A research team at the University of Arizona has developed a radar sensor that can reliably detect pedestrians and cyclists at intersections, including those partially hidden behind larger vehicles. The technology addresses a critical blind spot in current traffic monitoring systems. Over half of all fatal crashes in the United States occur at or near intersections, and two-thirds of pedestrian deaths happen at crossings—precisely where existing sensors often fail to see vulnerable road users.
At a Glance
- MmWave radar classifies road users as pedestrian, vehicle, or environmental clutter
- Performs reliably in rain, fog, darkness, and smoke
- Successfully tested at a real intersection in Tucson, Arizona
- Designed to complement existing camera and LiDAR systems
- Compact and inexpensive enough for widespread city deployment
- Researchers are expanding the system to identify additional transportation modes
The Problem With Current Traffic Sensors
Today's intersection monitoring relies on a mix of technologies, each with serious gaps. Cameras excel at visual classification but degrade sharply in heavy rain, fog, or when headlights create glare. They also struggle with smoke from nearby fires or simply with the deep shadows that appear during winter evenings. Conventional radar tracks moving objects but cannot reliably distinguish a pedestrian from a roadside tree swaying in the wind. LiDAR creates detailed three-dimensional maps but has trouble measuring object speeds accurately—yet speed is precisely what matters most for predicting collision risk.
The result is that most intersections operate with incomplete information. A cyclist approaching from behind a stopped bus, or a pedestrian stepping out from between parked cars, may be invisible to the sensor systems managing traffic signals or alerting drivers. According to federal data, intersections account for more than half of all fatal and injury crashes on American roads. The concentration is not random—intersections are where vehicles, pedestrians, and cyclists must share space and make rapid decisions based on limited information.
How the New Radar Works
The Arizona team built their sensor around millimeter wave radar, a technology already common in automotive safety systems but rarely used for detailed object classification at road crossings. The sensor emits radio waves at extremely high frequencies, then analyzes the returning signals to build a point cloud of detected objects. A machine learning component using a Gaussian mixture model sorts these point clouds into categories: vehicle, pedestrian, or clutter such as roadside vegetation or roadside furniture.
What sets this system apart is its ability to function in conditions that disable other sensors. Millimeter waves pass through rain, fog, and smoke without significant degradation. The sensor operates equally well at noon or midnight. Yet unlike simple motion-detecting radar, it can classify what it detects—not just that something is there, but what kind of road user it represents.
Our Analysis
The practical implications are substantial. Imagine an intersection where the traffic signal knows a pedestrian is approaching from behind a bus it cannot see. The signal could extend the crossing phase by a few seconds, preventing a scenario where a pedestrian steps into the crosswalk while the light is already turning green. Or consider a driver receiving a warning that a cyclist hidden in their blind spot is about to enter the intersection.
This is not science fiction. The Arizona team tested their system at an actual Tucson intersection, demonstrating that the technology works in real-world conditions rather than just controlled lab environments. The system detected pedestrians behind vehicles that blocked camera sight lines, which is precisely the scenario that causes so many real-world crashes.
From a deployment standpoint, the technology appears viable. Millimeter wave radar components are compact and relatively inexpensive, especially compared to LiDAR systems. A city looking to upgrade its most dangerous intersections could do so without the seven-figure infrastructure investments that full sensor fusion systems require.
There are limitations worth noting. The current system focuses on classifying objects into broad categories. Distinguishing between a pedestrian in a wheelchair and a person pushing a stroller, or between an adult cyclist and a child on a bicycle, would require additional development. The system also needs integration with traffic signal controllers and warning systems to actually improve safety—this sensor alone does not prevent crashes.
What This Means for You
Say you are a city transportation manager responsible for a cluster of intersections with high pedestrian crash rates. Current options are limited: more signage, pedestrian hybrid beacons, or expensive full-scale sensor upgrades. This radar system offers a middle path—an affordable way to understand who is actually using an intersection, including the users that existing loops and cameras miss. That data alone can inform better signal timing. As the technology matures, it could also power real-time warnings delivered through connected vehicle systems or smartphone apps.
Even as a regular driver or pedestrian, this development matters. Autonomous vehicle developers face the same occlusion problem—this sensor demonstrates a pathway toward vehicles that maintain awareness of hidden road users. Over time, cars that can track a cyclist who disappears behind a truck for several seconds are cars that will not make left turns into oncoming cyclists.
What to Watch Next
The Arizona Connected and Automated Vehicle lab plans to expand the system's object identification capabilities. Future versions may classify additional transportation modes and handle more complex multi-object scenarios. If these improvements succeed, the technology could eventually feed directly into connected vehicle platforms and autonomous driving systems that currently struggle with exactly these occlusion challenges.
Key takeaways
- Millimeter wave radar reliably classifies pedestrians in conditions that blind cameras and confuse conventional radar
- The technology addresses the specific occlusion problem responsible for many intersection crashes
- Compact, affordable design makes widespread deployment feasible for cities with limited budgets
Frequently asked questions
How does the sensor classify different road users without using cameras?
It uses a Gaussian mixture model to analyze radar point cloud characteristics and categorize detected objects as pedestrian, vehicle, or environmental clutter based on size, shape, and motion patterns.
Why are intersections particularly dangerous for pedestrians and cyclists?
Federal data shows over 50% of fatal and injury crashes occur at or near intersections, with two-thirds of pedestrian deaths happening at crossings where vehicles, pedestrians, and cyclists must negotiate right-of-way simultaneously.
How does this technology handle poor weather conditions?
Millimeter wave radar operates reliably in rain, fog, darkness, and smoke—conditions that significantly degrade camera performance and create false positives in conventional radar systems.
Could this system work with existing traffic infrastructure?
Yes. The design emphasizes compact size and affordability, meaning cities could deploy it at existing intersections without major infrastructure overhauls or the high costs associated with full LiDAR installations.
When might this technology appear on public roads?
The system has been tested in real-world conditions at a Tucson intersection. Researchers are working to expand object classification capabilities, with eventual integration into traffic management systems and connected vehicle platforms as a realistic near-term goal.