Injury severity prediction from two-vehicle crash mechanisms with machine learning and ensemble models

  • Ji, Ang and Levinson, D. (2020) Injury severity prediction from two-vehicle crash mechanisms with machine learning and ensemble models. IEEE Open Journal of Intelligent Transportation Systems. [doi]

Machine learning algorithms aim to improve the power of predictors over conventional regression models. This study aims to tap the predictive potential of crash mechanism- related variables using ensemble machine learning models. The results demonstrate selected models can predict severity at a high level of accuracy. The stacking model with a linear blender is preferred for the designed ensemble combination. Most bagging, boosting, and stacking algorithms perform well, indicating en- semble models are capable of improving upon individual models.

The New New Normal: Mobility and Activity in the `After Times’

We may be nearing “peak city”. This shift undermines all of the place-based strategies that economic development organisations have been promoting for decades. It’s a topic David Levinson will be addressing at the Festival of Urbanism 13-26 November. This article was originally published in The Fifth Estate, November 2, 2020.

The Dot-Com boom and Y2K crisis compressed a decade of technological investment into a two-year period. As a bubble, it was naturally followed by a stock market crash, but the Internet is bigger and more important than before. Since Y2K we have seen the advent of smartphones (the Internet in our pocket), social media, Wikipedia, ride-hailing (mobility on demand in our pocket) and a major refactoring of many if not most businesses around the new information reality.

COVID-19 has initiated as profound a transformation, acculturating the population to videoconferencing and  work from home and virtual conferences and webinars and distance learning and online shopping (for everything) and remote medicine and  Zoom weddings and funerals and many other activities that once would have been experienced in person. This transformation did not begin from scratch, all of these changes had already begun, but they were accelerated by events.  The home has at least temporarily returned to its historic pre-urban role of being the restaurant, the workplace, the schoolhouse, the theatre — making it more crowded and more intense.

Eventually, like Y2K, COVID-19 will cease to be an issue and we will enter the `After Times’. The fear of others will dissipate but not disappear.  Its effects on society will linger. 

Many people who previously stood all day at a hot-desked, open plan, office in the CBD  have discovered they like working at home at least some of the time, so long as they also don’t need to supervise children.  Couple that with finding that 37% of all jobs in the US can be done entirely at home, 39% in Australia, a greater share of those in metro areas, and a greater share still in Central Business Districts. Many more jobs can be engaged with at least part-time at home. Thus, many firms  have discovered they don’t need to pay for expensive CBD real estate. Soon governments and universities will discover the same. Many shoppers and eaters like getting deliveries rather than going out, and distribution infrastructures have scaled up to enable this. Dark kitchens designing meals for delivery will ensure they are no longer mere allusions to tasty food.  Many students find they like attending class from home, which may be China, and now they are able to.

It’s not that some of the people don’t miss some of the ways of the `Before Times’ — commuting at least some times provides a spatial and mental separation between home and the workplace. In a study we conducted during the Sydney lockdown (Aoustin 2020), almost everyone was traveling less, and respondents were asked how they experienced the decrease in time spent in transport. On a scale of 1 to 7, 51% said they missed this time little or not at all, 17% were indifferent, but 32%  did miss it. Travel to work does not need to take place at 8:00 am five days a week, 50 weeks a year, packed into a shiny metal box.

This prospective future no doubt would come as a disappointment to Urban Triumphalists, who insist the value of cities is due to economies of agglomeration resulting from face-to-face interaction, as well as organisations like the Property Council. While historically in-person contact has driven economies of agglomeration, and why be in cities but to be near other people, the question remains: Must it always be so? Mega cities were largely non-existent in the pre-Industrial Revolution period when the economies of agglomeration were often outweighed by the diseconomies. Cities will not be abandoned quickly, transitions are long, but we may be nearing `peak city’. This shift undermines all of the place-based strategies that economic development organisations have been promoting for decades. 

