This Week In Public Transit

Transcript 30 lines

Cold Open Stats Overview Paper Walkthrough free_promo

Cold Open

Jenny Why does a trip feel easy in one neighborhood and exhausting in another, even when the place you need is technically nearby?

Davis Because nearby on a map can still mean two transfers, a dark walk, and a bus that only behaves when you don't need it.

Jenny Right, and I keep wanting to separate closeness from reachability, because a clinic, a job, or a community center doesn't help much if getting there burns the whole afternoon.

Davis So the phrase is access is actually used, not just promised.

Jenny And that gets very real in Mexico City, where who reached the Utopías community centers depended not just on distance, but on transit links, what each center offered, and whether the walk there felt safe ... welcome to This Week In Public Transit on paperboy.fm.

Stats Overview

Jenny This week the search found 13 transit papers, and 9 qualified, meaning they passed the relevance screen. Those 9 papers involve 34 authors across 6 countries, so it's a smaller stack, but still geographically spread out.

Davis The qualified count fell from 11 to 9, down about 18 percent, and the methods help explain the feel of it. We have 2 qualitative studies, 2 case studies, and 2 simulations, so this isn't one big wave of bus-tech papers; it's a mixed week about how systems actually work.

Jenny Here's the odd bit: query hits rose from 12 to 13, up about 8 percent, while the qualified set shrank. That says the net was a little wider, but more papers got screened out, and I'd want to know whether the misses were adjacent topics like mobility tech without a real public transit core.

Davis Country coverage stayed flat at 6 countries, with the USA showing up twice and Macao, North Macedonia, Mexico, Canada, and Korea once each. But unique authors fell from 39 to 34, down about 13 percent, so the week is not less international; it's a little less crowded.

Jenny The author mix is mostly established. Out of 34 authors, 20 are experienced, 8 are emerging, and 6 are first-time authors, meaning first-ever paper in the metadata, not just new to this show.

Davis Theme-wise, public transit leads with 2 papers, then single papers on ridership, financial sustainability, emerging technologies, and service planning. That fits the through-line: no magic tool fixes transit right now; agencies are trying to match reliability, access, equity, and capacity to how people actually move.

Paper Walkthrough

Paper 1 The Future of Transit

Davis Alright, let's get into the papers with The Future of Transit, a twenty-twenty-six perspective paper by Adam P. Cohen that treats post-pandemic transit as a management problem, not a gadget problem.

Davis The practical claim is that agencies already know the four big levers: reliability, travel time, frequency, and customer experience. Cohen uses the Santa Clara Valley Transportation Authority as a case example, and the point is that old peak-commute planning doesn't fit as well when hybrid work and scattered trips are now normal.

Jenny So is this a research finding, or is it more like a well-informed strategy memo for agencies?

Davis More the second, but a useful one. The paper synthesizes peer-reviewed studies, government and industry reports, public data, agency case studies, and expert interviews, then turns that into a portfolio: improve core service first, then test simpler fares, better transfers, late-night service, demand-responsive service, which means flexible service built around requests, and land-use moves like transit-oriented development.

Davis The limit is important, though. This isn't one causal test showing that, say, late-night buses raise ridership by a specific percent; it's a moderate-strength synthesis built from cases and expert judgment.

Jenny That actually feels honest to me, because the thread here is reliability under constraints. If you're an agency with budget pressure, workforce shortages, and riders who don't all travel at eight in the morning anymore, the takeaway is not chase the shiniest pilot; it's decide what reliability means, measure it, and only then ask which fare, partnership, or real-estate move earns its keep.

Paper 2 Real-time bus arrival prediction and bus bunching prevention using a multi-agent LSTM approach

Jenny That word measure is the bridge here, because the next paper asks what happens when reliability gets measured minute by minute, not in a quarterly dashboard. It's called Real-time bus arrival prediction and bus bunching prevention using a multi-agent LSTM approach, and it's by S. Rawat, Alyona Abaidullina, Dilnaz Alimbayeva, Deepak Kumar, and Praveen Kumar in Frontiers in Built Environment.

