Research On Trajectory Planning And Tracking Control Algorithm For Autonomous Vehicle On Structured Road

Autonomous driving technology is significant to improving travel efficiency,ensuring driving safety,and building intelligent transportation systems.Structured road scenes are currently important environment for developing autonomous driving technology.Trajectory planning technology and tracking control technology are the core parts of the automatic driving system.However,there are still shortcomings in applying the existing technology to structured road scenes.For trajectory planning technology,the completeness and quality of trajectory decision-making need to be improved,and trajectory optimization has problems such as high dimensionality and large complexity;for tracking control technology,there are problems such as difficulty in describing the nonlinear dynamics model of the vehicle,complicated multi-state variable constraints,and low solution efficiency.The trajectory planning module and the tracking control module are closely connected and have high functional coupling.Both are the core modules in automatic driving technology.Research on safe,reliable,and high-real-time algorithms is of great significance and engineering value.For this reason,this paper takes the trajectory planning algorithm and tracking control algorithm of structured roads as the research objects.The main research contents are as follows:(1)In order to construct the Frenet coordinate system accurately and efficiently in the structured road scene,design the reference line optimization algorithm based on the segmented quintic spline curve and the matching point search algorithm based on the two-stage method.The high-discrete,low-precision original reference line position data is segmented and the quintic spline curve is used for fitting,and the optimal smooth reference line is optimized.Convertting the state quantity of Cartesian coordinates to the Frenet coordinate system based on the reference line.For the search for the matching point of the reference line,a two-stage search algorithm combining the second-order minimization method and the Newton method is proposed,which is fast in the first stage Shrink the range of matching points,and perform accurate calculation to solve the matching points in the second stage,reducing the computational time for constructing the Frenet coordinate system.Matlab programming simulation verifies that the smoothing proposed has good smoothness;verifies that coordinate conversion algorithm has accuracy and efficiency.(2)Design a trajectory planning algorithm combining sampling search and numerical optimization based on the Frenet coordinate system.In order to ensure the completeness of trajectory planning propositions and the rapidity of optimization solutions,a layered trajectory planning framework of collaborative decision-making and decoupling optimization is proposed.The upper-level constructs the T-S-L spacetime solution space and constructs a sampling search method based on the dynamic programming algorithm,The solution obtains the decision trajectory containing the vehicle rough path information and speed information;the lower-level decouples the trajectory optimization into path optimization and speed optimization,adopts the tunneling idea to pave the convex optimization space with the decision trajectory as the baseline,and proposes non-uniformity The discrete form reasonably allocates the computational power of numerical optimization,and adopts Taylor’s expansion idea to establish precise obstacle boundary collision constraints to ensure the accuracy of collision detection in the discrete space,ensure the safety of obstacle avoidance,and reduce the scale of optimization propositions.Comprehensive consideration of driving safety,driving comfort,vehicle kinematics and traffic rules and other factors to construct a path and speed optimization model in the form of secondary planning,reduce the optimization parameters and constraints of the overall model,and reduce the difficulty of solving.It is verified by Matlab programming simulation that the trajectory planning algorithm can plan a safe and comfortable path and speed curve,and has a better solution speed.(3)Design a tracking control algorithm that decouples the longitudinal and lateral control to match the path and speed information.The constructed longitudinal controller has a layered structure.The upper layer uses PID control to obtain the desired acceleration,and the lower layer formulates the acceleration/deceleration switching logic,and calculates the throttle and the brake cylinder pressure based on the dynamic inverse model;The constructed lateral controller is a model predictive controller that adopts the vehicle dynamics model,reasonably simplified the vehicle dynamics model to obtain the linear predictive model,set the control quantity constraint and the state quantity constraint,and design the optimal control model in the form of quadratic programming.In order to improve the convergence speed of optimization,using the characteristics of model prediction rolling optimization,the vector obtained by historical optimization is data-expanded and used as the initial solution of the next optimization to achieve the effect of hot start.Finally,verifies that the trajectory tracking control algorithm proposed has good real-time performance and accuracy.(4)Establish an autonomous driving algorithm test platform based on Carsim and Matlab/Simulink,using real vehicle dynamics parameters to complete the parameter configuration of the vehicle model,build a structured road scene close to reality,and design overtaking,following,and corner avoidance.In three working conditions,experiments and results analysis of the trajectory planning algorithm and tracking control algorithm are carried out.The experimental results show that the reference line smoothing algorithm proposed can effectively smooth the reference line;the proposed trajectory planning algorithm can plan a reasonable,feasible,safe and comfortable motion trajectory,with high solving efficiency;the proposed tracking control algorithm can be effective and accurate Track trajectory in real time,and meets the system requirements.