Face-to-face encounters will remain important for a few sectors, those with high trust issues like politics, or requiring hands-on physical interactions, it is clearly being abstracted away. The risk we as a civilisation face is that of the explore/exploit trade-off – if the urban triumphalists are correct that new ideas emerge more from in-person interaction, while we can continue to exploit existing ideas, we will generate fewer of them and the rate of progress will slow. 

I am not convinced this is true, certainly not to the degree it once was. Perhaps more ideas can be generated from the vast increase of total interactions online, even if those interactions are intermediated electronically, than the serendipitous in-person encounters traditional place-based thinking privileges. This is not necessarily Zoom meetings, certainly not group meetings, but may instead be real-time (or nearly so) text-based interactions, bulletin boards, messaging, Slack, Twitter, Miro, and the like.

The implications of these changes on physical places are several. 

If more is to be done at home for more of the day, the demand for more space per person at home will increase, and the demand for person space in offices will diminish. While physical distancing requirements may remain at offices, giving those in the office building more space as well, it also drives up costs of offices, further inducing firms to increase their virtuality. The demand for new office construction will be permanently reduced, and we may see buildings or sites retrofit for other purposes, perhaps residential, as they run through their lifecycle. The demand for housing in contrast will tend toward the larger, with in-home offices for every member of the household becoming standard for those who can afford it. This of course drives houses to places where land is cheaper, the edge of the metropolitan sphere, or beyond, rather than the center, as the commute to the center, which may once have been daily, is now reduced to weekly. The challenge will be to make good suburbs and desirable small towns, where people can still engage in meaningful out-of-home activity, while accommodating their demands for larger structures.

Daily travel changes are already visible: though vehicle miles traveled in the US are largely back to normal (with more rural and less urban travel), public transport levels are not. It will be a while before, if ever, public transport returns to the pre-virus normal, even in places like Sydney which were not nearly as severely hit as China, the US, and Europe. Work-from-home, fear of exposure on public transport (and not just personal fear, official fear being promulgated by governments), and just a general economic downturn and unemployment are all factors to date.  

Substitutes like walking and biking (and especially the newly cost-effective e-biking) are likely to pick up some of the slack for those who continue to travel to work in the CBD, though more needs to be done to facilitate safe bicycling, in particular instituting a much larger network of separated and protected bike lanes. Even auto travel will change though, while total vehicle travel may remain stable or drop only a small amount compared with the Before Times, the nature of that travel differs so it is less peaked. This implies less demand for new road infrastructure, as the usage of roads is more balanced across the day. 

The After Times are post-post industrial. The industrial districts that were fashionably converted to urban office precincts will now get reconverted, perhaps to residential, which the market will always demand – people have to live somewhere, even if they can work anywhere.

If we will indeed interact primarily intermediated by the Internet, we will have finally moved to the next stage of human development, that of spaceless places. The flip side of spaceless places are placeless spaces. We will abandon spaces we no longer need. The CBD office building is the first target. New and expensive transport links to connect to the central city, or relieve peak congestion, will also be seen as white elephants.

Fig 4. Tract-to-Tract Commutes of 80km/50 miles or less in Minneapolis-St. Paul. http://dx.doi.org/10.1371/journal.pone.0166083.g004
Fig 4. Tract-to-Tract Commutes of 80km/50 miles or less in Minneapolis-St. Paul. http://dx.doi.org/10.1371/journal.pone.0166083.g004
Covid-19 virus. Source wikipedia
Covid-19 virus. Source wikipedia

Urban Infrastructure: Reflections For 2100

I have a chapter in

Urban Infrastructure: Reflections For 2100: An Edited Volume Imagining Infrastructure Transitions And Goals At End-Of-Century. Edited by Sybil Derrible and Mikhail Chester.

