transferring link data between networks · the integrated transport network (itn) and traffic...

44
Transferring Link Data between Networks An Algorithmic Approach Steven O’Hare Senior Transport Modeller & Practice Champion for Modelling Mott MacDonald

Upload: others

Post on 13-Oct-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Transferring Link Data between NetworksAn Algorithmic Approach

Steven O’Hare

Senior Transport Modeller & Practice Champion for Modelling

Mott MacDonald

Page 2: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Transferring Link Data between Networks

Typical project

• Inherit an existing model of an area with the intention of using it to model a transport intervention

• This requires significant updates to both the supply and demand representation, including the networks

• Once updates applied, model needs to be validated to show that it provides a reasonably accurate representation of reality, which includes journey times/speeds

The Challenge

• Doing all of the above requires data and several important data sources come in network form e.g. the Integrated Transport Network (ITN) and Traffic Master Journey Time Data

• However, the network representation for the data seldom overlaps with the network representation in the model and does not come with a correspondence list to transfer data across

• To make use of this data we have to find a way to make this transfer

15 June 2018Mott MacDonald | Presentation 2

Context

Page 3: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 3

Page 4: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 4

Page 5: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Typically Current Practice

Approach

• One or two unlucky graduates are tasked with manually transferring data between the networks

• This is:

• Extremely time consuming

• Boring and monotonous for the graduates involved

• Prone to human error, which is often randomly distributed and difficult to detect

Implications

• Process is so time consuming that transfer is limited to a select number of important routes

• This does not extract the full value that is available in the data

• Need a method to get data for all of the links transferred across en masse using a technique that is reliable and not resource intensive

15 June 2018Mott MacDonald | Presentation 5

Manual Transfer

Page 6: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

The GIS Option

Approach

• Use of geoprocessing techniques in GIS packages to create relationships between links

• Relationships based on their spatial characteristics

• E.g.

• Search from the midpoint of each link in the model network to find the closest link in the data network

• Search out from each link in the model network to find all links in the data network within a certain radius and then accept relationships where link orientations are similar

Problems

• Parallel links

• Elevated carriageways and tunnels

Page 7: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

The GIS Option

15 June 2018Mott MacDonald | Presentation

A challenging example

7

Page 8: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

An Algorithmic Approach using Path Searching Algorithms

15 June 2018Mott MacDonald | Presentation 8

Node Match

• Develop a node based correspondence list between model network and data network

• Match nodes that are within a distance threshold of each other

Link to Link

Link to Path

Path to Path

Apply Hierarchy

• Identify links in model network where a) start and end nodes have matches and b) where the

matched start and end nodes in the data network form a link in that network

• Data transfers directly to model network link

• Identify links in model network where a) start and end nodes have matches and b) where the

matched start and end nodes in the data network do not form a link (but a path) in that network

• Aggregate data from path in data network and transfer to model network link

• Identify links in model network where only start or end node is matched with a node in the data

network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

• Where multiple allocations for a model network link prioritise in following order: 1) Link to Link,

2) Link to Path, 3) Path to Path

Page 9: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 1: Node Match

Method

• Use spatial characteristics to create a correspondence list between nodes in model network and nodes in data network

• Accept matches within a defined distance threshold

Potential Criticism

• Relies on accuracy of spatial characteristics and GIS techniques

• However:

• Node relationships not as complex as links

• Distance cut-off to limit incorrect matches

Page 10: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 2: Link to Link

Method

• Links in model network where a) start and end nodes have matches and b) where the matched start and end nodes in the data network form a link in that network

Model Network

Data NetworkMatch Match

Page 11: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 2: Link to Link

Method

• Links in model network where a) start and end nodes have matches and b) where the matched start and end nodes in the data network form a link in that network

Model Network

Data NetworkMatch Match

Page 12: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 3: Link to Path

Method

• Identify links in model network where a) start and end nodes have matches and b) where the matched start and end nodes in the data network do not form a link (but a path) in that network

• Aggregate data from path in data network and transfer to model network link

Model Network

Data NetworkMatch Match

Page 13: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 3: Link to Path

Method

• Identify links in model network where a) start and end nodes have matches and b) where the matched start and end nodes in the data network do not form a link (but a path) in that network

• Aggregate data from path in data network and transfer to model network link

Model Network

Data NetworkMatch Match

++

Page 14: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

Page 15: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

Page 16: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

Page 17: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

Page 18: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

Page 19: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

dA

Page 20: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

dA

dB

Page 21: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

dA

dB

Page 22: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Method

• Identify links in model network where only start or end node is matched with a node in the data network. Apply breadth first search to find a second matched node

• Find shortest paths between node pairs and use distance weighted data transfer process

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

dA

dB

LA1A2 LA2A3

LB1B2 LB2B3 LB3B4

Page 23: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

dA

dB

LA1A2 LA2A3

LB1B2 LB2B3 LB3B4

Value on Link A1A2 = X + Y * ( (LA1A2 – LB1B2) / LB2B3 )

X Y Z

Page 24: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 4: Path to Path

Model Network

Data NetworkMatch Match

A1

B1

A2 A3

B2 B3 B4

dA

dB

LA1A2 LA2A3

LB1B2 LB2B3 LB3B4

Value on Link A2A3 = Y * ( (LB2B3 + LB1B2 – LA1A2) / LB2B3 ) + Z

X Y Z

Page 25: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Step 5: Apply Hierarchy

Method

• Where multiple allocations for a model network link prioritise in following order:

1. Link to Link

2. Link to Path

3. Path to Path

Page 26: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Example Application

Greater Manchester SATURN Model

• Used to investigate impacts of a motorway scheme

• Need to transfer Traffic Master Journey Time data from an ITN layer base onto the SATURN model network for the purposes of:

• Calibrating speed/flow curves

• Validating modelled journey times against observed data

Page 27: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 27

Page 28: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 28

Page 29: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 29

Page 30: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 30

Page 31: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 31

Page 32: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 32

Page 33: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 33

Page 34: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 34

Page 35: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 35

Page 36: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 36

Page 37: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 37

Page 38: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 38

Page 39: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 39

Page 40: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Transferring Link Data between Networks

Issues

• Identify how to measure success of the method

• Additional step required to exclude links in data and/or model network that do not have matched pair. This was a late addition to the example application shown here

Future Work

• Fine tuning of parameters/thresholds and also sensitivity testing to better understand behaviour

• Applications to other data sources with other network based representations

15 June 2018Mott MacDonald | Presentation 40

Future Work

Page 41: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Transferring Link Data between Networks

Current Environment

• Increasing processing challenges posed by large and complex datasets

• Opportunity cost: the more time we spend processing data to make it usable, the less time we get to spend adding value elsewhere, e.g. building “better” models, doing more sensitivity tests

Conclusion

• Big datasets are good and are potentially really useful but…

• … we can only extract their full value if they come with good translation tools to get them into a base that matches up with other data sources

15 June 2018Mott MacDonald | Presentation 41

Broader Thoughts

Page 42: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

Thank you

[email protected]

Page 43: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 43

Page 44: Transferring Link Data between Networks · the Integrated Transport Network (ITN) and Traffic Master Journey Time Data • However, the network representation for the data seldom

15 June 2018Mott MacDonald | Presentation 44