1. VehiCaching: Embracing User Request on Vehicle Route with Proactive Data Transportation
Wonkwang Shin, Byoung-Yoon Min and Dong Ku Kim
School of Electrical and Electronic Engineering
Yonsei University, Seoul, Korea
2015 IEEE 81st Vehicular Technology Conference (VTC Spring)

2. Outline Introduction Network Model
VehiCaching: Proactive Multiple-Access
VehiCaching: File Placement Strategies
Simulation Results and Remarks
Conclusion

3. Introduction
“Cache” the contents on overlaid layers of the networks, and off-load specific contents by immediate users' access without realtime backhaul
We propose movable caching nodes, namely VehiCaching (VC), assisted cellular networks that can potentially achieve proactive data off-loading

4. Introduction
It replaces wired/wireless backhaul transmission by equipped storage utilization and moving pathway
How users proactively determine to access between macrocell and VC layer
The enhanced file placement strategy to contain pre-known user demand on vehicles' route

5. Network Model

6. Network Model random-walk mobility model [10]
have some unrealistic moments to reflect the properties of vehicular model in the terms of dramatically changing mobility direction
limited random-walk model
a mobility direction is uniformly selected within restricted maximal and minimal direction, for example −60° to +60°
the direction is renewed whenever vehicles move per unit time t and all vehicles go forward by their selected direction with the same speed
[10] MM. Zonoozi and P. Dassanayake “User Mobility Modeling and Characterization of Mobility Patterns” vol. 15, no. 7, pp , Sep., 1997

7. Network Model
Propagation loss
Received signal power

8. Network Model Propagation loss Received signal power
the path-loss in dB scale at time t

9. Network Model Propagation loss Received signal power
the value of path loss at unit distance

10. Network Model Propagation loss Received signal power
the path loss exponent

11. Network Model Propagation loss Received signal power
the distance between the transmitter and the receiver

12. Network Model Propagation loss Received signal power
shadowing term with standard deviation σ

13. VehiCaching: Proactive Multiple-Access
It is not difficult to predict mobility of public transportation because it moves along the predetermined route
VCs serve users located near its trajectory by files in its storage which is filled by system in advance

14. VehiCaching: Proactive Multiple-Access
Since VCs serves users planed to be served by macro BS, data is off-loading from the macro-layer
Therefore, in case of vehicles, it is desirable to serve lower power than macro BS to prevent give harm to the user who is served by macro BSs

15. VehiCaching: Proactive Multiple-Access
To control proper load balance between the macro and pico layers, a positive bias is applied to the received SINR measured from pico BSs
This approach can expand picocell range and referred to as the range expansion (RE)

16. VehiCaching: Proactive Multiple-Access

17. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t

18. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t
the subset of VCs
the set of APs

19. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t
1 if the file f exists in storage of AP ai and 0 otherwise

20. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t
1 if i th user requests the file f and 0 otherwise

21. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t
received signal power from AP ai

22. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t
variance of white noise with normal distribution

23. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t
1 if ai is contained V is satisfied, 0 if not

24. VehiCaching: Proactive Multiple-Access
index of AP that associated with the user i at time t
RE bias for load balancing

25. VehiCaching: Proactive Multiple-Access
the user throughput is their instantaneous rate multiplied by the fraction of resources
therefore, i th user is allowed to use about 1/Nai resources by round robin scheduling
throughput of i th user at time t

26. VehiCaching: Proactive Multiple-Access
the user throughput is their instantaneous rate multiplied by the fraction of resources
therefore, i th user is allowed to use about 1/Nai resources by round robin scheduling
throughput of i th user at time t
the number of users located within the coverage of AP ai*

27. VehiCaching: File Placement Strategies
it is desirable that many users are connected to the vehicle in aspect to data off-loading
vehicle has to have user request files in limited storage as much as possible

28. VehiCaching: File Placement Strategies
File Placement by Statistical Priority
assume that video downloading files we focused on this system follow Zipf distribution
VCs store the data in accordance with this Zipf distribution within the limited storage when there is no prior information about the user
difficult to obtain a performance gain by mobility

29. VehiCaching: File Placement Strategies
File Placement by User Demand Awareness
if users' file requests are known a priori, a system performance of VC increases because it can sort files to store in limited storage

30. VehiCaching: File Placement Strategies
Storage range (Rs)
the range which decides the request of how far users from VCs trajectory should be reflected to file placement in VC
If users within Rs, VC stores their request file and otherwise, VC not stores their request file and vacancy of storage is filled by statistical priority

31. VehiCaching: File Placement Strategies
Storage range (Rs)
the range which decides the request of how far users from VCs trajectory should be reflected to file placement in VC
If users within Rs, VC stores their request file and otherwise, VC not stores their request file and vacancy of storage is filled by statistical priority
user set in storage range Rs

32. VehiCaching: File Placement Strategies

33. VehiCaching: File Placement Strategies
if users located in VC transmission range are not connected at VC, they receive a lot of interference from VC
outage probability for target throughput Ttarget

34. Simulation Results and Remarks
Using
Only macro BS
File placement by statistical priority
Proposed file placement by user demand awareness
Simulated average throughput during 50 seconds and observed snapshot results per second

35. Simulation Results and Remarks

36. Simulation Results and Remarks

37. Simulation Results and Remarks

38. Simulation Results and Remarks

39. Simulation Results and Remarks

40. Simulation Results and Remarks

41. Simulation Results and Remarks
Complementary
Cumulative
Distribution
Function
(0.3 Mbps target throughput)

42. Simulation Results and Remarks
VCs can service the data to users located in macrocell edge
Complementary
Cumulative
Distribution
Function
(0.3 Mbps target throughput)

43. Simulation Results and Remarks
users around VC undergo more interference
Complementary
Cumulative
Distribution
Function
(0.3 Mbps target throughput)

44. Simulation Results and Remarks
Complementary
Cumulative
Distribution
Function
(0.3 Mbps target throughput)

45. Conclusion We focus on two aspects of VC
proactive multiple-access
file placement strategy
We then propose a method to effective file placement containing user demand compared to that of statistical priority, which newly define the storage range as the range of consideration of users request

46. Conclusion
We evaluate the performance gains on throughput and outage probability compared with conventional priority- based caching and non-caching cellular networks
The results show that proposed VC including its file placement scheme is necessary to movable caching system