As you’d expect, my chapter is a bit on the snarkastic side. It begins something like this:

While digging for bitcoin in my copious Sydney backyard, I uncovered golden tablets revealing to me the near future history of transport in Sydney. …

Book Overview

Infrastructure systems deliver basic and critical services. They are the pillars of civilization. In the twenty-first century, infrastructure will need to change to fit the needs of a new world. What shape will they take? What function will they provide? Who will they serve and why? In this book, forty experts from around the world share their reflections for infrastructure at 2100. The book is a series of science fiction short stories, essays, and poems. Climate change, sustainability, resilience, and technology are recurring themes in the reflections. Written in 2020, it is impossible to predict how infrastructure will be in 2100. The goal of this book is not to make accurate descriptions of the future. Instead, it is to provide a dialogue and visions of what we could hope for or fear. Only time will tell on which side of the balance we end up leaning.

Moving Array Traffic Probes

Recently published:

  • Davis, Blake, Ji, Ang,  Liu, Bichen, and Levinson, D. (2020) Moving Array Traffic Probes. Frontiers in Future Transportation. doi: 10.3389/ffutr.2020.602356 [doi]

This paper explores the potential of moving array ‘probes’ to collect traffic data. This application simulates the prospect of mining environmental data on traffic conditions to present a cheap and potentially widespread source of traffic conditions. Based on three different simulations, we measure the magnitude and trends of probe error (comparing the probe’s `subjective’ or time-weighted perception with an `objective’ observer) in density, speed, and flow in order to validate the proposed model and compare the results with loop detectors. From these simulations, several conclusions were reached. A single probe’s error follows a double hump trend due to an interplay between the factors of traffic heterogeneity and shockwaves. Reduced visibility of the single probe does not proportionately increase the error. Multiple probes do not tend to increase accuracy significantly, which suggests that the data will be still useful even if probes are sparsely distributed. Finally, probes can measure the conditions of oncoming traffic more accurately than concurrent traffic. Further research is expected to consider more complex road networks and develop methods to improve the accuracy of moving array samples.

Keywords: Autonomous vehicles, Probes, Traffic state estimation, Floating car data, NetLogo

Transport Findings -> Findings

An announcement from FindingsPress:

October 19, 2020 AEST

You may have noticed that Transport Findings has become Findings. We believe the core idea of open access, peer-reviewed, short form research articles that is central to Findings has applications well beyond the transport domain, and we don’t want to limit ourselves (or you). We could have started a lot of small journals, but it is more cost effective, and probably also more beneficial, to keep everything under one journal name, with multiple sections and editors.

So everything we have published to date is in the section Transport Findings, as will undoubtedly be many future papers. But we are pleased to announce that we have opened up a new section Urban Findings, edited by Somwrita Sarkar, which will be launching soon. Urban Findings welcomes submissions following the Findings model of short, to-the-point research findings in the broad field of urbanism. You can see the Editorial Board here:

So at this time we are about Findings in the domains of Transport and Urbanism, because those are the practical limits of our current expertise, but we see no reason in principle that there should not be other sections.

If you have ideas about a topic area that you would both like to see articles for, and are willing to edit, please let us know. Editors of the new section would have to help recruit an editorial board, solicit articles, find reviewers, and, of course, make editorial decisions.

Unfortunately, we can only pay you in social capital, but those rewards are enormous, you will be helping assemble the knowledge of humanity, brick-by-brick, finding-by-finding.

Roderick Distinguished International Webinar: Jennifer Whyte from Imperial College London on `Infrastructure projects and digital delivery.’

The University of Sydney’s First Roderick Distinguished International Webinar is scheduled on Thursday, 19 November 2020, from 6-7pm via Zoom.

In this webinar, we will hear from Prof. Jennifer Whyte from Imperial College London. The talk will focus on Infrastructure projects and digital delivery.

Please CLICK HERE to register at your earliest convenience.

