Strategic Coalition-Based Coordination of Autonomous Vehicles Under Deceptive Behaviour at Intersections

Reliable coordination in connected and autonomous vehicle (CAV) systems depends on shared information that may be inaccurate or strategically distorted. Such uncertainty can alter perceived priorities at unsignalised intersections and degrade coordination performance. This paper develops a Bayesian game-theoretic framework to analyse multi-vehicle interactions under deceptive or unreliable information. While much prior work studies deception mainly through malicious attacks, false data injection, or unreliable sensing, this paper considers a complementary case in which misreporting is modelled as strategic uncertainty under rational decision-making. Vehicles are organised into conflict-based directional groups to maintain tractability while preserving vehicle-level outcomes. Results show that increasing deception fundamentally alters the coordination structure and induces regime-switching behaviour in optimal decisions. In a four-vehicle intersection scenario, fixed decision rules fail to adapt as deception rises, leading to growing inefficiencies or a higher risk of conflicting actions. By contrast, the proposed group-based framework maintains stable performance and substantially reduces unsafe simultaneous movements compared with independent vehicle-level decisionmaking. These results underscore the importance of deceptionaware coordination for robust and efficient intersection management under uncertainty.