Enabling GLOSA for adaptive traffic lights

Green Light Optimized Speed Advisory (GLOSA) systems aim at giving ideal target speed recommendations to the driver when approaching a traffic light to lower CO₂ emissions (and fuel consumption) and to reduce the number of unnecessary stops. These systems have been shown to work well with static traffic light programs, unfortunately, a large portion of traffic lights in inner cities are adaptive and can change their behaviour with almost no lead time. This paper presents and validates (using field tests and simulation) a method to help overcome this problem and forecast fully and semi-adaptive traffic lights. First, we transformed the state graph of the traffic light controller into a transition graph focusing on signal changes and their occurrence probability. We then reduced routing possibilities within the graph using real life observations and recorded detector data of the traffic light. We further optimized our system in terms of needed storage and computationally efficiency. Our results show that in 80% of all cases we could predict signal changes 15s in the future with a high enough accuracy to enable GLOSA for adaptive traffic lights.