FME World Tour 2015, Boston. April 14th, 2015

Spatial Decision Support and Analytics on a Campus Scale:

Bringing GIS, CAD, BIM and Tableau Together

Alexander StepanovGIS ArchitectA&F Administrative Systems

Niels La CourSenior Planner Campus Planning

University of Massachusetts, Amherst

Introduction

• Smart Campus (what is “Campus”?)

• Analytics: Descriptive, Predictive & Prescriptive (being ‘smart’

through optimization of operations and use of existing

resources).

• What are prerequisites for bringing optimization and operations

research (OR) methods to support decision-making process

• GIS as enabling technology

Campus Environment

Campus is a complex system (decomposition)

Processes

• Academic/Teaching

• Research

• Operations

• Residential Life

“No buffer space”

“Changes all the time”

“Divide & Conquer”

Campus as Systems: Buildings

Class Scheduling

ADA Pickup Locations

3. Campus is a Complex System

Spatial Data: CAD, GIS, BIM, NavisWorks

Non-Spatial Data: MS SQL Server, SharePoint, Excel, Other DBs, Text files

Diff ITSystems

• Specialization (to cope with complexity)• Multiple databases• “Silos” & “Silos Paradox”• Living organization/changes!

Location/Space & Time as KEY to join and relate entities/features in the DBs.

How can we create & maintain ‘comprehensive’ spatial model/db of Campus efficiently.

4. Data & Analytical Workflows (WF)Client

Applications & Systems

DesktopMobile

Web

3D ModelingOPTIMIZATION [ SAS/OR]

SCENARIO VISUALIZATION

ANALYTICAL TOOS

Spatial Data Warehouse

Set of ETL workflows

“PULL” system vs “PUSH System”

LEAN approach

Principles of such integration (CAD&GIS)

… given the fact that

majority of the

existing buildings

data are in CAD/BIM

systems/formats.

2.2 3D Modeling: Tools: FME

Library of Transformers Data Sources Data DestinationsTransformers

ETL = Extract – Transform - Load

~300 formats in

~250 formats out

Ability to embed Python sci libraries

Easy to express ideas

4. Data & Analytical Workflows (WF)

Client Applications &

SystemsDesktopMobile

Web

3D ModelingOPTIMIZATION [ SAS/OR]

SCENARIO VISUALIZATION

ANALYTICAL TOOS

“BRAIN”

“HEART”

“LUNGS”

Decision Problems on Campus Scale

Decision Problems on Campus Scale

• Within Spatial Context

• Effect ‘other’ sub-systems

• KPIs have spatial component

• Awareness of trends/projects

• Require integration of multiple data-sources

• Snapshot of performance matrix

• Snapshot of current ‘state’ of the system

Example: “Backfill” Space Assignment

New building is built on campus

Researchers/Groups are moved to the new building, vacating their offices and labs.

What would be an optimal policy to assign the vacant space to other groups/users?

Demand - Supply - Assignment

What parameters and constraints? Performance metrics?

Everything has a SPATIAL component

“Backfill” Space Assignment“Backfill” Space Assignment

Demand SupplyCij

Xij = {0,1}

Creating room/space inventory

3.1 Spatial DB of Space Inventory

Input:CAD Floor PlansCAFMGIS Layers with BLDGfootprintsElevation profiles

3 step process: (3 feature classes)

3.1 Unique Campus Space ID & 3D representation of rooms

0318_04_0020A

Spatial database is a collection of rooms(spaces), which are registered in physical space and have unique “global” campus ID.

Connection to the enterprise DB

From CAD Space to ARCH SpaceOrigin (0,0) is shifted to specific tile/cell

STEP 1: CAD to GIS (Architectural Space)

Every SPACEhas PRIMARY KEYBLGD_FLOOR_ROOM

ETL: CAD to Arch Space

2D (2.5D) Inner Space Modeling: “Architectural Space”

2D (2.5D) Inner Space Modeling

Step 2: Getting Shift, Scale and Rotation for Buildings

Step 2: Getting Shift, Scale and Rotation for Buildings

Step 3: GIS “Arch” to GIS “Real World”

Every SPACEhas PRIMARY KEYBLGD_FLOOR_ROOM

Ability to JOINExternal datasets

ETL: Arch to GIS Space

2-way translation: ETL to add/update data in CAD via GIS (“Breadcrumbs”)

2D (2.5D) Inner Space Modeling: 2-Way process

Each space/room has LOCATION_ID to join with DBs

Room/Space Spatial Database

3 Feature classes:

Floors (~1,800 )Rooms (~32,000)Arch Details (windows, doors, stairs) ~ 8 million

Every feature has LOCATION_ID

Buildings Networks I(“Pine branches”)

• A building is more than just set of

Rooms/spaces.

• From inventory/reporting to

optimization and analysis

• Vertical & Horizontal Conveyors

Example2: Indoor Routing & Navigation

Buildings as

complex systems

Getting ideas for modeling building networks

IBM (Supply Chain Optimization)

Network/Graph G(N,E)

edges E:Eh - horizontal edges (room -> floor

centroid)Ev - vertical connections b/w floors

nodes N:Nr - room centroidsNf - floor centroids

G(V,E) should be connected to pedestrian networks.

