harnessing the power of high performance computing to assess and

45
Harnessing the power of high performance computing for creating sustainable, resilient, and liveable cities Maria-Cristina Marinescu Barcelona Supercomputing Center, Spain Jorge Garcia Vidal Universitat Politecnica de Catalunya, Spain

Upload: vuminh

Post on 14-Feb-2017

215 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Harnessing the power of high performance computing to assess and

Harnessing the power of high performance computing for creating sustainable, resilient,

and liveable cities

Maria-Cristina MarinescuBarcelona Supercomputing Center, Spain

Jorge Garcia VidalUniversitat Politecnica de Catalunya, Spain

Page 2: Harnessing the power of high performance computing to assess and

2

R & D Areas

We develop HPC software for science and engineeringWe do research in various areas: modeling, simulation, visualization

Page 3: Harnessing the power of high performance computing to assess and

3

R & D Areas

We develop HPC software for science and engineering

We do research in various areas: modeling, simulation, visualization

Integrate and analyze data

Simulate

Visualize

Urban processes and structures

Dicover patterns Compute metrics (liveability, resilience, contamination, etc) Understand the effect of changes/decisions beforehand Predict (timed) events offline and in real time

To plan, predict, react...

Page 4: Harnessing the power of high performance computing to assess and

Data integration

Page 5: Harnessing the power of high performance computing to assess and

Goal

● Urban Management and Planning– Data Integration from multiple data sources (errors, uncertainty, ..)

– Query facilities

– GIS Visualization

● Big Data Solutions if needed for performance

● Ecological Urbanism Paradigm– Environmental, social, and economic sustainability and resilience of cities

– Evaluated by specific indicators (KPI)

● Standard-based Semantic Data Model for Smart Cities

Page 6: Harnessing the power of high performance computing to assess and

Barcelona Urban Ecology Agency

● Ecological Urbanism: degree to which a city adheres to the following principles

Efficientland useEfficientland use

Quality ofPublicSpace

MobilityAnd

Services

HabitabilityIn Housing

and Buildings

Biodiversity

Social Inclusion

and Interaction

Self-Sufficiency

Organization

Page 7: Harnessing the power of high performance computing to assess and

City Semantics

7

Urban City Model• Ontology to include urban models• Instantiation engine (Open Linked Data repositories)• SPARQL queries• Integration with openGIS

RDF graph

Ontology for smartcities (OWL-DL)

SPARQLquery

Data(RDF, XML, GML, JSON, etc)

$RDB

XML

SQL query

Xpath,Xquery

IoT

internet

Page 8: Harnessing the power of high performance computing to assess and

Example Scenario

Ratio of health related incidents per medical center within given neighbourhood?

Relate Data

Integrate with GIS

Measure

Query and

Visualize

Page 9: Harnessing the power of high performance computing to assess and

Technical Benefits

● Open World Concept– Valid data even if there are unknown features (e.g. we cannot know the

number of residents of a new building complex but cannot assume there are no residents) – semantically precise although incomplete

● Multi-classification– E.g. a street can be pedestrian or not, depending on the hour of the day

● Automatic inference– E.g. A public school is an educational facility, a facility, a public building,

a building, a geolocalizated structure, etc.

● Concept Browsing● Allows the access and the sharing of the City Data via Web● Eases data sharing between cities using the same city model● Uses a Standard Query Language (SPARQL)● GIS Capabilities

Page 10: Harnessing the power of high performance computing to assess and

● Modeling relevant and shared city data● Why is this important?

