Vous êtes ici : Accueil > Université > Composantes > Faculté d’Informatique > Séminaires IRIT UT Capitole
- Recherche,
Who detours and who gets stuck? Heterogeneous behavioural adaptation under traffic disruptions
le 9 juin 2026
12h45
Manufacture des Tabacs
MF103
MF103
Yihang Bai, School of Geography, University of Leeds
Abstract: How do drivers actually respond when a road is unexpectedly closed? The challenge is that travel behaviour observed under such non recurrent disruption reflects two distinct forces at once: some drivers actively adapt, while others are passively exposed to conditions they never chose. Most impact assessments cannot separate the two, because the behaviour of the same driver on a normal day is never observed. Existing work approximates this counterfactual with shortest paths or aggregate flows, which fail to capture individual routing preferences. It also largely overlooks a second constraint, namely that a route which is feasible in the network may not be behaviourally accessible once cognitive and navigational complexity are accounted for.
This talk presents a counterfactual, behaviourally grounded approach in which each trip's habitual expectation is reconstructed from individual route choice models combined with agent based simulation, so that observed behaviour can be revealed against that baseline. The approach is illustrated through a recent study of non recurrent traffic disruptions in the West Midlands, UK, using large scale GPS trajectories and a calibrated MATSim model of the region. Four behavioural archetypes emerge from the data, capturing how drivers differ in congestion exposure, route diversion and adaptation effectiveness. A central finding is that navigational complexity, although never used to define the groups, systematically separates them after the fact, suggesting that it may act as a hidden filter on who can reroute effectively. Building on these results, the talk discusses implications for network resilience and outlines how multi-agent simulation can be extended to represent heterogeneous driver responses under disruption.
This talk presents a counterfactual, behaviourally grounded approach in which each trip's habitual expectation is reconstructed from individual route choice models combined with agent based simulation, so that observed behaviour can be revealed against that baseline. The approach is illustrated through a recent study of non recurrent traffic disruptions in the West Midlands, UK, using large scale GPS trajectories and a calibrated MATSim model of the region. Four behavioural archetypes emerge from the data, capturing how drivers differ in congestion exposure, route diversion and adaptation effectiveness. A central finding is that navigational complexity, although never used to define the groups, systematically separates them after the fact, suggesting that it may act as a hidden filter on who can reroute effectively. Building on these results, the talk discusses implications for network resilience and outlines how multi-agent simulation can be extended to represent heterogeneous driver responses under disruption.
Mis à jour le 1 juin 2026
