Autonomous Driving on Curvy Roads Without Reliance on Frenet Frame: A Cartesian-Based Trajectory Planning Method

Curvy roads are a particular type of urban road scenario, wherein the curvature centerline changes drastically. This paper is focused on trajectory planning task for autonomous driving curvy road. The prevalent on-road planners in Frenet frame cannot impose accurate restrictions curvature, thus easily making resultant trajectories beyond ego vehicle’s kinematic capability. Regarding Cartesian frame, selection-based methods suffer from curse dimensionality. By contrast, optimization-based more flexible to find optima continuous solution space, but new challenges how tackle intractable collision-avoidance constraints and nonconvex constraints. An iterative computation framework proposed accumulatively handle complex Concretely, an intermediate problem solved each iteration, which contains linear tractably scaled softened Compared with existing planners, our proposal less sensitive initial guess especially when it not kinematically feasible. efficiency planner validated by both simulations real-world experiments. Source codes this work available at https://github.com/libai1943/CartesianPlanner .