Mobile Data Mashups per applicazioni di Urban Computing Emanuele Della Valle Irene Celino . [email_address] [email_address] . http://emanueledellavalle.org http://swa.cefriel.it . Joint work with: Daniele Dell’Aglio, Kono Kim, Zhisheng Huang, Volker Tresp, Werner Hauptmann, and Yi Huang

Agenda Introduction Cities are alive Mobile users’ questions Urban Computing Data Mashups Are Data Mashups up to address Mobile users’ needs? Powerful visualization Simple programming abstractions Does everything boil down to plumbing? Requirements for a Mobile Data Mashup Environment LarKC as a backbone for a Mobile Data Mashup Environment What’s LarKC? Asking LarKC Plugging components in LarKC Configuring a LarKC pipeline Demo of current state of development of Urban Baby LarKC What’s next? Conclusions and outlooks GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

Introduction

Cities are alive

Mobile users’ questions

Urban Computing

Data Mashups

Are Data Mashups up to address Mobile users’ needs?

Powerful visualization

Simple programming abstractions

Does everything boil down to plumbing?

Requirements for a Mobile Data Mashup Environment

LarKC as a backbone for a Mobile Data Mashup Environment

What’s LarKC?

Asking LarKC

Plugging components in LarKC

Configuring a LarKC pipeline

Demo of current state of development of Urban Baby LarKC

What’s next?

Conclusions and outlooks

Cities Are Alive Cities born, grow, evolve like living beings. The state of a city changes continuously, influenced by a lot of factors, human ones: people moving in the city or extending it natural ones: precipitations or climate changes GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino [source http://www.citysense.com ]

Cities born, grow, evolve like living beings.

The state of a city changes continuously, influenced by a lot of factors,

human ones: people moving in the city or extending it

natural ones: precipitations or climate changes

Some Mobile Users’ Question “ Is public transportation where I am?” “ Is the event where I am the one that attract more people right now?” “ Where are all my friends meeting?” “ Is the traffic moving where I’m going?” GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

“ Is public transportation where I am?”

“ Is the event where I am the one that attract more people right now?”

“ Where are all my friends meeting?”

“ Is the traffic moving where I’m going?”

Urban Computing as an Answer to Them GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino [source IEEE Pervasive Computing,July-September 2007 (Vol. 6, No. 3) ]

Urban Computing The integration of computing, sensing, and actuation technologies into everyday urban settings and lifestyles. [source IEEE Pervasive Computing,July-September 2007 (Vol. 6, No. 3) ] GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

The integration of computing, sensing, and actuation technologies into everyday urban settings and lifestyles.

Example of Urban Computing Application GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino 5th cycle ASP Multidisciplinary Projects, Torino 24.1.2009

Data Availability Some years ago, due to the lack of data, Urban Computing looked like a Sci-Fi idea. Nowadays , a large amount of the required information are available on the Internet at almost no cost, e.g., Maps (Google,Yahoo!, Wikimapia, OpenStreetMap ), events scheduled (Eventful, Upcoming…), voluntarily provided users location (Google Latitude), mass presence and movements ( multimedia data with information about location (Flickr…) relevant places (schools, bus stops, airports...) traffic information (accidents, problems of public transportation...) city life (job ads, pollution, health care...) GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

Some years ago, due to the lack of data, Urban Computing looked like a Sci-Fi idea.

Nowadays , a large amount of the required information are available on the Internet at almost no cost, e.g.,

Maps (Google,Yahoo!, Wikimapia, OpenStreetMap ),

events scheduled (Eventful, Upcoming…),

voluntarily provided users location (Google Latitude),

mass presence and movements (

multimedia data with information about location (Flickr…)

relevant places (schools, bus stops, airports...)

traffic information (accidents, problems of public transportation...)

city life (job ads, pollution, health care...)

