Recruiting students

As a new academic staff* of the Faculty of Engineering and Information Technology, I am recruiting graduate students.

A current list of my PhD and research master’s project opportunities can be seen here:

PhD and master’s project opportunities


An overview of the application process is outlined at

When lodging an application for admission, you are expected to have already discussed your research proposal with a potential supervisor (me), and will need a support letter.

You will need to draft a statement of research interest, this typically includes a proposed topic area, a review of research already in that area, specific research questions, and a brief idea of how the research will be conducted. I will review that with you, and if I am interested, may support your application. Even with the support of a potential supervisor, you will still need to meet the University of Sydney’s rigorous graduate admission criteria, and opportunities with financial support available are even more competitive.





* I am “academic staff” not “faculty”, the “faculty” is the “college”, a college is basically some hybrid between a “house” (in the Harvard system, without quite the percentages) and a “fraternity/sorority” in the land-grant system. Also the “school” is the “department.”


What if the University of Minnesota had 100,000 students?

This is a thought experiment.

According to the US Department of Education (reported in wikipedia), the University of Minnesota has (Fall 2013) 63,929 students, ranking 19th in the US. Of course, many of those larger schools are for-profit, non-residential, cater to part-time students, or college systems. Among “peers”, (land-grantish research schools) larger schools include Arizona State, Central Florida, Ohio State, and Maryland.

Alternatively, the Twin Cities campus is the 9th largest public university in the US by enrollment, with 48,308 students (undergraduate about 34,500), falling several places, and several thousand students from 2009, when it was 4th.

But US universities are pikers compared to some others in the world. The National Autonomous University of Mexico has over 324,000 students. The University of Paris has XIII parts, which collectively have about 330,000 students.

There are numerous economies of scale to be had from a larger school (i.e. costs rise sub-linearly with the number of students because the delivery of education has a large fixed cost component like buildings and chairs and lectures and software and administration), and the evidence about quality of education with size of school (and even to an extent size of classes) is mixed (though everyone seems to like smaller classes better). (The evidence about student satisfaction tends to favor small elite liberal arts schools with a lot of hand-holding, land grants like the University of Minnesota will never compete in that market). These economies of scale should lower per pupil costs. Rankings like US News favor small class sizes, but also favor large universities.

Clearly the University of Minnesota rejects more students than it accepts (it has a 44% acceptance rate). If it just one year raised that to 88% several things would happen. First there would be more students (though it is not clear enrollees would double, more than double, or less than double, since rejected students might be more or less likely). Second, because it lost some of its exclusivity, fewer top tier students would apply. Many faculty would similarly find alternatives – not wanting to deal with the headaches twice as many students brings, though I imagine teaching slots would be filled with the glut of PhDs on the market in most fields. Class sizes would undoubtedly increase, but more classes could be offered. [After World War II, Georgia Tech, which saw an influx of GIs, expanded its night school, so that there would be a full second shift of courses]. Yet the quality of the University would on average drop.

Trends working in the opposite direction of increasing campus size include MOOCs, encouraging more online, distance learning. Getting those to work well has been difficult, and they are yet to be mainstream. Moreover a seldom-stated purpose of university, keeping students out of the labor market and in a semi-protected environment between home and the harsh cruel world cannot be accommodated with MOOCs. Demographics also move against rapidly expanding university size overall. However any one university should be able to increase market share, both with domestic and international students. Another purpose of university, providing a mixing bowl for organization of individuals into groups and couples, requires in-person attendance.

However if this were done more gradually but intentionally, with an aim of attracting more and better students, it should be possible to grow the campus to 100,000 over a decade or two.

From a physical perspective, the campus has significant room to grow, especially to the North East, but also infill and intensification to the south-east, north, and on West Bank. Even more growth is possible on the St. Paul campus, which abuts the largely underutilized State Fair Grounds.