Jenny Plain version: they built a better bus-arrival predictor, then plugged it into simulated buses that could coordinate before the line collapsed into bunching. The data came from New York City's MTA SIRI feed, which is the real-time service data agencies use for vehicle locations and arrival estimates, covering the twenty highest-frequency bus routes.

Jenny The model was an LSTM, short for long short-term memory, which is a neural network built to learn time patterns in sequences. It got mean absolute error down to two point zero two minutes, which means the average miss was about two minutes, and that was a twenty-eight point five percent improvement over historical baseline models.

Davis How much confidence should we put in a simulation where buses make decisions without a central dispatcher? That's the part that sounds useful, but also a little like giving every bus its own tiny operations manager.

Jenny Fair question. They used MTA real-time data, filled sparse gaps with a forward-looking imputation method that kept more than ninety percent of the original telemetry, trained the LSTM across routes, and then put the predictions inside autonomous bus agents, meaning simulated buses that communicate locally with nearby buses instead of waiting for a central command.

Jenny In the one hundred twenty-step simulations, those agents could trigger decentralized stop-skipping, which they call EXPRESS mode, when spacing got bad. EXPRESS mode activated in nineteen point two percent of steps on the B82 and twenty-five point eight percent on the Bx9, but the strongest evidence is still specific New York routes inside simulation, so live operations would be the real test.

Davis The practical takeaway is bigger than the gadget. If an agency already has real-time bus data, this says prediction can move from passenger information to proactive headway control, which fits that reliability-under-constraints thread: don't just tell riders the bus is late; use the next two minutes to stop three buses from arriving together.

Paper 3 Evaluating the reach and equity of public spaces in Mexico City using mobility data: the case of the Utopías in Iztapalapa

Davis That last paper was about using two minutes of bus prediction to keep service from bunching, and this one asks what happens after the bus can get you somewhere. It’s called “Evaluating the reach and equity of public spaces in Mexico City using mobility data,” a twenty twenty-six Frontiers in Big Data paper on thirteen Utopías community centers in Iztapalapa, an underserved borough of one point eight million people in eastern Mexico City.

Davis The plain version is that free community centers don’t automatically deliver equity just because they exist nearby. The authors find that visitor patterns depend on transit connectivity, on whether a site has something distinctive like a planetarium or dinosaur park, and on safety, because a higher violent-crime share nearby is associated with more local and lower-income visitor catchments, meaning the area people actually come from shrinks.

Jenny Do location traces tell us who is benefiting, or only who managed to show up? Because if the data come from phones, I want to know whether we’re measuring access for everyone, or access for people visible to apps.

Davis That’s the right pressure point, and they’re pretty careful about realized access, meaning actual visits rather than a map circle around a facility. They combine six months of location-based mobility traces, basically anonymized movement data from devices, with GTFS transit data, which is the standard public transit schedule format, plus a georeferenced crime registry and a catalog of each center’s activities; the evidence is rich, but it’s still one borough and one network of community centers, so I wouldn’t export the exact pattern everywhere.

Jenny That lands for me because it makes the access-is-actually-used thread very literal. If a city builds the pool, the music room, or the planetarium, the equity question is still whether the trip is connected, safe enough, and worth making, not whether someone can draw a fifteen-minute buffer around it.

This Week In Public Transit

Transit’s Next Tests Are Practical

Episode

Cold Open

Stats Overview

Paper Walkthrough

Paper 1 The Future of Transit

Paper 2 Real-time bus arrival prediction and bus bunching prevention using a multi-agent LSTM approach

Paper 3 Evaluating the reach and equity of public spaces in Mexico City using mobility data: the case of the Utopías in Iztapalapa

free_promo

Other Episodes

This Week In Public Transit

Transit’s Next Tests Are Practical

Episode

Cold Open

Stats Overview

Paper Walkthrough

Paper 1 The Future of Transit

Paper 2 Real-time bus arrival prediction and bus bunching prevention using a multi-agent LSTM approach

Paper 3 Evaluating the reach and equity of public spaces in Mexico City using mobility data: the case of the Utopías in Iztapalapa

free_promo

Other Episodes

Get This Week In Public Transit by email