Decision model based on driving-mode misidentification for mixed AV–HDV straight–left conflict interactions at two-phase signalized intersections

The coexistence of autonomous vehicles (AVs) and human-driven vehicles (HDVs) has complicated the interaction between left-turning and straight-moving vehicles at intersections. Existing studies predominantly assumed the type of interacting vehicle was known, failing to account for the uncertainty in the identification of vehicle types by human drivers and their differentiated decision-making toward AVs versus HDVs. This study explores the potential impact of AVs on driver decision-making by proposing a hybrid game-based dynamic decision-making framework for human-machine mixed driving at intersections, simulating the challenges human drivers face in identifying interacting vehicle types and the interactive behaviors under different vehicle combinations in mixed-traffic flows, thereby revealing the potential influence of AVs on human decisions and mixed-traffic flows. Case analyses indicate that 1) accurate identification of AVs by human drivers can reduce average vehicle delay by 30 % and collision risk by 54.4 %, with higher interaction efficiency observed when the left-turning vehicle is an HDV; and 2) as AV penetration rates and driver recognition accuracy improve, vehicle delay and collision risk decrease significantly, with the enhancement of recognition accuracy exhibiting the most pronounced effect on intersection performance at low AV penetration rates. This study provides a novel theoretical framework for analyzing vehicle interaction mechanisms in complex mixed-traffic environments during the early stages of AV adoption, offering new theoretical foundations for addressing straight-left conflicts at intersections in mixed driving conditions.