[W. Liu, R. Hu, R. Shinkuma, and T. Takahashi, IEICE Trans. Commun., vol. E98.B, no. 11, pp. 2141-2150, Nov. 2015]
[T. Nishio, R. Shinkuma, T. Takahashi, and G. Hasegawa, EURASIP Journal on Wireless Communications and Networking, vol. 2011, no. 1,
Aug. 2011]
[N. Maki, T. Nishio, R. Shinkuma, T. Mori, N. Kamiyama, R. Kawahara, and T. Takahashi, IEICE Trans. Inf. & Syst., vol. E95-D, no. 12, pp.
2860-2869, Dec. 2012]
Payment by quality of service (QoS)
Monetary payment
• Radio frequency band in physical layer
[Zhang, 2010]
• TXOP in data link layer [Nishio, 2012]
• Packet queuing in network layer [Liu, 2015]
• Flow rate in transport layer [Nishio, 2011]
• Content delivery in application layer [Maki,
2012]
Physical
Data link
Network
Transport
Session
Presentation
Application
15. 15
Modeling of utility brought about by
cooperative computing-resource sharing
Computing-resource sharing network
Modeling of utility
Service consists of multiple tasks;
task-oriented cooperation might
cause inefficient cooperation
Service-oriented
cooperation is proposed
c) Mixed case d) Parallelized sequences
a) Sequential case b) Parallel case
Sharing computing resources via
networks
• Utility
• Problem formulation
Is it different from forwarding?
16. 16
Modeling of utility brought about by
cooperative computing-resource sharing (2)
:
Problem formulation
・ : service start time with cooperation
・ : service start time w/o cooperation
・ : processing time for task l of node i
・ : energy consumption with cooperation
・ : energy consumption w/o cooperation
(Reduced service
latency)
• Proposed (service-oriented) cooperation
reduced latency up to upper-bound
level
• Proposed model achieved best fairness
Results
17. 17
Delaying requests from peak time to
off-peak time
Delaying as cooperative action
[R. Shinkuma, Y. Tanaka, Y. Yamada, E. Takahashi, and T. Onishi, Elsevier Computer Networks, vol. 137, pp. 17-26, Jun. 2018]
[Y. Yamada, R. Shinkuma, T. Iwai, T. Onishi, T. Nobukiyo, and K. Satoda, vol. 146, pp. 115-124, Dec. 2018]
Delaying should not cause another
peak traffic
(Current traffic) (Predicted traffic after
delaying)
18. 18
Delaying as cooperative action (2)
(Expected utility by choosing `Yes')
(Expected utility by choosing `No')
(Probability of choosing `Yes')
TA: sensitivity of
nodes to delay
Modeling of decision making
and behavior
• Random utility model: statistic
model for making binary
decisions (delay or not)
Results
• Reproduced traffic, estimated
from signal quality measured in
Kawasaki-city, Kanagawa, Japan
• Off-peak is reduced by
cooperation instead of forcing
nodes to delay
19. 19
Moving to location where more efficient
resource is available:
Moving as cooperative action
[T. Kangawa, M. Yoshino, R. Shinkuna, and T. Takahashi, IEICE Trans. Communications, vol. J90-B, no. 12, pp. 1263-1273, Dec. 2007 (in
Japanese)]
[M. Yoshino, K. Sato, R. Shinkuma, and T. Takahashi, IEICE Trans. Commun., vol. E91-B, no. 10, pp. 3132-3140, Oct. 2008]
[T. Kakehi, R. Shinkuma, T. Murase, G. Motoyoshi, K. Yamori, and T. Takahashi, IEICE Trans. Commun., vol. E95-B, no. 6, pp. 1965-1973,
Jun. 2012]
Move
Assign to new users
Assignedresources
20. 20
Conclusion
Recommended design of cooperation mechanism in heterogeneous networks
1. Cooperative actions:
Sharing resources, delaying, moving
3. Utility model:
Service-oriented
4. Incentive mechanism:
NBS-based
5. Decision-making:
stochastic models like random utility theory are ready-to-use but
not enough to reflect individual characteristics
further research is needed
Psychological factor related to social relationships
2. Cost model:
Editor's Notes
Z. Han, D. Niyato, W. Saad, T. Başar, and A. Hjørungnes, ``Game theory in wireless and communication networks: theory, models, and applications,'' Cambridge university press, 2012
Lambda 1, 2, 3, 4 corresponding to <n,c><c,c><c,n><n,n>