Speaker
Prof. Jennifer Whyte is a Professor in the Department of Civil and Environmental Engineering at Imperial College London, and holds the Royal Academy of Engineering and Laing O’Rourke Chair in Systems Integration. Her research is on the delivery of major infrastructure projects, and on the integration of systems, modular and digital delivery strategies. As a member of Construction Leadership Council in the UK, she has strong links to industry and policy, giving advice based on her research. She has been visiting faculty at Stanford and is Director of the Centre for Systems Engineering and Innovation at Imperial College London. She is the incoming Head of the School of Project Management and Director of the John Grill Institute at the University of Sydney.


Title: Infrastructure projects and digital delivery
The talk will focus on how is the delivery of major infrastructure projects transformed by pervasive use of digital technologies and digital information? This lecture will explore how the practices of delivering infrastructure are changing, arguing for the need for focus on systems integration and the realization of value from projects. It draws on research on London megaproject, Heathrow Terminals, London 2012 Olympics, Crossrail, Tideway, High Speed 2.

Transportist: October 2020

Welcome to the latest issue of The Transportist, especially to our new readers. As always you can follow along at the  transportist.org or on Twitter

The New Normal

While VMT in the US is back to normalpublic transport levels are not.

It will be a while before, if ever, that public transport returns to the pre-virus normal, even in places like Sydney which were not nearly as severely hit as China, the US, and Europe.

There are several reasons public transport demand will remain low, and these changes are perhaps permanent:

  • more people work from home at least a few days a week, especially CBD office workers who would otherwise be packed both onto trains and into hot-desked offices.  
  • people are instructed to avoid trains and buses to ensure distancing, which people who can will voluntarily do anyway.
  • unemployment rates are higher than previously.

Substitutes like walking and biking are likely to pick up some of the slack for those who work in the CBD, though more needs to be done to facilitate safe bicycling in and around Sydney (and most other English speaking cities), in particular following the lead of other global cities in instituting a much larger network of separated and protected bike lanes.

Research 

  • Cui, Mengying, and Levinson, D. (2020) Internal and External Costs of Motor Vehicle PollutionTransportation Research Record. [doi]On-road emissions, a dominant source of urban air pollution, damage human health. Emissions increase air pollution intake (and damage health) of travelers (internal costs), and of non-travelers (external costs). This research constructs a framework modeling the microscopic production of emission cost from the vehicle and link level and applies it to a metropolitan road network. It uses project-level Motor Vehicle Emission Simulator (MOVES) simulations to model link-specific on-road emissions, and then employs the RLINE dispersion model to estimate on- and off-road concentrations of pollutants from vehicles. The internal and external emission costs are measured accordingly by counting the health damage costs of travelers and gen- eral population because of exposure. The framework is applied to the Minneapolis-St. Paul (Twin Cities) Metropolitan Area as a proof-of-concept. The estimates show that highways have higher emission concentrations because of higher traffic flow, but that the internal and external emission costs per vehicle kilometer traveled are lower. The emission costs that commuters impose on others greatly exceeds that which they bear. This modeling process is replicable for planners and practitioners assessing emission costs in other regions.

Walk Sydney

  • After a year at the helm, in a peaceful and planned transition of power involving neither vote fraud nor Supreme Court intervention, I transitioned from being President to being an ordinary Committee member of WalkSydney this month. Good luck to our new President Barnaby Bennett.

Transport Findings

  1. Hassanvand, Mina. 2020. “Long-Distance Person Travel: A Cluster-Based Approach.” Findings, September.
  2. Roy, Avipsa, Daniel Fuller, Kevin Stanley, and Trisalyn Nelson. 2020. “Classifying Transport Mode from Global Positioning Systems and Accelerometer Data: A Machine Learning Approach.” Findings, September. https://doi.org/10.32866/001c.14520.
  3. Zimny-Schmitt, Daniel, and Joshua Sperling. 2020. “Quantifying Airport Employee Commuting and Related Energy Use: A Comparison of Six US Airports.” Findings, September. https://doi.org/10.32866/001c.16663.
  4. Fischer, Jaimy, Trisalyn Nelson, and Meghan Winters. 2020. “Comparing Spatial Associations of Commuting versus Recreational Ridership Captured by the Strava Fitness App.” Findings, September. https://doi.org/10.32866/001c.16710.
  5. Aldred, Rachel, and Anna Goodman. 2020. “Low Traffic Neighbourhoods, Car Use, and Active Travel: Evidence from the People and Places Survey of Outer London Active Travel Interventions.” Findings, September. https://doi.org/10.32866/001c.17128.
  6. Chen, Peng, and Jihao Deng. 2020. “Integrating Affordable Housing with Transit: Where Are the Transit Deserts?” Findings, September. https://doi.org/10.32866/001c.17244.