– It simplifies the development of applications that require integrated access to city data sources (cross-domain)

– It enables solution reuse as we move from one city to the next

– It allows extending the metadata with new categories (Sanitation, Crime) without modifying the application or the data sources

● In terms of resilience, it facilitates:

– Data pattern discovery – prediction tool● E.g. Conditions under which an event triggers, correlations between monitored parameters

– Computing quality metrics – analysis tool

A Semantic Data Model for Smart Cities

Page 11: Harnessing the power of high performance computing to assess and

Data integration and inference

E.g: Integrate transportation networks -metro, bus, train, …

Scenario 1: An accident on the Vittorio Emmanuele – Termini tramSolution: use existent connection via train through Tuscolana

Scenario 2: An accident on the Lepanto -Flaminio tramSolution: introduce temporary bus line

Scenario 3: explosion on metro tram that affect electricity networkIF additionally we integrate electricty network we know potentially affected neighbourhoods… where there may be a hospitalSolution: prepare generators for hospital

Page 12: Harnessing the power of high performance computing to assess and

Simulation

Page 13: Harnessing the power of high performance computing to assess and

13

Atmospheric Simulations

Atmospheric Transport - FALL3D– Forecast of volcanic ash for air control

Atmospheric physics - modelling– Cloud formation

Page 14: Harnessing the power of high performance computing to assess and

Atmospheric Transport at Microscale (CFD)

Page 15: Harnessing the power of high performance computing to assess and

CFD Indoors/Outdoors

Page 16: Harnessing the power of high performance computing to assess and

CFD Indoors/Outdoors

Page 17: Harnessing the power of high performance computing to assess and

CFD Indoors/Outdoors

Page 18: Harnessing the power of high performance computing to assess and

Water Flooding

18

Page 19: Harnessing the power of high performance computing to assess and

LIDAR Geometry example

Page 20: Harnessing the power of high performance computing to assess and

City Geometry for Simulation

Page 21: Harnessing the power of high performance computing to assess and

City Geometry for Simulation

Page 22: Harnessing the power of high performance computing to assess and

Agent Based Simulation

Page 23: Harnessing the power of high performance computing to assess and

Agent Based Models

Useful for – Social simulations– Crowd simulations– Mobility simulations (cars+pedestrian+…)– Epidemic evolution– . . .

But ABM scale bad in parallel computers, then there are serious limits in:– Agents complexity– Number of agents

Page 24: Harnessing the power of high performance computing to assess and

Agent-Based Modeling

● Bottom-up approach based on the creation of a group of actors that interact inside a defined environment.

● Evolutionary approach

● Agent Heterogeneity

● Complex Behaviour

● Environment

Page 25: Harnessing the power of high performance computing to assess and

Classical approaches to markets

● Physical-based Models– The actors of the system are forces, and the system looks for an

equilibrium

● Game Theory– The actors of the system are players that try to maximize their profits.

How can we model...

Heterogeneity? Behaviour? Realistic Scenarios?

Page 26: Harnessing the power of high performance computing to assess and

27

PANDORA

PANDORA: Scalable execution of agent-based models– Able to run models with millions of agents– C++ framework for large-scale social simulation– Automated generation of parallelized code– GIS support (Grass), Statistical support (R package)– Python scripting to define ABM– Cassandra: Cassandra is an analysis tool, implemented to interpret

results and detect patterns

Page 27: Harnessing the power of high performance computing to assess and

Potential Applications

Smart cities

Financial Markets

Policy Analysis

Traffic Simulation

Prediction in dynamic environments

● Explore mixed numerical – ABM simulations for decision making in case of emergency

Understand social activities

Page 28: Harnessing the power of high performance computing to assess and

Visualization

Page 29: Harnessing the power of high performance computing to assess and

Pasive (predictive) big data analysis

Page 30: Harnessing the power of high performance computing to assess and

Big Data Visualization

Page 31: Harnessing the power of high performance computing to assess and

Big Data Visualization

Page 32: Harnessing the power of high performance computing to assess and

Data visualization for crisis management

• Data dashboards: Clarity and recognition speed are main goal• Intuitive displays needed to communicate maximum of relevant

information as immediately as possible (it is an operational, NOT an analytical tool)

• Heterogeneous team of data designers and interface experts• We are developing technology for innovative and collaborative

interfaces

Page 33: Harnessing the power of high performance computing to assess and

Active data visualization and processing

• Use modern means of communication for:• Fast and cheap one way communication (govt to

citizens)• Empower citizens for massive feedback

• Example: Blackouts during hurricane Sandy• Integrate power visualization with online big data sources

Page 34: Harnessing the power of high performance computing to assess and

Massive data collection, visualization and processing

• Measure citizen response to events (even in real time)• Example: “Happiness” can be extracted from social network

geolocated activity

Page 35: Harnessing the power of high performance computing to assess and

Questions?

