Accurate Bike Routing for Lane Prediction in GLOSA Apps via Infrastructure Reference Models

Green Light Optimal Speed Advisory (GLOSA) helps drivers stop at fewer red traffic lights, reducing unnecessary energy expenditure. While GLOSA has been mainly explored for cars, GLOSA apps tailored to cyclists (bike-GLOSA) may motivate more people to travel sustainably. However, there are still technical hurdles that limit the usefulness of bike-GLOSA. One of these hurdles is lane prediction at subsequent intersections. A previously understudied but practical solution for lane prediction is calculating highly accurate routes that match cyclists' intentions. However, this solution demands more detail to accurately resolve lane topologies at intersections than available in established routing foundations such as OpenStreetMap (OSM). To solve this problem, institutionally maintained infrastructure reference models may provide the needed accuracy and quality assurance. This paper presents specific methods to integrate such a reference model into a bike-GLOSA app, at the example of Hamburg. We propose metadata from routing foundations as a practical solution for future, situationally aware GLOSA apps. Our results show that routing accuracy and lane prediction can be substantially improved with this type of routing foundation. We discuss implications for GLOSA apps and highlight future research directions.