Friday, February 27, 2015

Pedestrian Crowd Experiment at Monash University

Congratulations to Prof. Majid Sarvi and his team on completing an exciting experiment on pedestrian crowd dynamics today. The complex experiment was held at Clayton campus and was covered by Channel 10 News. Results of the study will be come out in a few months.

Link to the video:

Thursday, February 26, 2015

Monash City Science web analytics by Google Analytics

Web analytics (thanks to Google Analytics) are quite interesting and addictive to follow. During the past week, my research webpage has had 1,695 views from 1,093 viewers mainly because of releasing two new visualizations. 70% of the viewers were from Melbourne, 5% from Sydney, 1.5% from Brisbane, 1% from Canberra, and the rest from different cities in and out of Australia. On average every viewers spent 1 minute and 6 seconds surfing around the page, perhaps interacting with the visualizations.

This is all great to know because I now have a better understanding of where the majority of my audience are and whether really people interact with my maps. I think one minute interaction time is a pretty good number.

Since the beginning, when I first launched the website in June 2014, I have had 27,865 views from 16,304 viewers: 46% from Melbourne, 6% from Sydney, 3% from Brisbane, 2% from London, and the rest from other cities including Adelaide, Canberra, and Perth.

I wonder if we develop a visualization for Sydney or Brisbane, are we going to get more viewers from those cities? I would say yes, probably.

Sunday, February 22, 2015

Connected Future by Telstra

Nice to see "connected" business initiatives by mobile companies such as Telstra. I think Optus has also started something similar in Australia. Picture downloaded from Telstra Enterprise twitter account.

Melbourne Car Ownership Map

I am happy to announce the release of our latest visualization developed at the Monash City Science Research Group: Melbourne Car Ownership Map [link].
An important question in urban planning and travel demand forecasting is where households with fewer or more cars live? and why? In this project, we visualize the spatial distribution of car ownership by dwelling in Melbourne metropolitan area using Census (2011) data. Every dot in the map represents a dwelling.
As expected, households who live within CBD and inner city areas have fewer cars, mostly 0 or 1 car per dwelling. As we go further out to the suburbs, car ownership rate increases. 
Click on the map below to re-direct to the interactive visualization.

Wednesday, February 18, 2015

Visualization: Melbourne Truck AADT Map (2012)

Yesterday I had a discussion with one of my colleagues, Colin Caprani who is a lecturer in the Civil Engineering Department at Monash University on a potential collaboration. Colin is in the structure group. His research specialization includes bridge traffic loading and ITS-infrastructure interaction.

The following visualization is a result of our conversation plus a couple of my hours today. I spent about an hour to find and process the data and a good few more hours to visualize it in CartoDB. What I see in the map is a clear larger truck volumes on freeways, around the Port of Melbourne, and Essendon airport. I could also identify a few other truck corridors in Melbourne such as the Princess Highway, Nepean Highway, Bell Street, etc. Perhaps a more comprehensive analysis could reveal more insightful patterns and findings.

Link to the map:

At first, I was trying to develop a heatmap showing the spatial distribution of truck volumes as an indication of where road damages are more likely to occur. I used "Torque Heat" which is a combination of heatmaps and Torque in CartoDB but I struggled. 

The default torque-aggregation-function in torque heat is count(cartodb_id). Since we wanted to visualize where truck volumes are higher, I changed the function to (round(avg(trucks)). But it didn't work as I expected. See the code below. It took me a few hours to figure out why. The problem was not with the aggregation function; rather, it was the image-filters: colorize-alpha () which caused a confusion. Apparently the colorize-alpha() filter only works with count data. According to its developer website, "Colorize-alpha is an image-filter and works at the layer level. Technically, we accumulate alpha channel values from densely positioned markers and convert this value to the color."

Well, I ended up using a bubble chart which is as informative but simpler to implement. If I figure out how to fix the image-filtering problem, I'll update the map.

CartCSS code for the Heatmap (which didn't give me what I wanted)
/** torque_heat visualization */
Map {
  image-filters: colorize-alpha(blue, cyan, lightgreen, yellow , orange, red);
  marker-file: url(;
  marker-fill-opacity: 0.4;
  marker-width: 35;
#truck_data_2012_updated[frame-offset=1] {
#truck_data_2012_updated[frame-offset=2] {

Tuesday, February 17, 2015

Three Dimensional (3D) Trajectories

Interested to learn more about 3D trajectories? See the following two papers recently published in Transportation Research Record.

