LiDAR in Smart City Intersections to Save Pedestrians

Technology links self-driving cars to smart infrastructure in support of Vision Zero.

A visualization of a self-driving vehicle’s onboard LiDAR sensor detecting opposing traffic with laser pulses. (Source: Nvidia.)

A visualization of a self-driving vehicle’s onboard LiDAR sensor detecting opposing traffic with laser pulses. (Source: Nvidia.)

According to the U.S. Department of Transportation’s National Highway Traffic Safety Agency, a pedestrian is struck and killed by a vehicle on average once every 88 minutes. That amounts to over 100 pedestrians leaving their homes for a walk, run or bike ride every week and never making it back because they were killed in an entirely preventable accident. In 2019, the most recent full year with available data, over 6,500 pedestrians were killed. It’s simply a staggering loss of life.

The 2019 pedestrian fatality total was the highest in 30 years, exceeding the previous 30-year high from 2018 and up nearly 50 percent over the past decade. “This alarming trend signifies that we need to consider all the factors involved in this rise, identify the high-risk areas, allocate resources where they’re needed most, and continue to work with local law enforcement partners to address the chronic driver violations that contribute to pedestrian crashes,” said Governor’s Highway Safety Association Executive Director Jonathan Adkins.

There are several trends driving the troubling and steady increase in pedestrian fatalities over the last 10 years. Low fuel prices and improved efficiency have led to an increased adoption of SUVs and midsize vehicles. Pedestrians struck by an SUV are more than twice as likely to be killed than those struck by a car. There are also more people than ever congregating and clustering in large cities, where it is more common to use your own two feet as your primary mode of transportation. Lastly, there has been a dramatic increase in our smartphone addiction over the past 10 years. Distracted driving—and walking for that matter—is a major factor contributing to the increased rate of pedestrian fatalities worldwide.

The University of Nevada, Reno’s Nevada Center for Applied Research is using Velodyne’s LiDAR sensors in its Intelligent Mobility initiative to collect data aimed at making transportation more efficient, sustainable and safe. (Source: Velodyne LiDAR.)

The University of Nevada, Reno’s Nevada Center for Applied Research is using Velodyne’s LiDAR sensors in its Intelligent Mobility initiative to collect data aimed at making transportation more efficient, sustainable and safe. (Source: Velodyne LiDAR.)

Based on what’s causing the troubling trends in pedestrian fatalities—namely, human nature—it’s clear we’re facing a major challenge. Barring a wholescale rejection of SUVs and a return to flip phones, pedestrian fatalities are likely to remain at their current levels. Technology got us into this mess, but technology may also get us out of it. The Vision Zero movement, which wants to get us to zero pedestrian deaths a year, hinges on major advances in engineering and artificial intelligence.

As Elon Musk has said for a few years now, self-driving electric vehicles are the way of the future. Self-driving vehicles have been billed as the ultimate answer to making our roads safer for all users, including pedestrians. However, a car that can drive itself without running into other motor vehicles but lacks the ability to react quickly to pedestrian and bicycle movements makes us no safer than we are currently.

Full self-driving capabilities could still be another five years away, and it could be decades before the entire global fleet is transitioned away from human driving. In the meantime, the technology that will fuel the ability for cars to safely drive themselves can be deployed to help make streets safer for pedestrians. Autonomous vehicles drive themselves based on the feedback provided to their software platform and algorithm from radar, LiDAR and camera systems. These systems also have functionality that could make them extremely useful in smart, connected infrastructure systems.

Velodyne LiDAR, one of the leading companies in the race to develop a top sensor for fully autonomous vehicles, partnered with the University of Nevada, Reno to install a LiDAR sensor at the entrance to the university’s campus. Additional sensors have been installed around the city of Reno and the state of Nevada in conjunction with a statewide research initiative called Intelligent Mobility. The program aims to turn Nevada streets and roads into a lab and revolutionize transportation, mobility and smart infrastructure. If proven successful, LiDAR sensors could become an integral part of the built environment, functioning to provide constant data and feedback to a network of connected autonomous vehicles. The City of San Francisco has also implemented a pilot program to install LiDAR sensors that will collect pedestrian data.

