Tech to make biking safer with AI-powered alerts gets $200k in federal funding

In a world increasingly focused on smart cities and autonomous vehicles, one Pittsburgh startup channels technology into an increasingly important area: bicycle safety. 

When Clark Haynes, CEO and founder of Velo AI, took up biking more regularly during the pandemic, he noticed just how perilous roads can be for cyclists. This realization spurred him to create Velo AI, a start-up dedicated to enhancing cycling safety through innovative technology. 

“I’ve been a bike commuter my entire career,” Haynes told Technical.ly. “Every decision — where to live, what job to take — was centered around whether I could bike to work. But despite all the advancements in tech, our cities still don’t feel safe for people on two wheels.”

With a recent $200,000 grant from the US Department of Transportation’s Complete Streets AI program, Velo AI now sets its sights on how data can provide real-time alerts to cyclists facing impending danger. 

Velo AI’s flagship product, Copilot, is a direct result of this vision. The device, which attaches to the back of a bike, constantly records and monitors the environment around the cyclist. It alerts the rider in real-time through a smartphone app, flashing lights or sound signals whenever a vehicle approaches too closely. 

“It’s like having eyes in the back of your head,” Haynes said. 

Bike commuting nearly doubled in Pittsburgh between 2010 and 2019, making it the 16th most popular city for bike commuters, according to a 2022 report from BikePGH. A person on a bike is involved in a crash on the road every six days, the report found.

Collecting data leads to predicting areas for improvement

Velo AI’s Copilot device doesn’t just warn cyclists of immediate dangers — it also collects valuable data about near-misses, dangerous intersections and more.

Using everyday bike rides to gather critical data caught the attention of federal officials. The goal of the DOT sponsorship is to deploy Velo AI’s technology at a larger, more inclusive scale, particularly through partnerships with organizations like Pittsburgh’s bike share program, POGOH. That should expand the presence beyond just the 20 to 30 cyclists testing Copilot in Pittsburgh as of 2023. 

Velo AI’s team will install a simplified version of Copilot in the front baskets of POGOH bikes. Unlike the consumer version, which monitors the environment 360 degrees, this iteration focuses on what’s ahead. Haynes explained that they are particularly interested in choke points, or places where bikes and vehicles are most likely to clash.

“With POGOH, we can equip e-bikes with our technology and gather data from a much broader demographic,” said Haynes. “These bikes go everywhere, from upscale neighborhoods to underserved areas, and that diversity of data is critical for us.”

With every mile biked, Copilot gathers information that feeds into a larger system designed to help city planners, transportation departments, and mobility advocates understand where and why cyclists are at risk. 

The data collected will be analyzed in partnership with Carnegie Mellon University’s (CMU) Mobility Data Analytics Center (MAC) — the Velo AI founders are, in fact, CMU alumni. Then, it’ll be used to create predictive tools and advanced statistical models that can identify high-risk areas before accidents happen, essentially allowing cities to be proactive rather than reactive, according to Haynes. 

MAC plans to combine Velo’s data with other traffic information, like satellite images, aerial photos, traffic counts and road layout, to create a web-based data platform for visualization and analysis.

“This platform will estimate risks for vulnerable road users,” said Sean Qian, director of MAC, “and predict how changes, such as adding bike lanes or lowering speed limits, could affect those risks.”

Velo AI has built a robust coalition of support for this project. The company is working with local and national groups, including the Ehrlichman Group, BikePGH and the League of American Bicyclists. 

According to the City of Pittsburgh Crash Data Dashboard, 56 car accidents involved a bicycle in 2023, including one fatality. Since it launched in 2019, the number of bike accidents totaled 197.

That’s why the mission, stakeholders say, is truly a matter of life and death. 

“A cyclist should not have to die in order to provide a data point that high-risk safety conflicts exist in specific areas,” said Courtney Ehrlichman, founder and CEO of the transit strategy firm Ehrlichman Group.

A background in autonomous vehicles informs Velo’s approach

Haynes’s journey to Velo AI didn’t start with a business plan. After working on humanoid robots at Carnegie Mellon and later contributing to Uber’s Advanced Technologies Group (ATG) in Pittsburgh, Haynes was well-versed in the potential of autonomous systems.

At Uber, Haynes focused on how self-driving cars could navigate complex urban environments. He was involved in advanced robotics and AI work, all aimed at creating self-driving cars. 

Haynes worked as the prediction team lead, developing methods to predict how real-world road users interact with self-driving vehicles, drawing heavily from machine learning, data science, robotics, and software engineering.

This experience provided him with a deep understanding of transportation systems and their potential vulnerabilities. 

Over time, it became clear to Haynes that the future of safe streets might not solely lie in fully autonomous vehicles. He realized that much more could be achieved with simpler ideas, which led to the founding of Velo AI in 2021.

“Autonomous vehicles are incredible, but they’re just one piece of the puzzle,” Haynes said. “What we’re doing at Velo AI is taking some of that same technology — machine learning, computer vision, sensor fusion — and applying it in a way that’s more immediately impactful.”