Estimating Network Fundamental Diagram Using Three-Dimensional Vehicle Trajectories: Extending Edie’s Definitions of Traffic Flow Variables to Networks
ABSTRACT. This paper evaluates measurement methods for traffic flow variables taken at the network level. Generalized Edie’s definitions of fundamental traffic flow variables along highways are extended for considering vehicles traveling in networks. These definitions are used to characterize traffic flow in networks and form the basis for estimating relationships among network density, flow, and speed in the form of a network fundamental diagram. The method relies on three-dimensional vehicle trajectories to provide estimates of network flow, density, and speed. Such trajectories may be routinely obtained from particle-based microscopic and mesoscopic simulation models and are increasingly available from tracking devices on vehicles. Numerical results from the simulation of two networks, in Chicago, Illinois, and Salt Lake City, Utah, are presented to illustrate and validate the estimation methodology. As part of the verification process, the study confirms that the traffic flow fundamental identity (Q = K • V) holds at the network level only when networkwide traffic flow variables are defined consistently with Edie’s definitions.
How to cite this paper?
Saberi, M., Mahmassani, H., Hou, T., Zockaie, A. (2014) Estimating Network Fundamental Diagram using Three-Dimensional Vehicle Trajectories: Extending Edie's Definitions of Traffic Flow Variables to Networks. Transportation Research Record: Journal of the Transportation Research Board. No. 2422, 12-20.

Exploring Areawide Dynamics of Pedestrian Crowds: Three-Dimensional Approach
ABSTRACT. The main objectives of this paper are to evaluate existing measurement methods of pedestrian traffic flow and to propose a three-dimensional approach that extends Edie’s definitions of fundamental traffic variables to multidirectional walking areas by using three-dimensional pedestrian trajectories. Pedestrian crowds have an areawide fundamental diagram that is similar to a network fundamental diagram of vehicular traffic. Pedestrian traffic in a multidirectional area exhibits hysteretic behavior similar to that of some other many-particle physical systems. Some of the underlying dynamics of bidirectional pedestrian streams are explored with empirical data. Pedestrian streams behave somewhat differently from ordinary fluids with regard to the viscosity concept in the models based on fluid dynamics. The velocity profile for both unidirectional and bidirectional pedestrian streams is hyperbolic (with higher values on the boundaries and lower values in the middle), opposite that of fluids. The formation and dissipation of self-organized pedestrian lanes also are explored. A modification to Helbing’s social force model is proposed with regard to the attractive force between pedestrians.

How to cite this paper?
Saberi, M., and Mahmassani, H. (2014) Exploring Areawide Dynamics of Pedestrian Crowds: Three-Dimensional Approach. Transportation Research Record: Journal of the Transportation Research Board. No. 2421, 31-40.

Dynamics of Urban Network Traffic Flow During a Large-Scale Evacuation

Recently published in Transportation Research Record

This paper explores some of the dynamics of urban network traffic flow during a large-scale evacuation in the context of the network fundamental diagram (NFD). The structure of the evacuation demand can significantly affect network performance. A radial-shaped structure results in a more stable network recovery compared with a directional evacuation structure. This study confirms the existence of unloading-reloading hysteresis when a network is subject to successive cycles of loading and unloading. If a network undergoes a complete or near-complete recovery, the reloading path in the NFD follows almost the same path as in the initial loading. Results suggest that the linear relationship between average network flow and trip completion rate does not always hold, as previously thought. The relationship becomes highly scattered and nonlinear when the network is highly congested and under disruption and the number of adaptive drivers is sufficiently large. Frequent route switching by adaptive drivers can artificially increase the average network flow but does not necessarily increase the network output (trip completion rate). Adaptive driving increases fluctuations in the NFD; however, it reduces hysteresis and gridlock while increasing network capacity.
How to cite this paper?
Zockaie, A., Mahmassani, H., Saberi, M., Verbas, O. (2014) Dynamics of Urban Network Traffic Flow during a Large-Scale Evacuation. Transportation Research Record: Journal of the Transportation Research Board. No. 2422, 21-33.