My guess is doubling enrollment would require far less then twice as much new classroom space, about a doubling of dorm and residential space, a bit less than doubling of graduate student offices and labs and faculty offices – less assuming many of the new teaching responsibilities would be borne by adjuncts rather than regular faculty.

Around campus, business would boom. New apartments, with ground floor retail and restaurants would fill in any of the empty parcels of Stadium Village and Dinkytown and West Bank.

Transit ridership to campus would more than double, since the capacity for cars on campus would decrease significantly (all those new buildings will use up surface parking lots, reason itself to support an expansion, pedagogy be damned) while total travel demand in terms of trips to campus rose.

Such a change would challenge Boston’s 250,000 students (not sure how this was calculated, must be “Greater Boston”) as the lead college town in the US. According to city-data, Boston’s college population is 14.6% of the total. The City of Minneapolis has 11.3%, near the top for large cities. Increasing enrollment by 50k students would increase the share of the City’s population in college by up to 12% (depending on where they lived, the maximum assumes they all live in Minneapolis proper), putting Minneapolis at about 23%. Clearly much of Boston’s attractiveness to college students is the surfeit of students already there. We don’t really operate in that league here yet.

If we think of college as a temporal port, just as immigrants from Europe used to land mostly in New York, and occasionally disperse from there, immigrants from youth land in college, and disperse slowly from that point. This should greatly increase innovation, as the region’s engine of growth scaled up and the region attracted boatloads of Generation Z/Post-Millennials to its lake shores.

MARC – Multi-Agent Route Choice Game

Our  Multi-Agent Route Choice (MARC) game is designed to engage students in the process of making route choice, so that they can visualize how traffic gradually reaches a user equilibrium (UE). In addition, the Braess’ paradox phenomenon, a concept not generally taught by undergraduate transportation courses, is embedded into the game so that students can explore this phenomenon through game-play.

The software, developed by Xuan Di, is now available for download.

The paper evaluating the application is now available:

School Miles vs. School Choice

The urbanist community has a nit about neighborhood schools. At one level. If all schools were interchangeable (like we imagine fire stations to be), people should use the closest one (just as you want the nearest fire truck when there is a fire). This is a “simple” covering problem in operations research, where you try to locate a set  of facilities (say schools) to serve some number of people (say on average 500 students)  at the lowest possible transportation cost, perhaps subject to some maximum transportation cost (no school is more than 12 minutes away).

Once upon a time (about the same time as all the model railroads in the world are set, that is, c. 1950, at the cross-over between Steam and Diesel so you can use both trains on the same layout), schools may have been interchangeable, since people were obviously undifferentiated.

I went to elementary school in the planned community of Columbia, Maryland, [in a generally well-off, well-educated suburban county with far more racial and income diversity than suburban Minnesota] where the elementary school was designed as the centerpiece of the Neighborhood, and the middle and high school were the center of the Village. Since the land use was planned along with the schools, it was probably as close to optimal at the time as any place in America.

I am old, so this was before the era of magnet and charter schools.  Most of the time I could walk home from elementary school (for a few years I was basically across the street). If I remembered my childhood fondly as an elysiatic paradise, (sadly for a variety of reasons, I don’t), I might want to impose that on future generations. Even then, there was school choice. Students could attend any public school which was not over-crowded. I attended an out-of-Village Middle School (which was in fact closer to my house).

It turns out however, that demographics change. Neighborhoods with lots of 5 year olds in 1972 have many fewer today. The best location for a school in 1967 is not the best location in 2017.

It also turns out that economies of scale change. The ideal size of school in 1967 according to 1967 standards is not the same as today. Schools are typically larger to provide more services, more diversity, and so on. [This has probably reached the point of negative returns, and is to the detriment of educational quality. Minneapolis’s Hans Christian Anderson Open School has 1000 students.]

