This paper is focused on planning fast, accurate, and optimal trajectories for autonomous parking. Nominally, this task should be described as an optimal control problem (OCP), wherein the collision-avoidance constraints guarantee travel safety and the kinematic constraints guarantee tracking accuracy. The dimension of the nominal OCP is high because it requires the vehicle to avoid collision with each obstacle at every moment throughout the entire parking process. With a coarse trajectory guiding a homotopic route, the intractably scaled collision-avoidance constraints are replaced by within-corridor constraints, whose scale is small and independent from the environment complexity. Constructing such a corridor sacrifices partial free spaces, which may cause loss of optimality or even feasibility. To address this issue, our proposed method reconstructs the corridor in an iterative framework, where a lightweight OCP with only box constraints is quickly solved in each iteration. The proposed planner, together with several prevalent optimization-based planners are tested under 115 simulation cases w.r.t. the success rate and computational time. Real-world indoor experiments are conducted as well.
Curvy roads are a particular type of urban road scenario, wherein the curvature of the road centerline changes drastically. This paper is focused on the trajectory planning task for autonomous driving on a curvy road. The prevalent on-road trajectory planners in the Frenet frame cannot impose accurate restrictions on the trajectory curvature, thus easily making the resultant trajectories beyond the ego vehicle's kinematic capability. Regarding planning in the Cartesian frame, selection-based methods suffer from the curse of dimensionality. By contrast, optimization-based methods in the Cartesian frame are more flexible to find optima in the continuous solution space, but the new challenges are how to tackle the intractable collision-avoidance constraints and nonconvex kinematic constraints. An iterative computation framework is proposed to accumulatively handle the complex constraints. Concretely, an intermediate problem is solved in each iteration, which contains linear and tractably scaled collision-avoidance constraints and softened kinematic constraints. Compared with the existing optimization-based planners, our proposal is less sensitive to the initial guess especially when it is not kinematically feasible. The efficiency of the proposed planner is validated by both simulations and realworld experiments. Source codes of this work are available at https://github.com/libai1943/CartesianPlanner.
Automated parking is a typical function in a self-driving car. The trajectory planning module directly reflects the intelligence level of an automated parking system. Although many competitions have been launched for autonomous driving, most of them focused on on-road driving scenarios. However, driving on a structured road greatly differs from parking in an unstructured environment. In addition, previous competitions typically competed on the overall driving performance instead of the trajectory planning performance. A trajectory planning competition of automated parking (TPCAP) has been recently organized. This event competed on parking-oriented planners without involving other modules, such as localization, perception, or tracking control. This study reports the TPCAP benchmarks, achievements, experiences, and future perspectives.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.