Robust Lane Assignment and Continuous Control for a Freeway CAV Priority Lane During Rule Transitions

Freeway connected and automated vehicle (CAV) priority lanes (CPLs) provide dedicated or preferential road space for CAVs, but access-rule transitions require CAVs, connected vehicles (CVs), human-driven vehicles (HDVs), and trucks to reorganize within a finite control zone. This paper develops a three-layer framework that links robust target-lane assignment, online executability screening, and continuous control through a common state and surrogate-safety-measure interface. A six-slot robust mixed-integer program supplies a feasible incumbent, robustness-enhanced beam search retains alternative lane sequences, and local integer-programming repair corrects restricted conflicts. The middle layer checks retained assignments using gap time, time to collision, deceleration rate to avoid a collision, and post-encroachment time. The lower layer uses extended-Kalman-filter-assisted control-barrier-function model predictive control (EKF-CBF-MPC). Across 180 matched closed-loop scenarios per upper-layer method, the complete framework achieves 98.2% target-lane completion and 100.0% planned-PET pass, compared with 89.5% and 49.8% for direct six-slot robust MIP. Mean access-rule satisfaction is 65.4% and 67.9%, respectively. In 100 matched parameter-mismatch runs, EKF initialized from the present HDV prior reduces one-second speed and net-gap prediction RMSE by 74.5% and 79.6% relative to the matching fixed-parameter predictor; a published highD prior gives similar online errors. Communication-fault experiments further evaluate state prediction, command expiry, lane-change deferral, and local fallback control. The results show improved execution reliability and conflict screening, with a modest rule-compliance and computation-time trade-off relative to robust MIP.