Are Data Mashups the Solution? GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino [source: http://www-01.ibm.com/software/lotus/products/mashups/ ] IBM Lotus Mashups [source: http://editor.googlemashups.com ] [source: http://pipes.yahoo.com/pipes/ ] [source: http://www.popfly.com/ ] [source: http://openkapow.com/ ]

Data Mashups Offers Powerful Visualizations GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino Google Charts API http://code.google.com/apis/chart/ http://maps.google.it/ http://maps.yahoo.com/ MIT Simile Timeline & Timeplot http://simile.mit.edu/timeline/ http://simile.mit.edu/timeplot/

Data Mashups Offers Simple Programming Abstractions GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

Not Everything Boils Down to Plumbing GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

Can Citysense Be Implemented with Pipes? Citysense was built to show you where the action is, right now. Using a billion points of GPS and WiFi positioning data from the last few years – plus real-time feeds – Citysense sees S.F. from above and puts the top live hotspots in your hand. You don't even need to sign up, just go to citysense.com on your BlackBerry, download, and open. GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino Live overall activity & top hotspots First of all see if it's a good night to go out. The city is 21% busier than normal for right now? Let's go. But where to? Check out the top hotspots in real-time and head out. What's at hotspot #1? Click over to Yelp or Google and find out what's going on at the #1 hotspot: Bars? Clubs? Restaurants? Then check out what's at #2 Show me where the unusually high activity is Even if you're a local, Citysense can give you the live details you need. When the Mission or Soma is busier than normal – you'll know immediately. Find out where everyone is going After dinner, drinks or a great night at a club, do you ever wonder where the afterparty is? Just press the "Locate Me" icon and see the top 5 places people go to from where you are now. [Source: http://www.citysense.com/moreinfo.php ]

Live overall activity & top hotspots First of all see if it's a good night to go out. The city is 21% busier than normal for right now? Let's go. But where to? Check out the top hotspots in real-time and head out.

What's at hotspot #1? Click over to Yelp or Google and find out what's going on at the #1 hotspot: Bars? Clubs? Restaurants? Then check out what's at #2

Show me where the unusually high activity is Even if you're a local, Citysense can give you the live details you need. When the Mission or Soma is busier than normal – you'll know immediately.

Find out where everyone is going After dinner, drinks or a great night at a club, do you ever wonder where the afterparty is? Just press the "Locate Me" icon and see the top 5 places people go to from where you are now.

Coping with reasoning heterogeneity precise and consistent inference for telling that at a given junction all vehicles, but public transportation ones, must go straight approximate reasoning when calculating the probability of a traffic jam given the current traffic conditions and the past history. GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino It means the systems allow for multiple reasoning paradigms ; e.g. [ source http://senseable.mit.edu/ ]

precise and consistent inference for telling that at a given junction all vehicles, but public transportation ones, must go straight

approximate reasoning when calculating the probability of a traffic jam given the current traffic conditions and the past history.

Coping with defaults heterogeneity 1/2 Open World Assumption vs. Close World Assumption While for the an entire city we cannot assume complete knowledge, for a time table of a bus station we can GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino [source: http://gizmodo.com/photogallery/trafficsky/1003143552 ]

Open World Assumption vs. Close World Assumption

While for the an entire city we cannot assume complete knowledge, for a time table of a bus station we can

Coping with defaults heterogeneity 2/2 Unique Name Assumption A square with several station for buses and subway can be considered a unique point for multimodal travel planning, but not when the problem is giving direction in that square to a pedestrian GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino 1 2 29 30 L3 L3

Unique Name Assumption

A square with several station for buses and subway can be considered a unique point for multimodal travel planning, but not when the problem is giving direction in that square to a pedestrian

Coping with scale Although we encounter large scale data which are not manageable , it does not necessary mean that we have to deal with all of the data simultaneously. Usually, only very limited amount data are relevant for a single query/processing at a specific application. GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino [source: http://gizmodo.com/photogallery/trafficsky/1003143552 ]

Although we encounter large scale data which are not manageable , it does not necessary mean that we have to deal with all of the data simultaneously.

Usually, only very limited amount data are relevant for a single query/processing at a specific application.

Coping with changing data Periodically changing data change according to a temporal law that can be Pure periodic law , e.g. every night at 10pm Milano overpasses close. Probabilistic law , e.g. traffic jam appear in the west side of Milano due to bad weather or when San Siro stadium hosts a soccer match. Event driven changing data are updated as a consequence of some external event. They can be further characterized by the mean time between changes : Slow , e.g. roads closed for scheduled works Medium , e.g. roads closed for accidents or congestion due to traffic Fast , e.g. the intensity of traffic for each street in a city GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

Periodically changing data change according to a temporal law that can be

Pure periodic law , e.g. every night at 10pm Milano overpasses close.