Further, it turns out that many urban parents are tired of the poor quality of urban schools, so many systems, including Minnesota have moved from a “single provider” model to a “single payer” model. This is the core of the Charter School movement, and is definitely popular among those who send their offspring to Charter Schools, if not every teacher’s union (for the record, my mom was an NEA member, now retired) or the establishment Department of Education, for the same reason entrenched interests always oppose change. Magnet schools are another response, within the existing public school system.

Both charters and magnets increase the transportation miles of children traveling to school. This is an expected outcome, and produces the usual side-effects.

Travel to these choice school may be undertaken with buses or parents driving (or biking for some older kids or nearby kids, or walking for really nearby kids, as there will always be some). But it will certainly be more motorized than students who are captive to neighborhood schools.

Charters and magnets also are designed to increase the quality of educational outcomes.  If that has on average occurred — I believe it has (See: The Unappreciated Success Of Charter Schools), as do other parents who send their children there (otherwise they wouldn’t) — then we are trading off quality of education for transportation costs.

We trade-off all the time.

  • We don’t require you take the job nearest your house, though that would reduce distance traveled.
  • We don’t insist you buy groceries from the nearest supermarket, though this would reduce distance traveled.
  • We don’t ask you to go the nearest college, and in fact encourage you to travel to see more of the world.


Why do we believe elementary school education (which by programming minds is far more important than grocery shopping location which is merely filling stomachs with different sources of calories) is somehow completely substitutable?

Why do we denigrate the professionals providing education by asserting they are interchangeable parts?

Why do we diminish children by asserting they are indistinguishable and whose needs can be best met only at the closest school?

The whole argument is basically dehumanizing those both learning and teaching for the sake of nostalgia, congestion (which “urbanists” like, right? Isn’t congestion a measure of vitality.), theories about public health (the evidence relating built environment and physical activity is weak at best), and pollution (will this argument actually change if we used electric vehicles powered by renewables?)


If local schools are best for your child, you should send her there. Maybe she can walk or ride a bike. Fantastic, we all win.

If another school is best, you should send your child to that school. Maybe she can take a bus. She we will get the best education possible, maybe go to an Environmental Magnet school, and hopefully learn other ways to reduce the negative effects of human life on earth.

Don’t assume what is best for your child is best for all children and their families, or what is best for the environment in the short run is best for the children, and don’t be sanctimonious.

Video: Timing Traffic Signals: Example

(from the Introduction to Transportation Engineering Series)

Video: Traffic Analysis at Signalized Intersections: Example

(from the Introduction to Transportation Engineering Series)

Video: Traffic Analysis at Signalized Intersections

(from the Introduction to Transportation Engineering Series)

Video: In Search for Lost Time

(from the Introduction to Transportation Engineering Series)

Mini Metro Review

Screen Shot of MiniMetro
Screen Shot of MiniMetro

I have spent too much time in the last month playing the highly addictive Mini Metro, by Dino Polo Club.

The game pitches itself as managing the growth of a metro (rail) system, but given the pliability of networks, it is probably better to think of it as a bus network, since lines can easily be moved and reconfigured, as well as extended.

The game has a number of attributes:


Nodes of activity (larger white shapes, outlined in black) appear pseudo-randomly over time. Development occurs randomly with some contiguity, new nodes are likely to appear near existing  existing nodes and lines, allowing a line extension or diversion to serve them. .

Notably, they get demand even without being added to the network (so you better add them). Sometimes they show up on the network, sometimes they appear to be on the network, but are bypassed by the network, which is really annoying, since the queue for this is subtle and not obvious on a busy screen.


Each node both produces and attracts trips. It attracts trips in the shape of the node, and produces demands of every shape but its own (internal trips can presumably satisfied without using the network). These demands are the little black shapes next to the node, which must get delivered to nodes of the same shape.

There are different types of demand (shapes). I like to think of them as squares representing downtown/employment, circles as residences, triangles as retail, and a bunch of one-time special generators (plus – hospital, wedge (intercity train station), star (airport), pentagon (stadium)). Circles are most common.

