Smart Traffic: A Case Study for Developing IoT Application ...

Smart Traffic: A Case Study for Developing IoT Applications in

Vietnam

Faculty of Computer Sc ience and Engineer ing

Ho Chi Minh City Univers i ty of Technology

VNU-HCM

Tran Vu Pham and Duc Hai Nguyen

Outline• Introduction: Smart Traffic at HCMUT

• Opportunities and Challenges

• Solutions

• Future work

• Conclusion

HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY 2

Introduction• Transportation development does not

keep up with economic and population growth.• HCMC in 2015:

• > 8 million people

• > 7 million motorbikes

• Traffic congestion becomes a major concern.• Traffic in HCMC:

• No. vehicles increases ~10%/year

• Street capacity increases ~2%/year


Existing Solutions


Building temporarilysteel bridges

Installing surveillancecameras

Broadcasting traffic conditionvia Radio

Smart Traffic at HCMUT


http://traffic.hcmut.edu.vn/

http://traffic.hcmut.edu.vn/

Opportunities• Management point of view

• Receive a lot of support and motivation from different management levels of the city.

• Have access to sensing data.

• Calculation results are tested and used in realistic conditions.

• Technical point of view• Applications of IoT and ITS technologies are still new in Vietnam

• Transportation in Vietnam has some unique characteristics

• A lot of rooms for new applications and research ideas


Challenges• High volume of data

• Sensing data are collected continuously from multiple sources.

• High sample rate is required to deliver high accuracy calculations

• Real-time processing• Required for instant actions

• Strict deadline

• Unreliable network infrastructure• Long delays

• Packet losses


Smart Traffic at HCMUTAn overview


Raw dataset

Smart Traffic at HCMUT

Our concerns

Real-time Processing• Challenges

• Huge data, short deadline

• Continuous processing

• Solutions for Smart Traffic• Parallelization is the key

• Develop a high performance storage system for quickly putting and getting records


Characteristics of Sensing Data


Spatial distribution Temporal distribution

S4STRD Architecture

07 Jan 2016 HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY 11

API

Application ApplicationApplication

Client side

RAM store

Disk-based store

Buffer

Disk Disk Disk

Server side

S4STRD Performance


Results from experiments on a cluster of three machines connected by Ethernet 1 Gbps

0

20

40

60

80

100

120

140

MongoDB S4STRD(with synchronization)

S4STRD(without

synchronization)

thro

ugh

pu

t (t

ho

usa

nd

rec

ord

/s)

Write performance

0

1

2

3

4

5

6

7

MongoDB S4STRD

Late

ncy

(m

icro

seco

nd

)

Read performance

Other Challenges• Low data utilization

• Usable data account for < 17%

• How to evaluate computation results• Multiple data sources with different

confident factors

• Need for a standard measurement that we can trust


17%

7%

38%

38%

Usable Miss deadline

Invalid Noisy records

Future Work• Improve data utilization

• Correction algorithms

• Filter, fusion modules

• Develop a method to evaluate our estimation• Multiple data source with different confident factors

• Decision making problems

• Need for a standard measurement that we can trust

• Expand the system:• Real-time routing finding

• Integrate into social networks


Conclusions• Smart Traffic

• Involves moving sensing devices, different types of networking infrastructure

• Big and real-time data analytics

• Aim at a very large community

• Our contributions• Give insight into development of IoT applications in Vietnam

• Solutions can be reused in other similar applications


The End

Thanks for listening.


Smart Traffic: A Case Study for Developing IoT Application ...

Documents

Transcript of Smart Traffic: A Case Study for Developing IoT Application ...