From the Transportation edition of November 5, 2026
UMTRI SAFETI wins $5.66M U.S. DOT award for AI intersection crash-prevention with MDOT and Ann Arbor
University of Michigan Transportation Research Institute (UMTRI) with Michigan DOT, City of Ann Arbor and SEMCOG · Roadway safety / traffic signals · Ann Arbor, Michigan, USA
- Publisher
- University of Michigan Civil and Environmental Engineering
- Original publication
- 2026-11-04
- Source retrieved
- 2026-11-06
What happened
UMTRI received $5.66 million from U.S. DOT for SAFETI (Safety Actions for Everyone at Traffic Intersections), advancing open-source roadside perception (MSight) into near-miss prediction that can warn drivers and pedestrians via dashboards, in-pavement lights, phones or signal changes. U.S. DOT attributes roughly one quarter of traffic fatalities and about half of injuries to intersections. Phase 1 (~18 months) prototypes at Mcity with GM, Volkswagen and Ford; Phase 2 deploys at four Ann Arbor intersections over a 3.5-year program. Privacy design: Ann Arbor fisheye cameras cannot resolve plates or faces; the algorithm classifies object type only. Public partners include MDOT, Ann Arbor and SEMCOG.
Why it matters
A U.S. DOT-funded, DOT/city-partnered path from lab perception to signalized intersection interventions with explicit privacy design—usable as a reference architecture for state DOT safety programs.
Evidence and measured results
University announcement of award and planned methods; no field crash-reduction results yet.
Limitations and uncertainty
Pre-deployment research; Phase 2 limited to four intersections; OEM partners may shape vehicle integration; crash outcomes unknown.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-05; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Intersections account for a large share of U.S. traffic deaths and injuries.
- Stakeholders
- State DOT safety engineer
- City traffic engineer
- MPO
- Law enforcement / Vision Zero
- Discovery questions
- Which intersections drive fatal and serious injury crashes?
- What camera and signal assets exist there?
- What privacy policy governs traffic video?
- Value hypothesis
- Near-miss prediction with multi-channel warnings may reduce conflicts if humans respond; unproven in the field here.
- Potential engagement
- Site selection and evaluation design for a bounded intersection safety AI pilot.
- Unsupported claims
- Crash or fatality reductions
- Nationwide readiness
- Driver/pedestrian compliance with alerts
Pre-sales engineering
Role takeaway
- Fit
- Fits cities/DOTs with instrumented intersections and signal control access.
- Architecture and integration
- Roadside perception to alerts and optional signal timing changes via V2X and controller interfaces.
- Prerequisites
- Camera/LiDAR coverage
- Signal controller integration path
- Privacy-compatible imaging
- Constraints
- 3.5-year research timeline
- OEM vehicle integration dependencies
- Four-intersection Phase 2 scope
- Security
- Prevent re-identification from video; harden roadside compute and V2X links.
- Proof of value
- Conflict prediction precision/recall at Mcity
- Human recognition of warnings
- False-alert rate in field
Delivery
Role takeaway
- Work
- Prototype, human-factors tests with robot VRUs, then four-intersection deployment.
- Dependencies
- U.S. DOT funding
- MDOT/Ann Arbor/SEMCOG
- GM/VW/Ford/Ouster
- Ownership
- UMTRI research lead; public agencies own field operations.
- Skills and adoption
- Operator training on alert dashboards and signal interventions.
- Governance checkpoints
- Phase 1 human-factors gate
- Privacy review before Phase 2
- Open-source release readiness
- Acceptance criteria
- Proposed: near-miss detection metrics met
- Proposed: warning comprehension thresholds
- Proposed: no plate/face leakage in audits
- Risks
- Alert fatigue
- Controller integration delays
- Public trust if mis-alerts occur
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Roadside cameras/LiDAR plus V2X, signal control hooks, smartphone and in-vehicle alerts on an open-source perception stack, with optional green extension for predicted red-light runners.
Governance
Who approves, reviews and stays accountable for outcomes?
Multi-year phased evaluation including human-factors testing before public deployment; open-source release planned.
Security and privacy
What data, permissions and controls need testing?
Fisheye and blurring choices aim to prevent plate/face identification; object-class-only inference stated.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Pedestrian and cyclist warnings are central; Phase 1 studies whether people understand interventions.
Procurement
What should contracts, pricing and exit terms secure?
Federal research award with university and OEM/public partners; not a commercial RFP award.
Operating model
Which teams own the service once it runs?
Research-to-operations path via MDOT and city partners at real intersections.
Publication history
- 2026-11-05Transportation · Issue 084 resources
Stable resource ID: umtri-safeti-usdot-intersection-crash-prevention-2026-11-04