| As an indispensable means of transportation in people’s life,automobile updates iteration speed is increasing year by year.Since the automobile performance and shape research has been basically perfected and there are few places to improve and innovate,most automobile companies are now turning to automotive automation research,auto-driving of automobiles may become the next air outlet in the automotive industry.Automatic parking is a key link of automatic driving,and it can be used as a breakthrough point in the research of automatic driving technology.The most important problem in parking process is to complete the motion planning.Therefore,this thesis studies the motion planning of automatic parking,mainly including the following:（1）Complete the preparation for parking route planning and optimization.Firstly,the kinematics model of the vehicle is established based on the Ackermann turning principle,and the kinematics expression of the vehicle and the four-point coordinates of the body equivalent to a rectangle are obtained as its motion constraints.Secondly,the obstacle avoidance constraints of the vehicle body are analyzed for two common parking scenarios,vertical parking and parallel parking.Two easily colliding points exist for vertical parking.For this point as the collision threshold,the boundary collision conditions are calculated.Parallel parking has more collision points and all three sides except the front of the body contain collision points,so the boundary collision conditions of all sides of the body are considered.Finally,the advantages and disadvantages of environment modeling method,route search strategy and route optimization strategy are analyzed,which lays a foundation for the following simulation experiments,parking route planning,and further optimization of parking route.（2）Planning parking paths based on improved Hybrid A* algorithm.First,a dynamic cost function is designed to adjust the number of search nodes according to the actual parking environment,which reduces the evaluation calculation of some unnecessary nodes.Secondly,the Reeds-Shepp curve is introduced to increase the direction of node expansion backward.To avoid multiple shifts during parking,penalty costs for turning around are also considered.To overcome the disadvantage of not taking obstacles into account when expanding nodes in Reeds-Shepp curves,a selective expansion scheme considering obstacles is proposed,which can not only speed up the Hybrid A* algorithm in path planning,but also ensure the generation of effective parking paths.Finally,the validity of the proposed improved Hybrid A* algorithm for parking route planning is verified by simulation experiments.（3）Optimizing parking paths based on improved B-spline curves.Firstly,the definition,properties and classification of B-spline curves are introduced in detail,which paves the way for selecting cubic quasi-uniform B-spline curves for path optimization later.Secondly,an optimal parking route strategy based on cubic quasi-uniform B-spline curve interpolation is proposed,and it is improved by adding control points to make the optimized path more similar to the original one.Finally,the validity of the proposed improved B-spline curve to optimize parking routes is verified by simulation experiments.（4）The simulation experiments of vertical parking and parallel parking are carried out based on the ROS platform.Firstly,the function and feature of ROS platform are briefly introduced,the data of map creation using raster method and the parameters of vehicle simulation are explained.Secondly,two groups of experiments are designed for vertical parking and parallel parking scenarios.The first group of experiments verifies the validity of this method for both scenarios.The second group of experiments proves that this method can complete parking efficiently and reliably in many cases.Finally,the experiment in this chapter is briefly summarized. |
