How NVIDIA Metropolis Is Paving the Way Toward Smarter Traffic
Traffic. Nobody likes it, but we all have to deal with it.
As the world’s cities grow more densely populated, scientists and entrepreneurs are looking for solutions to gridlock, pollution and the other byproducts of a world filled with cars.
Two sessions at the GPU Technology Conference earlier this month spoke to the role that data, deep learning and intelligent video analytics can play in easing traffic and improving quality of life for city dwellers the world over.
The Virtuous Cycle of Traffic
Kurtis McBride, CEO of Miovision Technologies, an IVA startup based in Ontario, Canada, spoke to a room full of developers about his company’s efforts — and their 40 percent year-over-year growth — to make traffic flow a little easier.
Miovision’s Open City platform gets data from existing city infrastructure and the company’s own video cameras, and applies AI to create insights from it.
For example, the company’s Smart Intersection optimizes traffic light timing to keep city buses moving more, and sitting at red lights less. The more efficiently bus lines run, the more likely residents are to opt for them instead of cars. Fewer cars on the roads means less traffic, lower emissions and more room for those buses to run their routes efficiently.
That’s a virtuous cycle on its own, but it gets better. Miovision’s business model hinges on selling quality information to its clients. The company uses deep neural networks to analyze the raw data they get from cameras and city infrastructure. The more raw data they collect, the better they can train their networks. Well-trained networks yield better data for clients. And so it goes, another virtuous cycle.
As of GTC, Miovision had begun preliminary work with the NVIDIA Metropolis platform for analyzing video streams, and in particular are excited about how the Jetson TX2 AI supercomputer on a module will aid their work. The company is planning small-scale trials of Open City running on the Jetson TX2 in cities across North America. And the TX2’s improved energy efficiency has them looking at future solutions that could run entirely on the sun’s rays.
“With TX2, we’re within striking distance of being solar-powered,” McBride said, citing Jetson TX2’s ability to run on only 7 watts of power. “We’re probably a generation or two away from realizing that, but when we do, solar will bring deployment costs way down for municipalities.”
Cycling the Green Waves
Intelligent traffic flows aren’t just for cars and buses. Economist, mathematician and computer scientist Edward Zimmerman spoke to GTC attendees about his ongoing work using deep learning to create “green waves” for bicyclists in Germany, and beyond.
A green wave is that wondrous phenomenon of cruising through one green light after another, as though the traffic gods were smiling directly upon you during your commute. As Zimmerman explained, green waves are more science than fiction, resulting from timing patterns and algorithms often devised to give priority to cars and buses over cyclists and pedestrians in high-traffic areas.
Self-professed “data guy” Zimmerman is working with GESIG, a Germany-based maker of signaling equipment, to develop low-cost systems that create on-demand green waves for urban cyclists. The project aims to use the GPU power and energy efficiency of the Jetson TX1 platform to identify cyclists through neural networks fed real-time data from cameras installed in traffic lights. The networks then analyze the data to identify opportunities to bestow green waves upon the bicycle riders — ideally doing so in harmony with mass transit vehicles already riding their own waves.
Like Miovision’s McBride, Zimmerman sees the NVIDIA Metropolis platform and Jetson TX2’s improved energy efficiency as a step towards solar-powered intelligence — in this case, in the shape of smart traffic lights powered entirely by the sun’s rays. A variant of the project was tested in the city of Bonn, Germany, with plans in the works to spread the green waves across Europe and beyond.