Optimizing Cooperative Off-Ramp Diverging for Connected and Automated Vehicles

Off-ramps are critical traffic bottlenecks, and the trajectory planning of Connected and Automated Vehicles (CAVs) has the potential to address this challenge. However, optimizing CAV trajectories at off-ramps remains challenging due to complex modeling and stringent real-time requirements, particularly when coordinating CAVs across multiple lanes. To address this challenge, we develop a novel cooperative method called system optimal trajectory planning for off-ramp diverging (SOTP-OD).This method employs a rolling horizon approach to systematicallyoptimize trajectories across multiple lanes, thereby ensuring safeand efficient traffic flow across the entire stream.To efficiently solve the proposed mixed integer nonlinearprogramming (MINLP) model and meet real-time requirements, wedesign a parallel algorithm that leverages the unique features ofthe off-ramp diverging problem.This algorithm incrementally resolves the MINLP by leveraging aseries of linear programming sub-models, each corresponding to aspecific lane. These sub-models are solved simultaneously using parallelcomputing, thereby significantly enhancing computational efficiency. Experiments using NGSIM data demonstrate that the proposed method outperforms commercial solvers in both computational efficiency andsolution quality. As a result, it effectively plans safe and efficient trajectories for all CAVs in off-ramp areas. Overall, our approach provides a scalable and efficient solutionfor CAVs, significantly improving traffic safety and efficiency at off-ramps.