Page 36: Harnessing the power of high performance computing to assess and

Digital

Data

Model

Geo Data

Semantic

City Data

Data Integration Different formatsHidden semantic relationships

Relate Data

Integrate with GIS

Data preprocessing

Map data to modelSpecialize model

for new city

Page 37: Harnessing the power of high performance computing to assess and

SemanticCity Data

DataProcessi

ng

QueryTechnol

ogy

Graphical

Interfaces

Data Processing and Visualization

Quality metrics

Average distance to recycling points per habitant

Green zones per habitant and block

Measure

Query and Visualize

Page 38: Harnessing the power of high performance computing to assess and

An ontology is …

Conceptually a graph– Precise semantics– Standard representation to allow interoperability (triples/ Linked

Open Data)• Same URI means same resource

– Models an open world, i.e. allows incomplete information• Adding edges / vertices is easier than adding information into tables

(which may require refactoring)

– Understandable and uniformly accessible from anywhere

Amenable data structure for– Defining data constraints– Discovering unknown relationships via logic inference

Page 39: Harnessing the power of high performance computing to assess and

Integration of Open Linked Data from any source

Composition of RDF graphs is another RDF graph

Page 40: Harnessing the power of high performance computing to assess and

SCRIBE● Open Source model developed by IBM

● Non-normative, authoritative, modular, extensible semantic model for Smarter Cities

● Consist of a Core Model that includes common classes (events and messages, stakeholders, departments, services, city landmarks and resources, KPIs, etc.)

Simple language Based on standards (OWL-QL, SPARQL) from the W3C Metadata annotations and Tagging

Authoritative Aligned with standards (CAP, NIEM, MISA/MRM, UCore) Validated with customer scenarios Validated with open city data

Page 41: Harnessing the power of high performance computing to assess and

City Data Analysis and Planning Tool

http://...Selected ElementsProperty: year-built

BuildingRoad…

196519701920…

Textual query

Numerical Result

BuildingYear-built: 1965Year-remodeled: 1970Area: 250m2…

Road Green Buildings: match the condition

Building data

Buildings

Result: 6,3m.Avg (height of Building with (year-built > 1960) OR (year-remodeled > 1980))

Search

Conditional query: Avg Building height

year-built > 1960 ORyear-remodeled > 1980..

Page 42: Harnessing the power of high performance computing to assess and

Monitor

Discovery

Planning & new quality metrics

• Changes in mobility network

• Best places to create business

• Neighbourhoods with insufficient green areas

• Conflictive neighbourhoods

• Violence indicators

• Pollution levels

User Benefits

Page 43: Harnessing the power of high performance computing to assess and

A Semantic Data Model as an Ontology

● Semantic network of Concepts– Model: Class + Relations + Constraints

– Knowledge base: Model + Instances

– InferenceAirport Address

Calle 26, nº103-109Bogotá, ColombiaEl Dorado

Geo

Facility

04-43N, 074-09W

Domain:FacilityRange: Address

Domain:FacilityRange: Geo

is-a

typetype type

Inferredtype

hasGeopositionhasAddress

Page 44: Harnessing the power of high performance computing to assess and
Page 45: Harnessing the power of high performance computing to assess and

An Agent-Based market

● “the agent-based method can provide an unprecedented understanding of the emergent properties of interacting parts in complex circumstances”

Farmer & Foley, The economy needs agent-based modelling, Nature, 2009)

● Simple modeling of complex emergent properties:● Self-fulfilling expectations● Asymmetric information● Non-equilibrium situations

● We would be able to evaluate different scenarios...● ...what if we introduce a new competitor?● ...what if a regulatory agency is created?