Demand grows in an inflationary way to ensure you lose in the end. I.e. you cannot respond fast enough to changing markets (there is no pause while I rework my network) button. I can’t figure out the exact formula for this, except it is too fast. It seems it is a function of accessibility or connectivity, though I am not clear how this is measured.


Each Sunday (after a week) you get a new locomotive. Sometimes you can build a new line, and connect more points, sometimes tunnels, sometimes carriages (2 of 3). It appears you get 2 of 3, but I can’t determine how you get one rather than the other. There are a maximum number of locomotives (4 per line), and lines in the system. Locomotives can have additional carriages.


You need to balance node shapes along the route, to minimize transfers. So your route should contain as many different shapes as possible, intermixed as much as possible. A chain of circle nodes is not helpful since circles don’t generate circle demands.

You need to balance the number of lines vs fewer (longer) lines with more locomotives and carriages. Long lines with few locomotives have long headways, and thus more crowding. I am not sure the extent this feedsback and dampens demand. I have lately taken the strategy of maximizing capacity on one line before opening the next.

Locomotives add frequency, carriages add capacity. Locomotives are more valuable, but carriages are a second best.

Tunnels cost money. Since all the game-boards have rivers, some tunnels are necessary, but if you choose the tunnel, you forego either the line or the carriage, so choose wisely.

Long lines with limited locomotives have longer headways between vehicles. In other words, frequency is endogenous.

No obvious architecture of system works best as far as I can tell (Grid, radial, U-shaped lines) though I am favoring circle lines now, with two locomotives going clockwise and two counter-clockwise. The key question seems to be how interconnected you make the lines, how many lines intersect with each other, and how to minimize transfers, especially at crowded platforms.

Extending lines is free by adding links, except that it adds to headways without also adding locomotives. Lines longer than 9 long are trouble, especially without additional locomotives.

Bus bunching occurs, and is endogenous. It appears vehicles can overtake, I am not clear on that.

I try to make lines connect on non-circles and non-triangles, since those are more special places, likely to have their own special demands, and I want to minimize transfers. However, geometry won’t always allow that since demands keep popping up which need to be served.

There seems to be some sort of reasonably intelligent transit passenger routing algorithm, determining which passengers take which trains (transfers vs. direct). They don’t just jump onto the first train.

Design of the network is very much like the Traveling Salesman Problem.


The game ends when one of your stations suffers over-crowding. In short, it’s too busy, we shouldn’t provide any service. Perhaps it should not be the end of the system, just the end of the manager, who gets fired. My high score on the Steam version is 2006: 6666 passengers over 322 days in Sao Paulo.


The game is still in Beta, and being actively developed, so wishes are useful.

  • I wish could easily uncouple trains from carriages and rebalance vehicles within and across lines without deleting whole line. (The newest version (beta) promises this.
  • The game does not work on iOS. I wish it did.
  • I wish in the game I could pause while I reworked my network, Changing the network seems a dangerous time, especially as the network gets crowded.
  • I wish game statistics worked better.



The game is very aesthetically appealing, especially in this new post iOS7/MacOS 10.10 (Yosemite) era, and the beautiful interface fits right in.  It is impressive what they can do with an interactive web-game, though it requires a plug-in.

A version of the game is free on their website, and a somewhat more advanced game is available on the Steam Platform for $6.99.

This version includes Multiple cities: London, NYC, Sao Paulo, Saint Petersburg, Paris,  and Hong Kong (as of yesterday) with more coming.

It has Leaderboards, so you can compare your pathetic score with others. The Leader has over an amazing 8000 points.

Importantly there is both the Commuter mode (with crowding) and a Scenic mode,  (with Free play, and without a crowding penalty.)

This is one of the most abstractly realistic, playable transportation games out there. There are more complex games to be sure, but none which seem to capture so much of the fundamental essence.

As a fan of games and simulations for education, I am making playing this game a lab for my Intro to Transportation Engineering class.