Multi modal intelligent traffic light control optimization
Traffic congestion is one of the increasingly serious problems in cities worldwide. Due to INRIX study, an average US/German driver loses 97/120 hours per year in congestion, with the total economy damage of $87/€5.1 billion. The problem has several reasons, one of which is a suboptimal control of traffic lights. Most of the controllers nowadays operate on a phase-cycle basis and do not take into account an instantaneous traffic situation, such as the number of vehicles waiting for the green signal at the intersection and the number of vehicles having recently left the neighboring intersections.
We offer an AI-based control system Mitco for traffic signal systems (“traffic light”). Mitco’s brain is an AI agent trained with deep reinforcement learning algorithms based on historical data and self-improving based on constantly collected data. Mitco can receive traffic information from radar detectors, cameras, and GPS transmitters at taxis, car sharing, and local public transport. In the future, the integration of further sources is planned, such as Vehicle2X communication, infrared sensors, induction loops and navigation providers (Google Maps, TomTom, etc.). Mitco analyzes this anonymized information per second and gives real time recommendations on optimal traffic lights phases, leading to a significant reduction of drivers’ lost time and environmental impact.