Using AI for the identification, monitoring and utilization of a personalized self-learning safe route network for home-school trips

◎Research at a glance

Road users navigating dense, heterogeneous traffic environments increasingly rely on routing systems to choose their paths. Within route-planning research, safety is emerging as an important criterion, reflecting the observation that the shortest or fastest path is often not the safest. Risk perception is a crucial component in traffic safety research, as it correlates with user behavior. Furthermore, it also supports proactive safety planning by identifying potential problems in the existing infrastructure, even in the absence of actual crash data.

At its core, this research develops a routing framework capable of modelling perceived safety, predicting it across the entire network, embedding the resulting estimates in route recommendations, and adapting those suggestions to individual user preferences over repeated trips. The foundational component of this framework is a perception model that estimates perceived safety from the underlying information in the physical design of the road network. Together, these components constitute a methodology for generating routes that are both safety-aware and personalized, and that connect the subjective experience of road users to the suggested routes. The study develops and evaluates the framework for home-to-school trips across Flanders, with school children and their accompanying adults as the key user population.