基于数值优化的智能车自主泊车路径规划算法

Autonomous Parking Path Planning Algorithm for Intelligent Vehicles Based on Numerical Optimization

为解决智能车在低速行驶状态下实现狭窄停车环境下的自主泊车问题,提出一种基于数值优化的自主泊车路径规划算法。首先将自主泊车路径规划问题转化为非凸优化问题,针对非凸优化问题的特点,通过引入辅助决策变量,对碰撞避免约束进行重构;然后构建泊车路径规划的最优目标模型函数;最后利用梯度下降法求解目标函数,获得最优的自主泊车路径。仿真实验结果表明:本算法目标模型函数将混合A*算法规划出的初始路径作为初始解,可以生成一条满足车辆运动学模型、无碰撞、平滑的路径,相较于混合A*算法,本算法生成的轨迹更平滑且更容易被跟踪。

In order to solve the problem of autonomous parking in narrow parking environment for intelligent vehicles at low speed,the paper proposes an autonomous parking path planning algorithm based on numerical optimization.Firstly,it transforms the problem of autonomous parking path planning into non-convex optimization problem,and reconstructs the collision avoidance constraint by introducing auxiliary decision variables according to the characteristics of non-convex optimization problem.Then,it constructs the optimal objective model function of parking path planning.Finally,it obtains the optimal autonomous parking path by using the gradient descent method solving the objective function.The simulation results show that the algorithm target model function of this paper uses the initial path planned by the hybrid A*algorithm as the initial solution,and can generate a smooth path that satisfies the vehicle kinematics model and has no collision.Compared with the hybrid A*algorithm,the trajectory generated by the algorithm in this paper is smoother and easier to be tracked.