News & Opinion

Books

Internal and External Costs of Motor Vehicle Pollution

Recently published:

On-road emissions, a dominant source of urban air pollution, damage human health. Emissions increase air pollution intake (and damage health) of travelers (internal costs), and of non-travelers (external costs). This research constructs a framework modeling the microscopic production of emission cost from the vehicle and link level and applies it to a metropolitan road network. It uses project-level Motor Vehicle Emission Simulator (MOVES) simulations to model link-specific on-road emissions, and then employs the RLINE dispersion model to estimate on- and off-road concentrations of pollutants from vehicles. The internal and external emission costs are measured accordingly by counting the health damage costs of travelers and gen- eral population because of exposure. The framework is applied to the Minneapolis-St. Paul (Twin Cities) Metropolitan Area as a proof-of-concept. The estimates show that highways have higher emission concentrations because of higher traffic flow, but that the internal and external emission costs per vehicle kilometer traveled are lower. The emission costs that commuters impose on others greatly exceeds that which they bear. This modeling process is replicable for planners and practitioners assessing emission costs in other regions.

Figure 6. Emissions cost estimates (cents/vehicle kilometers traveled): (a) internal and (b) external.

Transportist: September 2020

Welcome to the latest issue of The Transportist, especially to our new readers. As always you can follow along at the  transportist.org or on Twitter

Plants vs. Animals

I posit, as a first order approximation, humans are going to bifurcate as a species into plants and animals. Plants work from home and have everything delivered. Animals have out of home jobs and travel.

A quick Twitter poll shows nearly 2/3 of my followers admit to being plants, disturbingly high.

Books

  • 30-Minute City: Designing for Access is now Open Access

I am pleased to announce that you can now download a PDF version of The 30-Minute City: Designing for Access from the University of Sydney eScholarship Repository. (Free)

The 30-Minute City by David M. Levinson
The 30-Minute City by David M. Levinson
VIEW/OPEN
DATE
  • 2019-12
AUTHOR
  • Levinson, David M.
METADATA

This book describes how to implement The 30-Minute City. The first part of the book explains accessibility. We next consider access through history (chapter 2). Access is the driving force behind how cities were built. Its use today is described when looking at access and the Greater Sydney Commission’s plan for Sydney. We then examine short-run fixes: things that can be done instantaneously, or nearly so, at low budget to restore access for people, which include retiming traffic signals (chapter 3) and deploying bike sharing (chapter 5) supported by protected bike lane networks (chapter 4), as well public transport timetables (chapter 6). We explore medium-run fixes that include implementing rapid bus networks (chapter 7) and configuring how people get to train stations by foot and on bus (chapter 8). We turn to longer-run fixes. These are as much policy changes as large investments, and include job/worker balance (chapter 10) and network restructuring (chapter 9) as well as urban restoration (chapter 11), suburban retrofit (chapter 12), and greenfield development (chapter 13). We conclude with thoughts about the ‘pointlessness’ of cities and how to restructure practice (chapter 14). The appendices provide detail on access measurement (Appendix A), the idea of accessibility loss (B), valuation (C), the rationale for the 30-minute threshold (D), and reliability (E). It concludes with what should we research (F).

