Transit planners and shuttle riders in cities

AI maps risky spots for Detroit’s self-driving shuttles

Researchers mapped higher-risk spots on a shuttle route to help plan safer paths.
November 14, 2026
Peer reviewed stamp, level 1 of 5: checked by outside experts
The short version, TL;DR
What AI Achieved
Using 83-plus hours of Detroit shuttle-route data, researchers built AI safety maps of higher-risk spots to help plan routes. This is route analysis, not proof of fewer crashes. Read more.

A peer-reviewed conference paper describes how researchers used AI to map risky spots along Detroit’s self-driving shuttle route, so planners can think more carefully about where those vehicles travel. The work is a safety-map analysis of real street data, not a study that measured fewer crashes after the maps were used.

The City of Detroit launched an autonomous shuttle pilot in 2024, with support from a U.S. Department of Transportation demonstration program, to see whether self-driving shuttles could help residents who have trouble getting around. The American Center for Mobility and Michigan Technological University then studied the route itself.

With more than 83 hours of on-road data, the team built safety maps meant to point out localized high-risk zones and support better route planning.

That is a planning tool. It can highlight tricky places, such as areas with sudden maneuvers, work zones, or other hard cases. It does not, by itself, prove that the shuttle service has become safer.

What the researchers did

The authors include Vaishnavi Balambeed, Mojtaba Bahramgiri, Sushanta Das, Tony Geara, Jeffrey Naber, Shreeprasad Jadhav, and Peter Apostol. They collected data with fully instrumented vehicles on the Detroit shuttle route, then ran it through Driving Analytics, an AI platform developed by Michigan Tech and the American Center for Mobility.

The goal was to find edge cases and safety-critical situations, then turn those findings into maps of higher-risk zones. The authors say this approach can aid route planning and, in that way, help reduce risk. They also hope the work can inform policy by giving cities measurable ways to talk about readiness and remaining challenges.

The paper was presented at the ASCE International Conference on Transportation and Development in Detroit in 2026. The American Center for Mobility described the collaboration in an official announcement. Related work at the center has also used on-road findings to recreate hard cases, such as sudden pedestrian movement or blocked lanes, under controlled test conditions.

Why this matters for people

If you live along a future shuttle route, the everyday question is simple: are these vehicles being sent down streets where the hard cases cluster, or can planners steer around the worst spots, or at least test them more carefully first?

A map of risky intersections and street segments is a practical start. It can help a city choose a path, time a service, or decide what to practice on a test track before putting more shuttles in regular service.

Limits to keep in view

This research does not report a measured drop in crashes, injuries, or close calls after the maps were adopted. It analyzes hours of route data and proposes a method for spotting trouble. Whether those maps change real-world safety still has to be shown.

The dataset is more than 83 hours from one city’s shuttle corridor. Conditions in other cities, seasons, or vehicle types may look different. A winter storm, a stadium event, or a different shuttle design could change which spots rise to the top of the map, so the tool would need fresh data as street conditions change. And “risk reduction” here is a hoped-for result of better planning, not a before-and-after safety score.

Looking ahead for safer shuttle routes

For residents who may one day ride or share the road with an automated shuttle, careful mapping is a modest but useful step. It will not, on its own, make a busy city street safe. It can help people who plan those routes see the danger spots more clearly, and that is a start worth having.

You can read the American Center for Mobility announcement and the ASCE conference paper. For now, the honest claim is narrower, and that is a virtue. The team has shown a way to see risk more clearly. It has not yet shown that seeing the risk is the same as preventing harm.

If cities use maps like these and then measure what happens on the road, the next question will be whether those safer shuttle routes actually lead to fewer harmful events, not just better pictures of risk.

Like this? Get one every Friday.

The week’s real AI wins, in plain English, each with a stamp.

At a glance

Topic
Who it helps
Transit planners and shuttle riders in cities
Evidence level
Peer Reviewed
Confidence
55
out of 100
How soon
2 years
Original Source
American Center for Mobility
Published
July 9, 2026

AI Success Zone summarizes this work in plain English. The original source has the full details and is the final word.

More wins you may like