Probabilistic law , e.g. traffic jam appear in the west side of Milano due to bad weather or when San Siro stadium hosts a soccer match.

Event driven changing data are updated as a consequence of some external event. They can be further characterized by the mean time between changes :

Slow , e.g. roads closed for scheduled works

Medium , e.g. roads closed for accidents or congestion due to traffic

Fast , e.g. the intensity of traffic for each street in a city

Coping with Noisy, uncertain and inconsistent data Traffic data are a very good example of such data. Different sensors observing the same road area give apparently inconsistent information . a traffic camera may say that the road is empty whereas an inductive loop traffic detector may tell 100 vehicles went over it The two information may be coherent if one consider that a traffic camera transmits an image per second with a delay of 15-30 seconds, whereas a traffic detector tells the number of vehicles that went over it in 5 minutes and the information may arrive 5-10 minutes later. Moreover, a single data coming from a sensor in a given moment may have no certain meaning . an inductive loop traffic detector, it tells you 0 car went over Is the road empty ? Is the traffic completely stuck ? Did somebody park the car above the sensor ? Is the sensor broken ? Combining multiple information from multiple sensors in a given time window can be the only reasonable way to reduce the uncertainty. GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

Traffic data are a very good example of such data.

Different sensors observing the same road area give apparently inconsistent information .

a traffic camera may say that the road is empty

whereas an inductive loop traffic detector may tell 100 vehicles went over it

The two information may be coherent if one consider that a traffic camera transmits an image per second with a delay of 15-30 seconds, whereas a traffic detector tells the number of vehicles that went over it in 5 minutes and the information may arrive 5-10 minutes later.

Moreover, a single data coming from a sensor in a given moment may have no certain meaning .

an inductive loop traffic detector, it tells you 0 car went over

Is the road empty ?

Is the traffic completely stuck ?

Did somebody park the car above the sensor ?

Is the sensor broken ?

Combining multiple information from multiple sensors in a given time window can be the only reasonable way to reduce the uncertainty.

The LarKC Project GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino The Large Knowledge Collider a platform for infinitely scalable reasoning on the data-web Pipeline

LarKC at work for Urban Computing 1/2 GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino The Large Knowledge Collider project aims to develop a platform for massive distributed incomplete reasoning Traffic Monuments We are combining route planning techniques with reasoning on symbolic knowledge and traffic prediction produced by recurrent neural networks and continuous estimation of residual road capacity by real time analysis of data streams Inductive Loop http://www.larkc.eu PROBLEM : Which Milano monuments can I quickly visit from here?

LarKC at work for Urban Computing 2/2 We are combining route planning techniques with reasoning on symbolic knowledge, traffic prediction produced by recurrent neural networks, and continuous estimation of residual road capacity by real time analysis of data streams GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino You are here

We are combining route planning techniques with

reasoning on symbolic knowledge,

traffic prediction produced by recurrent neural networks, and

continuous estimation of residual road capacity by real time analysis of data streams

Conclusions and Outlooks LarKC aims at becoming an experimentation infrastructures for the development of advance semantic technologies. The public launch of the first open source release of the platform will take place in June 2009 We are developing our Urban Computing application as a showcase of the potentiality of the LarKC platform GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino The Large Knowledge Collider a platform for massive distributed incomplete reasoning http://www.larkc.eu

LarKC aims at becoming an experimentation infrastructures for the development of advance semantic technologies.

The public launch of the first open source release of the platform will take place in June 2009

We are developing our Urban Computing application as a showcase of the potentiality of the LarKC platform

Thank you for paying attention GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino Any Questions?

