Traffic management has relied on time-interval-based phases and simple induction loops for decades. MarshallAI, in collaboration with Dynniq Finland Oy, set out to prove that deep learning and visual sensing could take traffic management to a fully autonomous, optimized level—reducing CO2 emissions and eliminating bottlenecks without compromising safety.
Through the three-year, EU-funded AI4Cities project, MarshallAI deployed the Intelligent Interconnected Intersections (Ix3) system across multiple European cities. The goal was to drastically improve traffic flow and reduce emissions by replacing outdated sensors with complete, real-time situational awareness.
The Objectives: Flow Analysis & Prioritization
To deliver a scalable, intelligent flow solution, the project focused on:
- Comprehensive Flow Analysis: Understanding traffic volumes, waiting times, directions, and specific vehicle types to build a complete picture of the intersection’s demand.
- Traffic Optimization: Eliminating “dead seconds” by dynamically transitioning lights to remove unnecessary stops and idling when lanes are empty.
- Traffic Prioritization: Enabling the system to actively promote specific traffic users (such as public transport or heavy vehicles) based on flow efficiency and emission reduction goals.
The Solution: Autonomous Logic via Edge & Cloud
The Ix3 solution utilizes standard IP cameras to feed visual data to the MarshallAI machine vision platform. This data is processed in real-time and integrated directly with Dynniq’s traffic-light controllers to actuate immediate physical changes based on current demand.
Depending on the specific pilot phase, processing was successfully deployed via secure cloud servers (Helsinki) or directly on edge devices delivered by MarshallAI (Paris region).
Proven Results: Massive Reductions in Waste
The pilots proved that MarshallAI’s system can autonomously manage high-volume flows and eliminate systemic waste. Across three distinct phases, the system delivered remarkable efficiency gains:
- Phase 1 (Kalasatama, Helsinki): The system eliminated 162,944 car stops, 17,822 heavy vehicle stops, and 654 hours of irrational waiting time annually, equating to a 4% total emission reduction for the intersection.
- Phase 2 (Helsinki & Paris region): Expanding to a cluster of intersections in Suutarila achieved up to an 8% emission reduction potential. Concurrently, an edge-processed pilot in Saint-Maur-des-Fossés achieved a 3% reduction.
- Phase 3 (Meudon & Helsinki): Live pilots validated long-term viability with an average annualized reduction of 55,509 vehicle stops and 360 hours of irrational waiting time, resulting in a highly consistent 2% reduction in total baseline emissions.
Non-Intrusive, Turnkey Deployment
Deploying advanced AI to control critical infrastructure doesn’t have to be a sprawling IT project. The Ix3 pilots proved that MarshallAI can be deployed as a turnkey solution with near-zero disruption to existing operations. Customer involvement was limited strictly to identifying the target intersection, arranging basic installation permissions, and coordinating traffic controller programming with the team.
