Cycling in smart cities can be safer if enhanced with a smart traffic lights infrastructure. A distributed smartphone-based sensing approach is a cost-effective infrastructure to enable cyclist-aware traffic lights system. In this article, we treat cyclist movement on a trajectory with a Boundary model able to reduce GPS sensor power consumption, while performing time-of-arrival estimation to the nearest light. A global quantitative metric of model efficiency is proposed for assessing the overall behavior of the model, and a false-positives rating qualitative metric is used to assess the recall of the model. We evaluated the model with confined yet realistic cycling experiments and verify the precision of our model using an Android application installed in participants’ smartphones. We compared our model with previous literature, achieving a promising model for in-the-wild cycling scenarios.