RESEARCH ARTICLEContingency Overtaking Planning Based on Joint Frenet-Cartesian Constrained Iterative LQR Towards Extending FOV of Autonomous Vehicles

Motion planning for autonomous vehicles has attracted significant attention; however, autonomous overtaking by lane borrowing on two-way, two-lane national highways remains underexplored. The dense presence of heavy trucks on these highways, combined with overtaking blind spots and narrow road boundaries, presents significant challenges for motion planning systems to generate safe and feasible trajectories. The Frenét frame is widely adopted in motion planning primarily due to its efficiency in decoupling lateral and longitudinal motion and its ability to transform nonconvex road boundaries into more manageable forms. However, it exhibits notable limitations when describing the geometric dimensions of both the host vehicle and obstacles. In contrast, the Cartesian frame offers an intuitive framework for vehicle dynamics modeling yet faces significant challenges in addressing nonconvex constraints associated with curved lanes. To fully leverage the complementary advantages of these two frames, this paper proposes a joint Cartesian-Frenét Constrained Iterative Linear Quadratic Regulator (CILQR) planning method and introduces an optimization objective aimed at extending the host vehicle's front field of view (FOV). The augmented Cartesian-Frenét state space must satisfy four core constraints: (1) a Cartesian-Frenét coupled vehicle model to ensure the practical trackability of the generated trajectory; (2) a set of Frenét road constraints to provide critical safeguards for driving safety; (3) a set of Cartesian obstacle geometry constraints to ensure collision avoidance; and (4) the Cartesian-Frenét transformation relationship that serves as the central link between the two frames. The motion planning problem is then formulated and solved using CILQR to handle the aforementioned constraints. Simulation results show the capacity of the CILQR algorithm to improve planning safety by extending the FOV of autonomous vehicles.