High-performance Trajectory Optimization for Automated Parking via Half-space Constraining Theory
Trajectory planning is a vital function in vehicular automatic parking systems.Existing algorithms for automatic parking trajectory planning fail to balance generalizability,precision,time efficiency,and solution optimality.Numerical-optimization-based trajectory planning is considered in this work.Initially,the concerned planning task is formulated as a unified optimal control problem.Subsequently,a half-space constraining theory is introduced,together with a reference trajectory and a trust-region constraint modeling method,to simplify the nominal large-scale and nonconvex collision-avoidance constraints as linear inequalities.Finally,the simplified optimal control problem is solved numerically to derive an optimal parking trajectory.We name this proposed planner predefined space rapid optimization(PSRO)method.Extensive simulations indicate that PSRO outperforms prevalent trajectory optimizers such as OBCA and LIOM with respect to success rate,solution quality,and computational speed.
NETL
NSTL
万方数据