Mobile Data Mashups for Urban Computing Applications Emanuele Della Valle Irene Celino . [email_address] [email_address] . http://emanueledellavalle.org http://swa.cefriel.it . Joint work with: Irene Celino, Daniele Dell’Aglio, Kono Kim, Zhisheng Huang, Volker Tresp, Werner Hauptmann, and Yi Huang

Identifier strategy for Pipeline 2B Strategy based on common sense behavior: Detailed graph around starting and destination point (circles with center in the points and radius of 250 m) Main roads of the city Implemented in MixedStrategyIdentifier GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

Strategy based on common sense behavior:

Detailed graph around starting and destination point (circles with center in the points and radius of 250 m)

Main roads of the city

Implemented in MixedStrategyIdentifier

Towards Urban Baby LarKC Pipeline 3 GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino Urban City Decider SPARQL Result SPARQL Query Local Plug-in Manager SPARQL to GeoQuery Transformer Plug-in API Local Plug-in Manager SPARQL to GeoQuery Transformer Plug-in API Local Plug-in Manager Geo Location Identifier Plug-in API Local Plug-in Manager Geo Location Identifier Plug-in API Local Plug-in Manager Growing Data Set Selector Plug-in API Local Plug-in Manager PathFinding Reasoner Plug-in API Local Plug-in Manager SPARQL Endpoint Identifier Plug-in API

Adding Traffic Predictions Goal: Short and Mid-Term Traffic Flow and Speed Forecast Neural Network Architecture: We use a time-delay recurrent neural network to forecast the traffic flow and speed The neural network constructs a minimal set of indicators containing the traffic structure. Proceeding: Data: traffic data (flow and speed) and external inputs (e.g. temperature, holydays) Perform feasibility study to work out specific (prototype) neural network forecast models Develop demonstrator for traffic flow and speed forecasting based on prototype GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino B 1 calendar input t , #24 external input t , #4 A B 2 B 3 calendar input t+1 , #24 external input t+1 , #4 hidden, #8 C 1 traffic t , # 32 0-8h hidden, #8 C 2 traffic t , # 32 8-16h hidden, #8 C 3 traffic t , # 32 16-24h hidden, #8 traffic t+1 , # 32 0-8h C 1 hidden, #8 traffic t+1 , # 32 8-16h C 2 hidden, #8 traffic t+1 , # 32 16-24h C 3 previous day next day A A A A B 1 B 2 B 3

Goal: Short and Mid-Term Traffic Flow and Speed Forecast

Neural Network Architecture:

We use a time-delay recurrent neural network to forecast the traffic flow and speed

The neural network constructs a minimal set of indicators containing the traffic structure.

Proceeding:

Data: traffic data (flow and speed) and external inputs (e.g. temperature, holydays)

Perform feasibility study to work out specific (prototype) neural network forecast models

Develop demonstrator for traffic flow and speed forecasting based on prototype

Dealing with Streaming Data To deal with streams in the Semantic Web context we defined C-SPARQL an extension of SPARQL whose distinguishing feature is the support of continuous queries, i.e. queries registered over RDF data streams and then continuously executed. An example of C-SPARQL query REGISTER STREAM CarsEnteringCityCenterPerDistrict COMPUTED EVERY 5 MIN AS PREFIX c: <http :// linkedurbandata . org/ city #> PREFIX t: <http :// linkedurbandata . org/ traffic #> CONSTRUCT {? district t:has - entering - cars ? passages } FROM STREAM <http :// stream . org/ milantollgates .trdf > [ RANGE 30 MIN STEP 5 MIN ] WHERE { ? tollgate t: registers ? car . ? district c: contains ? street . ? tollgate c: placedIn ? street . } AGGREGATE {(? passages , COUNT , {? district })} GHOSTWAY Event, Vimercate, Milano 27-5-2009 Emanuele Della Valle and Irene Celino

To deal with streams in the Semantic Web context we defined C-SPARQL an extension of SPARQL whose distinguishing feature is the support of continuous queries, i.e. queries registered over RDF data streams and then continuously executed.

An example of C-SPARQL query

REGISTER STREAM CarsEnteringCityCenterPerDistrict

COMPUTED EVERY 5 MIN AS

PREFIX c: <http :// linkedurbandata . org/ city #>

PREFIX t: <http :// linkedurbandata . org/ traffic #>

CONSTRUCT {? district t:has - entering - cars ? passages }

FROM STREAM <http :// stream . org/ milantollgates .trdf >

[ RANGE 30 MIN STEP 5 MIN ]

WHERE { ? tollgate t: registers ? car .

? district c: contains ? street .

? tollgate c: placedIn ? street . }

AGGREGATE {(? passages , COUNT , {? district })}