Research On Motion Planning Method Of Intelligent Vehicle Based On Double Layer Model In Multi-obstacle Environment

With the rapid development and promotion of 5G network,driverless technology has gradually matured,and the Internet of Vehicles has also developed rapidly.Nowadays,the intelligent transportation system is becoming increasingly mature and stable.At the same time,the number of cars continues to increase,the traffic problem is increasingly serious,and under such a background,unmanned driving technology has become the current research hotspot.Path planning has always been a core and difficult problem in unmanned driving technology.When the driving environment is very complex,the difficulty of path planning will also rise sharply.Considering the influence of dynamic environment,non-holonomic constraints of vehicles and the requirements of safety,smoothness and real-time performance in autonomous driving,this paper proposes a path planning method for autonomous driving vehicles based on two-layered model,which is applied to the navigation of autonomous driving vehicles in complex multi-obstacle environments.Firstly,the advantages and disadvantages of various environmental modeling methods are compared and analyzed.In order to facilitate the design of path planning,this paper adopts the grid method to model the environment.Secondly,a two-layered model is built,including the high-level model that generates the global path and the low-level model that provides accurate navigation.The high-level model adopts BT-A* path planning algorithm,and this method optimized the point A*.On the one hand,two-way search is adopted to improve the search efficiency,and on the other hand,it inherits the advantage that the direction of pick point A* has A purpose when searching.The generated global path connects the starting point and the target point.The BT-A* algorithm expands the coding of obstacles and sets the grid around the obstacle as an infeasible area to prevent collision.The global path is secure enough.The planning speed of the high-level model is fast,and the global path obtained by the planning generally avoids the obstacles,which is safe to a certain extent.However,if the high-level model is selected as the final planning path,due to the influence of dynamic environment,the problem of frequent replanning of the high-level model will be brought,which will affect the planning efficiency.At the same time,the global path is relatively rough and does not consider the kinematics constraints,which will lead to a large error in vehicle tracking,and there are safety risks.To solve these problems,the low-level model based on model prediction is introduced to adopt the foresight strategy,and the vehicle kinematics model was introduced,the nonlinear dynamics model was linearized and the constraints were constructed.High-level model is firstly expressed in the discretization points the global path planning,due to the global path does not conform to the kinematic trajectory of the corner,so the design of new reference point on the corner,and the global reference path consistent with the kinematic trajectory,and designed with obstacle avoidance function of objective function,the function integrated weight coefficient,penalty function,and the speed,The value of the penalty function is related to the distance between the obstacle point and the vehicle center of mass,and the weight coefficient is related to the conservateness of the local path.According to the calculated discrete points,the quintic polynomial is used to fit the discrete points into a smooth real-time local path curve,which is input to the tracking module of the model prediction controller to perform the tracking function.The low-level model planning is more accurate and safe,due to the short planning distance,it is more efficient,and ensure the safety,ride comfort and comfort of the vehicle in the multi-obstacle environment.In the end,the path planning of the two-layered model is verified by Simulink and Carsim co-simulation,and verified by a real self-driving vehicle in several different driving scenarios.It is observed that the proposed method produces flexible,smooth and safe paths,enabling the vehicle to drive in complex environments.Through comparative test and statistical analysis,it is shown that this method is effective in vehicle path planning.