Foss4g 2013 Oliver May - Subsidized school transport for the shortest way to school

SUBSIDISED SCHOOL TRANSPORT FOR THE SHORTEST WAY TO SCHOOL

Oliver May – GIS developerDFC Software engineering

@olivermay_be

2

DFC Software Engineering

Belgium-based ICT company

http://www.dfc.be/

Located in Ghent, twin city of Nottingham

City ofGhent

Belgium-based

images: (c) Oliver May, BBC, http://en.wikipedia.org/wiki/File:Council-House-Nottingham.jpg

City ofNottingham

3

About me

Java Developer at DFC since 2008 OSS enthusiast

Geomajas Active contributor to the project Member of the Project Steering Committee

4

The project

SUBSIDISED SCHOOL TRANSPORT FOR THE SHORTEST WAY TO SCHOOL

5

The project

Facts

Students in special education get free school transport if they go to the closest school

Problem?

Processing applications is very time consuming Results may be inconclusive Why?

6

Time consuming – Why?

Paper applications via snail mail

7

Time consuming – Why?

“Calculating” closest school on printed map

Later using Google maps, Mappy, ...

8

Inconclusive – Why?

“Calculation” of closest school is done by a person

Prone to error

By wrong measurement By using different maps and online services

time vs. distance

Causes applicants to question the decision

9

The requirements

Digitalize the request form and work-flow Online web application

Automate the process as much as possible Fixed and known road data Shortest distance algorithm

Allow abnormalities in the process Ex: siblings, school full, boarding school,...

10

The process

School submitsonline application

Closest?

Accept application

Notify via email &send certificate

Yes

Abnormalities?

No

Reject application

Notify via email

System calculate school distances

Officer checks parameters manuallyYes

No

11

The process

School submitsonline application

Closest?

Accept application

Notify via email &send certificate

Yes

Abnormalities?

No

Reject application

Notify via email

System calculate school distances

Officer checks parameters manuallyYes

No

Less than 5% of applications

need to be checkedmanually

12

Architecture

Web frontend Javascript

Back-end server

DatabaseRoutingengine

geocoding

application logic

Pdf generation

13

Architecture - technologies

Web frontend Javascript

Back-end server

DatabaseRoutingengine

geocoding

application logic

Pdf generation

iTextSOAP WSGovernment

14

Technology choices

Geomajas Known technology (Java) Able to model and display domain logic Web based Hibernate integrated Built in GIS services (Geotools)

pgRouting Need for out of the box routing on spatial database Didn't want to re-invent the wheel

15

GIS data layers

SchoolsSchoolsSchools

SchoolsSchoolsStudent requests

SchoolsSchoolsRoads

700 schools

~350K road segments

~50K requests / year

16

The process

How did we design and implement the different steps in the process?

17

The process: Application form

School application form

Name, grade, study program Application details (weekly, daily,

type of transport, etc. ) Address (geocoded trough web

service) Municipality Street Number

18

The process: Manually geocoding

Problem: geocoding database might not contain address!

Allow entering of free text for manual processing

Manually process by pointing the location on a map by an officer

19

The process: Calculating the routes

For every school, the distance to the applicant's address is calculated using pgRouting

The closest school wins

Optimisations:

Don't bother about schools that are further as the crow flies than the closest school so far

20

The process: Calculating the routes (2)

Addresses need to be mapped to nodes 1. Find closest street 2. Calculate route from start and from end node

Consequence: we need to calculate up to 4 routes

21

The process: Calculating the routes (3)

The actual pgRouting query:

Arguments

'SELECT...' → the topology to route on source, target → nodes to calculate between directional calculation, reverse cost

22

The process: Manual processing

List of applications to manually process

Processes individual applications, based on abnormalities

Optionally provide feedback for the school

23

The process: Accepting applications

Fill out pdf document

Sign digitally with certificate

iText

24

The process: Rejecting applications

Same procedure

Attach a list of closer schools

25

The requirements

Digitalize the request form and work-flow Online web application

Automate the process as much as possible Fixed and known road data Shortest distance algorithm

Allow abnormalities in the process Ex: siblings, school full, boarding school,...

26

Results

The processing time of applications was reduced from weeks to minutes

Less discussion about rejections, because the data and algorithms are proven to be correct

27

And further

Real system was used for “simulations”

Interactive simulation tool was foreseen http://www.rechtopleerlingenvervoer.be/dichtstescholen/

28

Questions? Contact me:

[email protected] @olivermay_be At the geosparc booth (Atrium)

Thank you for your time!