Testing our ideas for modeling building networks

Workflow to build the network/graph G(V,E)

1 2

34

5

5

6

Input:

FC - Rooms (~32K)FC - Floors (~1,8K)

Output:

FC - Nodes (Pnts)FC - Edges (Lines)

Network/Graph G(V,E)

|E| ~ 32,000 (horizontal) +1,800 (vertical)

|V| ~ 32,000

Full set of attributes for rooms/floors

Processing time: ~ 2 min.

Connectivity to pedestrian/transportation networks

Bridging the “scale” divide (geo-design)Demo 2

Buildings Networks II

Building/Floor topology defines the performance metrics

(“Pine branch” networks are not enough)

IBM (Supply Chain Optimization) ESRI (Travel Time along actual road network)

Modeling paths between rooms

Source: J. MacGregor Smith (MIE, UMass Amherst)

Developing networks to model crucial building functions

Space Assignment

Location/Allocation

User/Space Interaction

2D/3D Routing

Evacuation Planning

Accessibility

Building Indoor Networks

Connection to ‘outdoor’ sidewalks/roads

RECTILINEAR STEINER MINIMAL TREES (RSMT) with Obstacles

•Rooms

•Doors

•Corridors (Circulation )

•Walls ( Obstacles )

Prof. J.MacGregor Smith (jmsmith@ecs.umass.edu) CAD/DB

Network Representation

(Floor/Building)

Example3: Indoor Routing & Navigation

Rectilinear Steiner Minimal Trees

is a graph connecting

all rooms/corridors with edges and

ensures min. total length

RSMT algorithm

(Prof. Smith J.M)

Data InterOp Extension

Example 3 RSMT: IMPLEMENTING MST with Custom Transformer

21 rows of code to implement MST transformer

4 rows to solve Minimum Spanning Tree sub-problem using NetworkX (LANL ) (http://networkx.lanl.gov/)

Example 3 RSMT: IMPLEMENTING MST with Custom Transformer

CAD to GIS (+ text/data mining)

3.2 EXTRACTING TEXT INFORMATION

SEE U702-2 (1955) LOC. APPROX.

6,000 notes

30 year of experience

Period 1940s – current

4 layers: STEAM/ELECTRIC/WATER/DRAINAGE

Non-structured text

3.2 EXTRACTING TEXT INFORMATION - FME WORKFLOW

Regular expressions allows one to do advance search for patterns ( learn once, re-use everywhere)

We use the same approach to extract utility attributes from CAD annotation layers

3.2 EXTRACTING TEXT INFORMATION

8 transformers ( + 8 transformers to parse file system for scanned documents )

3.2 EXTRACTING TEXT INFORMATION

Integration of Tableau (BI) and Spatial Data on Campus Scale

Gartner's Magic Quadrant for Business Intelligence and Analytics Platforms

Tableau is a powerful and

an intuitive BI package.

Many examples and use

cases are based on a

regional scale.

How we can integrate it

with GIS on a Campus

scale?

http://www.tableau.com/solutions/higher-ed

Tableau in Higher Education

Default Tableau map (service) UMass-Amherst GIS map ( + 200 Buildings)

How we can approach reporting on building level with Tableau?

a) How to bring GIS Building footprints into Tableau shapes?

b) How to bring centroids of Buildings as X/Y (lat/lon) into Tableau?

Example of case use: Teaching/Instruction/Operations happen in Buildings …

Converts BUILDINGS footprints from GIS representation to SQL Table

which represents a line by storing shape_id, point sequence and lat/long.

Most complex part is “VertexCounter” transformer ….

Available as a free custom transformer from FME Store …

Programming time – 0.

Example of simple FME workflow to bring any polygon/line GIS geometry into Tableau

Building Footprints in Tableau.

Now it’s ready to be connected with data tables and to run analytics

FME to convert Building Footprints to centroids (Polygons -> X/Y centroid -> Lat/Lon):

Student distribution at 10am on Thursday (Fall 2013)

Buildings and Campus as systems: GIS as Decision Support System

UNIQUE ANALYTICAL ABILITIES TO DECISION MAKERS

http://gis.larc.nasa.gov/spaceopt/overview.html

• Energy Use Optimization

• Facilities Management

• Asset Management (Equipment)

• Location/Allocation of Resources

• Public Safety and Security

• Maintenance

• Space and Group Management (NIH, the largest research organization)

• Optimization of Space Utilization

• Improve Collaboration

• Indoor Navigation and Routing

• Event Scheduling (‘time table for College’)

• Urban Planning

• Landscape Architecture and Design

• Sea Ports, Airports, Shopping Centers …

Conclusion

• GIS is an enabling technology for prescriptive analytics (Operations Research).

• GIS facilitates research collaboration among multi-disciplinary teams

• Geospatial approach dissolves silos within organization and stimulates organization-wide awareness.

• Spatial DB/ETL coupling allows to bring existing analytical tools and libraries to transform/analyze data.

• (Format agnostic, data centric)

• GIS approach may encourage collaboration among academic, research, operations and industry.

Acknowledgments

•Campus Planning, D&CM, UMass Amherst

•Prof. James MacGregor Smith, Department of Mechanical and Industrial Engineering

•Lukasz Czarniecki, Jonathan Contract, Niels La Cour and Ludmila Pavlova, Campus Planning

•Safe Software (http://www.safe.com)