Tech

NYC's AI-Powered Smart Crosswalk Pilot Rolls Out Across Midtown, Cutting Pedestrian Accidents by 34 Percent

The system, installed at 47 intersections between 34th and 59th Streets, uses computer vision to detect pedestrians, cyclists, and vehicles in real time and adjust signal timing to reduce conflict points.

By Keisha Brown · July 1, 2026 · 4 min read

NYC's AI-Powered Smart Crosswalk Pilot Rolls Out Across Midtown, Cutting Pedestrian Accidents by 34 Percent

NEW YORK -- The New York City Department of Transportation announced on Tuesday that its AI-powered smart crosswalk pilot program, which has been operating quietly at 47 intersections in Midtown Manhattan for the past five months, has reduced pedestrian-involved accidents at those locations by 34 percent compared with the same period last year -- a result the department called statistically significant and sufficient to justify a major expansion of the technology. The system uses computer vision cameras mounted on existing signal infrastructure to detect pedestrians, cyclists, delivery vehicles, and passenger cars in real time and adjust signal phase timing to reduce the risk of conflicts.

The technology, developed by a startup called Waypoint Signal in partnership with the DOT under a $22 million pilot contract, works by extending the pedestrian walk phase when the system detects a large crowd waiting to cross, shortening green phases when vehicle traffic is light, and triggering an automatic warning display when a cyclist or pedestrian enters an intersection against the signal. The cameras do not capture or store facial imagery and are not connected to any law enforcement database, according to the DOT -- a clarification the department made proactively in response to privacy concerns raised when the pilot was announced.

Transportation Commissioner Ydanis Rodriguez said at a press conference that the pilot results exceeded the department's projections. 'We hoped to see a 15 to 20 percent reduction in accidents,' Rodriguez said. 'Thirty-four percent is remarkable, and it is a direct result of the system making thousands of small adjustments per day that a static signal timing plan simply cannot make.' The most dramatic improvements came at intersections with high volumes of delivery truck activity, where the system's ability to detect trucks blocking sight lines and extend crossing times showed the greatest divergence from the baseline.

The DOT plans to expand the program to an additional 200 intersections citywide by the end of 2026, with priority given to locations that appear in the Vision Zero high-injury network -- a map of roughly 6 percent of street miles that account for more than 60 percent of serious and fatal pedestrian injuries. Community board responses to the expansion plan have been largely supportive, though several boards in lower-density outer borough neighborhoods asked the department to ensure that the expanded system addresses the specific risks posed by wide arterial roads with high-speed traffic, which differ significantly from the dense midblock conditions of the Midtown pilot.