LiDAR sensors work by transmitting a stream of laser light into the environment. When the light hits an object, it is reflected back to the sensor. As the laser scans, the sensor creates a three-dimensional rendering of the object by measuring the amount of time it takes for the reflected light to return. LiDAR systems are also able to determine specific data about the objects they are scanning and can distinguish between cars, trucks, pedestrians and bicycles. The data that LiDAR systems are able to collect and transmit may prove extremely valuable for continuing to refine the algorithms that power self-driving vehicles and smart cities. It is also important to highlight that roadside LiDAR systems have no facial recognition capability and do not violate any citizen’s privacy. This is a key differentiator for LiDAR from other technology that has been proposed as useful for smart cities.

“Unlike traditional radar or video methods for monitoring traffic, LiDAR sensors can convert vehicle data into information about vehicle and pedestrian trajectory,” Associate Professor Hao Xu explained. “The program can identify when and where speeding is occurring, for example, and it can provide a time-space diagram, showing how vehicles slow down, stop, speed up and go through an intersection during a light cycle.”

The key takeaway in Professor Xu’s statement is the type of data that LiDAR sensors are able to collect. The trajectory data, especially at intersections, is what self-driving algorithms need to function safely and reach their maximum potential. Autonomous vehicles are not entirely fail-proof at this point. They still struggle to predict and react to sudden movements at corners. For example, a fully self-driving vehicle may struggle to stop or alter its path to avoid a collision with a pedestrian who darts into traffic, such as what happened in a fatal accident with an Uber vehicle in self-driving mode back in 2018, or a vehicle attempting to speed through a yellow light.

Through the use of LiDAR sensors, several glaring holes in the current dataset that fuels self-driving algorithms can be filled. Key improvements will be driven by additional data related to near-crash analysis, connected vehicle systems, elevated traffic performance and adaptive signals, and automatic pedestrian and wildlife crossing warnings. LiDAR sensors track the trajectory of all movements within their field of vision. Near-crash analysis is particularly important, as it will allow self-driving vehicles to be programmed with countermeasures to avoid collisions based on hundreds of additional vehicles movements and maneuvers. 

State Departments of Transportation have collected crash datasets for decades, but these reports are of minimal value when it comes to building algorithms that allow a vehicle to pilot itself safely. With the added near-crash data, the technology can be refined to “teach” the machines how to react rapidly to avoid a crash or highlight potentially dangerous areas. Near-crash data is often overlooked when evaluating the safety of a stretch of road, but additional insight could enable engineers to make corrections and adjustments to improve safety.

Roadside LiDAR will also make it possible to make connected vehicles a reality. With enough LiDAR sensors mounted to existing infrastructure like traffic signals, bridges and buildings, there can be constant communication between vehicles and the sensors that will provide a continuous stream monitoring traffic conditions. Currently, most autonomous vehicles receive data solely from their own onboard sensors. Smart transportation networks, where vehicles can base their trips and path on conditions 10 miles away, will allow everyone to move from point to point as efficiently as possible.

“The Nevada Center for Applied Research is charting a path to create the smart cities of the future by enabling multimodal communication between infrastructure, vehicles and people,” said Jon Barad, vice president of Business Development at Velodyne LiDAR. “Their multidisciplinary research team is using Velodyne’s lidar sensors in innovative ways to collect and analyze data needed to improve efficiency and safety.”

Furthermore, if roadside LiDAR can be scaled up to cover the entire country, it will be possible to build predictive models for pedestrians, vehicles and bicycles. For example, will it be possible to predict, based on a vehicle’s trajectory, whether it will attempt to rush through a yellow light or how often pedestrians will attempt to cross against traffic? It may be possible to take the guesswork out of driving if enough data is collected.

LiDAR sensors will ultimately prove to play an integral role in smart cities, but most importantly, they will make it safer for pedestrians to go about their daily lives. Vision Zero is possible through the use of autonomous vehicles and increased investment in technology like LiDAR that will power smart cities. This technology is still in its infancy, but a truly synchronous, connected transportation network is within our reach.