Database Laboratory 2013-10-07 TaeHoon Kim

Database Laboratory2013-10-07

TaeHoon Kim

Work progress

/18

Work Progress

/18

Work Progress

1111 **** ****1110 **** ****1100 **** ****

Database LaboratoryRegular Seminar

2013-10-07TaeHoon Kim

Spatial Big-Data Challenges Intersecting MobilityAnd Cloud Computing

Shashi Shekhar, Michael R. Evans, Viswanath Gunturi,KwangSoo Yang

Computer Science & Eng. Faculty, University of Minnesota

MobiDE '12 Proceedings of the Eleventh ACM International Work-shop on Data Engineering for Wireless and Mobile Access

/18

Contents

1. Introduction2. Traditional Mobility Services3. Emerging Spatial Big Data4. New Challenges5. Conclusions

/18

Introduction Mobility is efficient, safe and affordable travel

In our cities, towns and other places of interest

Mobility services Routing and Navigation

From Google Maps to consumer GPS devices, society has benefited immensely from mobility services and technology

Scientists use GPS to track endangered species to better un-derstand behavior

Farmers use GPS for precision agriculture to increase crop yields while reducing cost

Hiker, biker, taxi driver know precisely where they are, their nearby points of interest, and how to reach their destinations.

6

/18

Introduction However, the size, variety, and update rate of mobility data

sets exceed the capacity To learn, manage, and process the data with reasonable effort

Such data is known as Spatial Big Data We believe that harnessing SBD represents the next gener-

ation of mobility services Examples of emerging SBD dataset include temporally de-

tailed(TD) roadmap Provide speeds every minute for every road-segment, GPS trace

data from cell-phones, engine measurements of fuel consumption, greenhouse gas(GHG) emissions

7

/18

Introduction A 2011 McKinsey Global Institute report estimates savings

of “about $500 billion annually by 2020” in terms of fuel and time saved by helping vehicles avoid congestion and reduce idling at red lights of left turns

8

/18

Introduction However, SBD raise new challenges

1. It requires a change in frame of reference, moving from a global snapshot perspective to the perspective the individual object traveling through road network

2. SBD increase the impact of the partial nature of traditional route query specification

3. The growing diversity of SBD sources makes it less likely that single algorithms, will be sufficient to discover answer ap-propriate for all situation

Other challenges Geo-sensing, privacy, prediction, etc

9

/18

Traditional Mobility Services Traditional mobility services utilize digital road map

Graph-based

Digital road map Road intersections are often modeled as vertices Road segments connecting adjacent intersections are repre-

sented as edges in the graph

10

/18

Traditional Mobility Services Route determination services, abbreviated as routing ser-

vices Best-route determination Route comparison

The first deals with determination of a best route given a start location, end location, optional waypoints and prefer-

ence function(fastest, shortest, easiest, pedestrian, public transportation …)

Route finding is often based on classic shortest path such as Dijktra’s, A*, hierarchical, materialization, other algo-rithms for static graphs

11

/18

Emerging Spatial Big Data Spatio-Temporal Engine Measurement Data

Datasets may include a time-series of attributes such as vehi-cles(weight, engine size), engine speed

Fuel efficiency can be estimated from fuel levels and distance traveled as well as engine idling from engine RPM

Fig3. Heavy truck fuel consumption as a function of elevation from a recent study at Oak Ridge National Laboratory

Explore the potential of this data to help consumers gain similar fuel savings and GHG emission reduction

12Figure3

/18

Emerging Spatial Big Data Spatio-Temporal Engine Measurement Data

Problem : These dataset can grow big Measurements of 10engine variables, once minute, over 100 mil-

lion US vehicles in existence, may have 1014 data-items per year GPS Trace Data

GPS trajectories are becoming available for a large collection of vehicles due to the rapid proliferation of cellphones, in-vehicle navigation devices

13

Make it possible to make per-sonalized route suggestions to users to reduce fuel consump-tion and GHG emission

GPS record taken at 1minute interval, 24 hour day, 7days a week

/18

Emerging Spatial Big Data Historical Speed Profiles

The profiles have data for every minutes, which can then be applied to the road segment, building up an accurate picture of speeds based on historical data

14

/18

New Challenges 1st : It requires a change in frame of reference, moving from

a global snapshot perspective to the perspective the indi-vidual object traveling through road network

15

Time

D1 : 20 D2 :

30

D1 : 20 D2 :

30

D1 : 10 D2 :

20

D1 : 20 D2 :

30

D1 : 20 D2 :

10

Time

/18

New Challenges 2nd : SBD increases computational cost because it magni-

fies the impact of the partial nature of the traditional route query specification

For example, traditional routing identifies a unique route(or small set)

but, SBD may identify a much larger set of solution What is he computational structure of determining routes that

minimize fuel consumption and GHG emission? : Eco-routing

3rd : The tremendous diversity of SBD sources substantially increases the need for diverse solution methods

For example, TD roadmaps cover an entire country, but provide mean travel-time for a road-segment for a given start-time in a week

16

/18

New Challenges 4th : Use of geospatial reasoning and SBD in sensing and in-

ference across space and time

5th : Privacy of geographic information inside SBDs is an im-portant challenge

While location information can provide great value to users and industry, streams of such data also introduce spooky privacy concerns of stalking and geo-slavery

6th : SBD can also be used to make predications the future path of a hurricane

17

/18

Conclusion This paper addresses the emerging challenges posed by

such datasets, which we call Spatial Big Data, specifically as they apply to mobility services (e.g transportation and rout-ing)

Challenges 1th : SBD requires a change in frame of reference, moving from

a global snapshot perspective to the perspective the individual object traveling through road network

2th : SBD increases computational cost because it magnifies the impact of the partial nature of the traditional route query specification

3th : Assumption that a single algorithm utilizing a specific dataset is appropriate for all solution 18