USDOT Selects UCLA, University Of Michigan And Derq For $5.5M Intersection Safety Phase One

The U.S. Department of Transportation (USDOT) has selected three teams to build prototypes that predict and mitigate conflicts between vehicles and pedestrians at intersections. The teams are led by the University of California, Los Angeles (UCLA), the University of Michigan Transportation Research Institute (UMTRI) and the intelligent transportation company Derq USA. The department has awarded $5.5 million across the three teams under Phase 1 of its Intersection Safety Systems (ISS) programme and expects to add a further $11.5 million for Phase 2 at operational intersections, according to the Federal Highway Administration solicitation and Derq’s announcement of 22 October 2026. The $17 million programme responds to an estimated 36,640 U.S. road deaths in 2025.

Three Of The Four Stage 1B Tier 1 Winners Move From Algorithms To Hardware

The ISS programme follows the Intersection Safety Challenge, a prize competition launched on 25 April 2023 with up to $6 million in prizes, and moves that work from software to physical prototypes. USDOT received 120 concept papers and selected 15 for Stage 1A awards, according to the department’s January 2024 announcement.

In Stage 1B, announced in January 2025, 10 teams shared $4 million after training sensor fusion and conflict prediction algorithms on data recorded on a closed course at the FHWA Turner-Fairbank Highway Research Center. The four Tier 1 winners, each paid $750,000, were Derq, the UCLA Mobility Lab, the University of Hawaii and the University of Michigan, as reported by ITS International.

Three of those four Tier 1 teams now hold the Phase 1 contracts, which leaves the University of Hawaii as the only top-tier Stage 1B winner outside the prototyping round. USDOT competed the round under solicitation 693JJ3-26-BAA-0004, which was posted on 5 June 2026 and closed to offers on 20 July 2026.

Camera-First, Multi-Sensor And LiDAR-First Designs Tested In Parallel

According to team descriptions published by the ITS Joint Program Office, the three prototypes differ most in their primary sensor and in how much control each takes over traffic signals. That gives USDOT a comparison of different architectures rather than three versions of one design.

UCLA’s InfraShield pairs an offline process for designing each deployment with an online edge-computing pipeline that accepts cameras, LiDAR, radar and optional thermal or infrared sensing. Mitigation is introduced in stages, from Observe and Diagnose through Warn and Advise to Control or Operational Support. The last stage is limited to agency-approved, controller-safe responses such as an all-red extension.

UMTRI’s SAFETI, short for Safety Actions for Everyone at Traffic Intersections, uses LiDAR as its primary sensor while accepting existing roadside cameras. It splits processing between an edge layer for real-time functions and a cloud layer for network-wide conflict analysis, and it delivers alerts through signal control, V2X messages and 5G notifications to smartphones and in-vehicle dashboards.

Derq USA’s system treats video cameras as the primary sensor and fuses them at the roadside edge with LiDAR or radar, signal controller events and vehicle-to-everything (V2X) messages. It classifies intent to anticipate red-light running and pedestrians about to enter the roadway, then selects a warning, dynamic signage or a change to intersection control. “Most systems today tell you what happened. Ours also predicts what’s about to happen and acts on it,” said Georges Aoude, co-founder and chief executive of Derq, in the company’s October 2026 press release.

Four Proving Grounds And Ten Public Agencies Behind The Teams

All three teams will test first on closed facilities, where conflicts can be staged without putting the public at risk. Work at live intersections is reserved for Phase 2.

UCLA will test InfraShield at GoMentum Station in Concord, California, with the Contra Costa Transportation Authority (CCTA) and the California Department of Transportation (Caltrans). CCTA is already running a one-year connected mobility demonstration with Monotch on Treat Boulevard in Concord.

UMTRI will use the Mcity Test Facility in Ann Arbor with the Michigan Department of Transportation, the Southeast Michigan Council of Governments and the City of Ann Arbor. It has the longest list of industry partners of the three teams: General Motors, Ford, Volkswagen, Shulman Technology, Ouster, Verizon, Amazon Web Services and NVIDIA. Ouster also supplies the sensors behind the Georgia Department of Transportation’s LiDAR network across more than 30 Atlanta intersections.

Derq will run its prototype at the Virginia Tech Transportation Institute’s Virginia Smart Roads and the Texas A&M Transportation Institute’s RELLIS Proving Grounds. Its public partners are the Texas and Virginia departments of transportation, the City of Houston and the Virginia cities of Alexandria and Falls Church, and its industry partners are WSP USA, the Intelligent Transportation Society of America and AT&T. Houston already deploys Derq’s AI traffic cameras at city intersections.

The Research Questions USDOT Has Written Into Phase 1

USDOT wants to know how far in advance conflicts between vehicles and vulnerable road users can be predicted, and under which conditions those predictions remain reliable. It also asks how quickly recurring unsafe conditions can be identified at a given intersection.

On mitigation, the department asks whether intersection control can be changed in real time. Its examples are removing permissive turns across busy crosswalks and holding a WALK signal until approaching vehicles have cleared. It also wants to learn which warning types work best and what each strategy does to safety and congestion.

The solicitation adds a commercial test, stating that the effort should consider the cost-benefit of broader ISS deployment. Each team will therefore have to price its system per intersection rather than only demonstrate detection accuracy.

A $17 Million Programme Against A National Pedestrian Toll

The solicitation cites National Highway Traffic Safety Administration data showing that, on average in 2024, a pedestrian was killed every 74 minutes and injured every 7 minutes. About 18% of pedestrian fatalities that year occurred at intersections.

At $17 million across both phases, the ISS programme is 2.8 times the $6 million prize ceiling of the original Challenge. The expected Phase 2 budget of $11.5 million is 2.1 times the Phase 1 figure, which is consistent with USDOT expecting integration with live signal controllers to cost more than algorithm development.

What Phase 2 Will Have To Prove

Phase 2 funding is expected but not yet awarded, and the published material does not state how many of the three teams will advance. Nor does it say which interventions agencies will permit at public sites, where any automated change must respect the minimum clearance intervals and conflict monitoring built into existing controllers.

Whether these prototypes reach city streets will most likely depend on the per-intersection cost the solicitation asks teams to weigh. An agency running hundreds of signals will compare that cost with a leading pedestrian interval, which needs little more than a timing change in a controller the agency already owns.