A major contributor to CO2 emissions in cities is traffic. City planners are always looking to reduce their carbon footprint and design efficient and sustainable infrastructure. NVIDIA Metropolis partner, MarshallAI, is helping cities improve their traffic management and reduce CO2 emissions with vision AI applications.
MarshallAI’s computer vision and AI solution helps cities get closer to carbon neutrality by making traffic management more efficient. They apply deep-learning-based artificial intelligence to video sensors to understand roadway usage, and inform and optimize traffic planning. When a city’s traffic light management system is able to adjust to real-time situations and optimize traffic flow, its increased efficiency can reduce emission-causing activities, such as frequent idling of vehicles.
As one of Finland’s fastest growing metropolitan areas, the City of Vantaa faces the challenge of quickly and safely transporting people on aging and constrained infrastructure. The city is deploying MarshallAI’s vision AI applications to optimize traffic management of intersections in real time.
The vision AI solution analyzes traffic camera streams and uses the information to adjust traffic lights according to the situation dynamically. These inputs are much richer than those from traditional sensors, capturing metrics on the amount and type of traffic users and the direction they’re driving.
MarshallAI leverages the powerful capabilities of NVIDIA Metropolis and NVIDIA GPUs. This includes the embedded NVIDIA Jetson edge AI platform—which provides GPU-accelerated computing in a compact and energy-efficient module to fuel their solution. The MarshallAI platform runs on NVIDIA EGX hardware, which brings compute to the edge by processing data from numerous cameras and providing real-time, actionable insights. MarshallAI’s traffic safety solution for the City of Vantaa automatically detects, counts, and measures the speed of vehicles, bicycles, and passerby.
“NVIDIA has made it possible for us to offer edge to cloud solutions depending on the client’s need; ranging from small, portable edge computing units to large-scale server setups. No matter the hardware constraints, the NVIDIA ecosystem enables us to run the same software stack with very little configuration changes providing optimal performance,” says Tomi Niittumäki, CTO of MarshallAI.
MarshallAI’s solution uses GPU-accelerated vision AI to process video data captured by camera sensors at traffic intersections. The system provides real-time and high-accuracy vehicle, pedestrian, and bicycle classifications, and speeds. It also tracks vehicle occupancy, paths, flows, and turning movements. These insights allow cities to react quickly to real-time situations and manage traffic effectively even during the most congested scenarios.
MarshallAI traffic management use cases
Understands traffic flow: Identifies and quantifies pedestrians, vehicles, and bicycles and detects the routes of all traffic users.
Data collection: Determines how much time traffic users spend waiting at red lights and taking unnecessary stops.
Optimizing traffic: Dynamically detects and responds to real-time traffic scenarios, eliminating unnecessary stops and idling caused by traditional time-based traffic light cycles.
Prioritizing traffic: Understands the quantity, wait time and directional velocity of vehicles on the road and can prioritize certain traffic users such as emergency vehicles.
MarshallAI’s machine vision and object detection solutions are extremely reliable. During their collaboration with the city of Vantaa, their average object detection rate was over 98% in all object classes. Different vehicle classes (car, van, bus, truck, articulated truck, and motorcycle) were treated distinctly and calculated separately.
By applying automatic traffic optimization solutions to an intersection, cities can save drivers up to every sixth traffic light stop and over a month’s worth of cumulative waiting time annually. This saves time and also reduces emissions.
MarshallAI solutions are working towards deploying across several cities such as Paris, Amsterdam, Helsinki and Tallinn, which are prioritizing the reduction of CO2 emissions and traffic congestion. The proof-of-concept installations in the Paris region and Helsinki have shown an emission reduction potential between 3% and 8% depending on the intersection, based only on optimization without any negative impact for traffic users.