Hierarchical Motion Planning And Control Of Intelligent Vehicles

As an important part of the autonomous driving system,trajectory planning,and trajectory tracking modules play a vital role in the safe self-driving of intelligent vehicles.The surround-ing environments of autonomous vehicles are changeable and the road conditions are complex.Therefore,the autonomous vehicle system has high requirements for trajectory planning calcula-tion stability.Also,the vehicle motion system is a nonlinear system,and the state variables change rapidly and frequently when driving at a high speed.Therefore,the real-time performance of the vehicle controller is highly required.In this paper,a hierarchical vehicle motion planning control algorithm is proposed as follow:1.In the motion planning,considering the direct trajectory calculation has a high dimension-ality and a large computational complexity,this article decomposes the motion planning problem into two problems of path planning and speed planning,respectively.In the pro-cess of path optimization,cost functions are selected appropriately,including the tracking cost function of the desired trajectory,the cost function of the control input,and the obstacle avoidance cost function.In the longitudinal velocity planning,terminal sampling method is used.The idea is to sample the terminal time and the terminal longitudinal motion state sep-arately according to different longitudinal motion modes.Then,the candidate longitudinal motion trajectories are generated by solving the boundary value problem.Finally,the can-didate trajectories are evaluated and an optimal longitudinal velocity trajectory is selected with the lowest cost function value.2.In the lower-level motion control,the trajectory tracking controller is divided into a longitu-dinal speed tracking and a lateral position tracking model predictive controller.In the design of the speed tracking controller,the speed tracking error is used to design the PID feedback control algorithm.In the design of the lateral position tracking controller,following the process of the model predictive controller design,the vehicle dynamics model is first estab-lished according to the Newton’s second law of motion,and the discrete prediction equation is obtained.Then,safety constraints and control input constraints are calculated.Next,the optimization problem is transformed into the standard form of quadratic programming.Fi-nally,in order to address the problem of model mismatch and uncertain disturbance,a linear quadratic regulator feedback correction law is calculated based on error system.Based on the state-of-art study in the field of motion planning and control,this dissertation conducts simulation comparisons with other works.The main contributions are summarized as follows:1.Based on the idea of model predictive control,a path planning algorithm considering vehi-cle kinematics is designed to meet the smoothness requirement of vehicle kinematics con-straints.2.By establishing a curve-based coordinate system and the longitudinal projection of mov-ing obstacles,the dimensionality of speed planning is reduced and calculation efficiency is improved.3.A LQR feedback correction control strategy based on the error system is designed,which improves the robustness of the lateral controller and effectively reduces lateral tracking er-rors under unknown disturbance.