URI

Research 

  • Ji, Ang and Levinson, D. (2020) An energy loss-based vehicular injury severity model. Accident Analysis and Prevention. 146 October 2020, 105730. [doi]

How crashes translate into physical injuries remains controversial. Previous studies recommended a predictor, Delta-V, to describe the crash consequences in terms of mass and impact speed of vehicles in crashes. This study adopts a new factor, energy loss-based vehicular injury severity (ELVIS), to explain the effects of the energy absorption of two vehicles in a collision. This calibrated variable, which is fitted with regression-based and machine learning models, is compared with the widely-used Delta-V predictor. A multivariate ordered logistic regression with multiple classes is then estimated. The results align with the observation that heavy vehicles are more likely to have inherent protection and rigid structures, especially in the side direction, and so suffer less impact.

This study focuses on path flow for road network, as the sum of individual route choices from individual travelers, associated with specific path type for each cost factor of auto travel that finds the optimal route with the minimum cumulative cost from the perspective of the corresponding cost analyst interest. The considered cost factors include time, safety, emission, and monetary costs, as well as their composite, internal and full cost of travel. We further explore the extent to which each cost factor explains the observed link traffic flows given an estimated home-to-work demand pattern. The results of the Minneapolis – St. Paul metropolitan area indicate that flows from multiple path types, associated with internal cost components, additionally to the factor of distance, provides the best fit.

Transportist Blog

Traffic Safety

On August 6, I was on ABC Radio Hobart at 09:35 AEST talking traffic safety https://abc.net.au/radio/hobart/live/… with Leon Compton @abchobart. I was later interviewed for ABC-TV on the same subject. I was talking about my blog post: 21 Solutions to Road Deaths. Apparently there has been a recent spike in Tasmania. My key talking points:

We use the term “crash” not “accident” as “accident” implies no one was at fault and lack of intention and crash is more neutral. Crashes have causes.

The safety community likes to say “Safety is a shared responsibility”, and the responsibility lies with drivers, road engineers, vehicle designers, public policy, and others. [Though I think it is often about shirking responsibility and putting it back on the victim rather than taking it themselves].

Every crash has individual causes, but there are trends.

Australia-wide, there were 484 total fatal crashes (January through June) – down compared to the same period in 2019, which saw 572. https://roadsafety.transport.nsw.gov.au/downloads/dynamic/nsw-road-toll-monthly.pdf

The Lockdowns associated with COVID-19 are a factor. A study from Ohio State Univeresity found that COVID, which decreased the amount of automobile (and all) travel saw an increase in Speeding.

Speeding is a known cause of crashes. Higher speeds have two major effects: 

  • 1. Speeding reduces the available time for drivers to react to events, increasing the likelihood of a collision. 
  • 2. Higher speed increases the severity of impact, increasing the likelihood of fatality.

Solutions include better driver training and testing, more rigorous enforcement, keeping intoxicated drivers off the road, better engineering of roads, lower speed limits (which are both enforced, and designed into the road).

Walk Sydney

Transport Findings

  1. Roy, Avipsa, Daniel Fuller, Kevin Stanley, and Trisalyn Nelson. 2020. “Classifying Transportation Mode from Global Positioning Systems and Accelerometer Data: A Machine Learning Approach.” Transport Findings, September. https://doi.org/10.32866/001c.14520.
  2. Young, Mischa, and Steven Farber. 2020. “Using Wait-Time Thresholds to Improve Mobility: The Case of UberWAV Services in Toronto.” Transport Findings, August. https://doi.org/10.32866/001c.14547.
  3. Du, Jianhe, and Hesham Rakha. 2020. “Preliminary Investigation of COVID-19 Impact on Transportation System Delay, Energy Consumption and Emission Levels.” Transport Findings, July. https://doi.org/10.32866/001c.14103.
  4. Branion-Calles, Michael, Kate Hosford, Meghan Winters, Lise Gauvin, and Daniel Fuller. 2020. “The Impact of Implementing Public Bicycle Share Programs on Bicycle Crashes.” Transport Findings, September. https://doi.org/10.32866/001c.16724.

News & Opinion

Books